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/fs_context.h> 12 #include <linux/sched/mm.h> 13 #include <linux/statfs.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 #include <linux/ctype.h> 31 #include <linux/fs_parser.h> 32 33 #include "f2fs.h" 34 #include "node.h" 35 #include "segment.h" 36 #include "xattr.h" 37 #include "gc.h" 38 #include "iostat.h" 39 40 #define CREATE_TRACE_POINTS 41 #include <trace/events/f2fs.h> 42 43 static struct kmem_cache *f2fs_inode_cachep; 44 45 #ifdef CONFIG_F2FS_FAULT_INJECTION 46 47 const char *f2fs_fault_name[FAULT_MAX] = { 48 [FAULT_KMALLOC] = "kmalloc", 49 [FAULT_KVMALLOC] = "kvmalloc", 50 [FAULT_PAGE_ALLOC] = "page alloc", 51 [FAULT_PAGE_GET] = "page get", 52 [FAULT_ALLOC_BIO] = "alloc bio(obsolete)", 53 [FAULT_ALLOC_NID] = "alloc nid", 54 [FAULT_ORPHAN] = "orphan", 55 [FAULT_BLOCK] = "no more block", 56 [FAULT_DIR_DEPTH] = "too big dir depth", 57 [FAULT_EVICT_INODE] = "evict_inode fail", 58 [FAULT_TRUNCATE] = "truncate fail", 59 [FAULT_READ_IO] = "read IO error", 60 [FAULT_CHECKPOINT] = "checkpoint error", 61 [FAULT_DISCARD] = "discard error", 62 [FAULT_WRITE_IO] = "write IO error", 63 [FAULT_SLAB_ALLOC] = "slab alloc", 64 [FAULT_DQUOT_INIT] = "dquot initialize", 65 [FAULT_LOCK_OP] = "lock_op", 66 [FAULT_BLKADDR_VALIDITY] = "invalid blkaddr", 67 [FAULT_BLKADDR_CONSISTENCE] = "inconsistent blkaddr", 68 [FAULT_NO_SEGMENT] = "no free segment", 69 [FAULT_INCONSISTENT_FOOTER] = "inconsistent footer", 70 [FAULT_TIMEOUT] = "timeout", 71 [FAULT_VMALLOC] = "vmalloc", 72 }; 73 74 int f2fs_build_fault_attr(struct f2fs_sb_info *sbi, unsigned long rate, 75 unsigned long type, enum fault_option fo) 76 { 77 struct f2fs_fault_info *ffi = &F2FS_OPTION(sbi).fault_info; 78 79 if (fo & FAULT_ALL) { 80 memset(ffi, 0, sizeof(struct f2fs_fault_info)); 81 return 0; 82 } 83 84 if (fo & FAULT_RATE) { 85 if (rate > INT_MAX) 86 return -EINVAL; 87 atomic_set(&ffi->inject_ops, 0); 88 ffi->inject_rate = (int)rate; 89 f2fs_info(sbi, "build fault injection rate: %lu", rate); 90 } 91 92 if (fo & FAULT_TYPE) { 93 if (type >= BIT(FAULT_MAX)) 94 return -EINVAL; 95 ffi->inject_type = (unsigned int)type; 96 f2fs_info(sbi, "build fault injection type: 0x%lx", type); 97 } 98 99 return 0; 100 } 101 #endif 102 103 /* f2fs-wide shrinker description */ 104 static struct shrinker *f2fs_shrinker_info; 105 106 static int __init f2fs_init_shrinker(void) 107 { 108 f2fs_shrinker_info = shrinker_alloc(0, "f2fs-shrinker"); 109 if (!f2fs_shrinker_info) 110 return -ENOMEM; 111 112 f2fs_shrinker_info->count_objects = f2fs_shrink_count; 113 f2fs_shrinker_info->scan_objects = f2fs_shrink_scan; 114 115 shrinker_register(f2fs_shrinker_info); 116 117 return 0; 118 } 119 120 static void f2fs_exit_shrinker(void) 121 { 122 shrinker_free(f2fs_shrinker_info); 123 } 124 125 enum { 126 Opt_gc_background, 127 Opt_disable_roll_forward, 128 Opt_norecovery, 129 Opt_discard, 130 Opt_noheap, 131 Opt_heap, 132 Opt_user_xattr, 133 Opt_acl, 134 Opt_active_logs, 135 Opt_disable_ext_identify, 136 Opt_inline_xattr, 137 Opt_inline_xattr_size, 138 Opt_inline_data, 139 Opt_inline_dentry, 140 Opt_flush_merge, 141 Opt_barrier, 142 Opt_fastboot, 143 Opt_extent_cache, 144 Opt_data_flush, 145 Opt_reserve_root, 146 Opt_reserve_node, 147 Opt_resgid, 148 Opt_resuid, 149 Opt_mode, 150 Opt_fault_injection, 151 Opt_fault_type, 152 Opt_lazytime, 153 Opt_quota, 154 Opt_usrquota, 155 Opt_grpquota, 156 Opt_prjquota, 157 Opt_usrjquota, 158 Opt_grpjquota, 159 Opt_prjjquota, 160 Opt_alloc, 161 Opt_fsync, 162 Opt_test_dummy_encryption, 163 Opt_inlinecrypt, 164 Opt_checkpoint_disable, 165 Opt_checkpoint_disable_cap, 166 Opt_checkpoint_disable_cap_perc, 167 Opt_checkpoint_enable, 168 Opt_checkpoint_merge, 169 Opt_compress_algorithm, 170 Opt_compress_log_size, 171 Opt_nocompress_extension, 172 Opt_compress_extension, 173 Opt_compress_chksum, 174 Opt_compress_mode, 175 Opt_compress_cache, 176 Opt_atgc, 177 Opt_gc_merge, 178 Opt_discard_unit, 179 Opt_memory_mode, 180 Opt_age_extent_cache, 181 Opt_errors, 182 Opt_nat_bits, 183 Opt_jqfmt, 184 Opt_checkpoint, 185 Opt_lookup_mode, 186 Opt_err, 187 }; 188 189 static const struct constant_table f2fs_param_background_gc[] = { 190 {"on", BGGC_MODE_ON}, 191 {"off", BGGC_MODE_OFF}, 192 {"sync", BGGC_MODE_SYNC}, 193 {} 194 }; 195 196 static const struct constant_table f2fs_param_mode[] = { 197 {"adaptive", FS_MODE_ADAPTIVE}, 198 {"lfs", FS_MODE_LFS}, 199 {"fragment:segment", FS_MODE_FRAGMENT_SEG}, 200 {"fragment:block", FS_MODE_FRAGMENT_BLK}, 201 {} 202 }; 203 204 static const struct constant_table f2fs_param_jqfmt[] = { 205 {"vfsold", QFMT_VFS_OLD}, 206 {"vfsv0", QFMT_VFS_V0}, 207 {"vfsv1", QFMT_VFS_V1}, 208 {} 209 }; 210 211 static const struct constant_table f2fs_param_alloc_mode[] = { 212 {"default", ALLOC_MODE_DEFAULT}, 213 {"reuse", ALLOC_MODE_REUSE}, 214 {} 215 }; 216 static const struct constant_table f2fs_param_fsync_mode[] = { 217 {"posix", FSYNC_MODE_POSIX}, 218 {"strict", FSYNC_MODE_STRICT}, 219 {"nobarrier", FSYNC_MODE_NOBARRIER}, 220 {} 221 }; 222 223 static const struct constant_table f2fs_param_compress_mode[] = { 224 {"fs", COMPR_MODE_FS}, 225 {"user", COMPR_MODE_USER}, 226 {} 227 }; 228 229 static const struct constant_table f2fs_param_discard_unit[] = { 230 {"block", DISCARD_UNIT_BLOCK}, 231 {"segment", DISCARD_UNIT_SEGMENT}, 232 {"section", DISCARD_UNIT_SECTION}, 233 {} 234 }; 235 236 static const struct constant_table f2fs_param_memory_mode[] = { 237 {"normal", MEMORY_MODE_NORMAL}, 238 {"low", MEMORY_MODE_LOW}, 239 {} 240 }; 241 242 static const struct constant_table f2fs_param_errors[] = { 243 {"remount-ro", MOUNT_ERRORS_READONLY}, 244 {"continue", MOUNT_ERRORS_CONTINUE}, 245 {"panic", MOUNT_ERRORS_PANIC}, 246 {} 247 }; 248 249 static const struct constant_table f2fs_param_lookup_mode[] = { 250 {"perf", LOOKUP_PERF}, 251 {"compat", LOOKUP_COMPAT}, 252 {"auto", LOOKUP_AUTO}, 253 {} 254 }; 255 256 static const struct fs_parameter_spec f2fs_param_specs[] = { 257 fsparam_enum("background_gc", Opt_gc_background, f2fs_param_background_gc), 258 fsparam_flag("disable_roll_forward", Opt_disable_roll_forward), 259 fsparam_flag("norecovery", Opt_norecovery), 260 fsparam_flag_no("discard", Opt_discard), 261 fsparam_flag("no_heap", Opt_noheap), 262 fsparam_flag("heap", Opt_heap), 263 fsparam_flag_no("user_xattr", Opt_user_xattr), 264 fsparam_flag_no("acl", Opt_acl), 265 fsparam_s32("active_logs", Opt_active_logs), 266 fsparam_flag("disable_ext_identify", Opt_disable_ext_identify), 267 fsparam_flag_no("inline_xattr", Opt_inline_xattr), 268 fsparam_s32("inline_xattr_size", Opt_inline_xattr_size), 269 fsparam_flag_no("inline_data", Opt_inline_data), 270 fsparam_flag_no("inline_dentry", Opt_inline_dentry), 271 fsparam_flag_no("flush_merge", Opt_flush_merge), 272 fsparam_flag_no("barrier", Opt_barrier), 273 fsparam_flag("fastboot", Opt_fastboot), 274 fsparam_flag_no("extent_cache", Opt_extent_cache), 275 fsparam_flag("data_flush", Opt_data_flush), 276 fsparam_u32("reserve_root", Opt_reserve_root), 277 fsparam_u32("reserve_node", Opt_reserve_node), 278 fsparam_gid("resgid", Opt_resgid), 279 fsparam_uid("resuid", Opt_resuid), 280 fsparam_enum("mode", Opt_mode, f2fs_param_mode), 281 fsparam_s32("fault_injection", Opt_fault_injection), 282 fsparam_u32("fault_type", Opt_fault_type), 283 fsparam_flag_no("lazytime", Opt_lazytime), 284 fsparam_flag_no("quota", Opt_quota), 285 fsparam_flag("usrquota", Opt_usrquota), 286 fsparam_flag("grpquota", Opt_grpquota), 287 fsparam_flag("prjquota", Opt_prjquota), 288 fsparam_string_empty("usrjquota", Opt_usrjquota), 289 fsparam_string_empty("grpjquota", Opt_grpjquota), 290 fsparam_string_empty("prjjquota", Opt_prjjquota), 291 fsparam_flag("nat_bits", Opt_nat_bits), 292 fsparam_enum("jqfmt", Opt_jqfmt, f2fs_param_jqfmt), 293 fsparam_enum("alloc_mode", Opt_alloc, f2fs_param_alloc_mode), 294 fsparam_enum("fsync_mode", Opt_fsync, f2fs_param_fsync_mode), 295 fsparam_string("test_dummy_encryption", Opt_test_dummy_encryption), 296 fsparam_flag("test_dummy_encryption", Opt_test_dummy_encryption), 297 fsparam_flag("inlinecrypt", Opt_inlinecrypt), 298 fsparam_string("checkpoint", Opt_checkpoint), 299 fsparam_flag_no("checkpoint_merge", Opt_checkpoint_merge), 300 fsparam_string("compress_algorithm", Opt_compress_algorithm), 301 fsparam_u32("compress_log_size", Opt_compress_log_size), 302 fsparam_string("compress_extension", Opt_compress_extension), 303 fsparam_string("nocompress_extension", Opt_nocompress_extension), 304 fsparam_flag("compress_chksum", Opt_compress_chksum), 305 fsparam_enum("compress_mode", Opt_compress_mode, f2fs_param_compress_mode), 306 fsparam_flag("compress_cache", Opt_compress_cache), 307 fsparam_flag("atgc", Opt_atgc), 308 fsparam_flag_no("gc_merge", Opt_gc_merge), 309 fsparam_enum("discard_unit", Opt_discard_unit, f2fs_param_discard_unit), 310 fsparam_enum("memory", Opt_memory_mode, f2fs_param_memory_mode), 311 fsparam_flag("age_extent_cache", Opt_age_extent_cache), 312 fsparam_enum("errors", Opt_errors, f2fs_param_errors), 313 fsparam_enum("lookup_mode", Opt_lookup_mode, f2fs_param_lookup_mode), 314 {} 315 }; 316 317 /* Resort to a match_table for this interestingly formatted option */ 318 static match_table_t f2fs_checkpoint_tokens = { 319 {Opt_checkpoint_disable, "disable"}, 320 {Opt_checkpoint_disable_cap, "disable:%u"}, 321 {Opt_checkpoint_disable_cap_perc, "disable:%u%%"}, 322 {Opt_checkpoint_enable, "enable"}, 323 {Opt_err, NULL}, 324 }; 325 326 #define F2FS_SPEC_background_gc (1 << 0) 327 #define F2FS_SPEC_inline_xattr_size (1 << 1) 328 #define F2FS_SPEC_active_logs (1 << 2) 329 #define F2FS_SPEC_reserve_root (1 << 3) 330 #define F2FS_SPEC_resgid (1 << 4) 331 #define F2FS_SPEC_resuid (1 << 5) 332 #define F2FS_SPEC_mode (1 << 6) 333 #define F2FS_SPEC_fault_injection (1 << 7) 334 #define F2FS_SPEC_fault_type (1 << 8) 335 #define F2FS_SPEC_jqfmt (1 << 9) 336 #define F2FS_SPEC_alloc_mode (1 << 10) 337 #define F2FS_SPEC_fsync_mode (1 << 11) 338 #define F2FS_SPEC_checkpoint_disable_cap (1 << 12) 339 #define F2FS_SPEC_checkpoint_disable_cap_perc (1 << 13) 340 #define F2FS_SPEC_compress_level (1 << 14) 341 #define F2FS_SPEC_compress_algorithm (1 << 15) 342 #define F2FS_SPEC_compress_log_size (1 << 16) 343 #define F2FS_SPEC_compress_extension (1 << 17) 344 #define F2FS_SPEC_nocompress_extension (1 << 18) 345 #define F2FS_SPEC_compress_chksum (1 << 19) 346 #define F2FS_SPEC_compress_mode (1 << 20) 347 #define F2FS_SPEC_discard_unit (1 << 21) 348 #define F2FS_SPEC_memory_mode (1 << 22) 349 #define F2FS_SPEC_errors (1 << 23) 350 #define F2FS_SPEC_lookup_mode (1 << 24) 351 #define F2FS_SPEC_reserve_node (1 << 25) 352 353 struct f2fs_fs_context { 354 struct f2fs_mount_info info; 355 unsigned long long opt_mask; /* Bits changed */ 356 unsigned int spec_mask; 357 unsigned short qname_mask; 358 }; 359 360 #define F2FS_CTX_INFO(ctx) ((ctx)->info) 361 362 static inline void ctx_set_opt(struct f2fs_fs_context *ctx, 363 enum f2fs_mount_opt flag) 364 { 365 ctx->info.opt |= BIT(flag); 366 ctx->opt_mask |= BIT(flag); 367 } 368 369 static inline void ctx_clear_opt(struct f2fs_fs_context *ctx, 370 enum f2fs_mount_opt flag) 371 { 372 ctx->info.opt &= ~BIT(flag); 373 ctx->opt_mask |= BIT(flag); 374 } 375 376 static inline bool ctx_test_opt(struct f2fs_fs_context *ctx, 377 enum f2fs_mount_opt flag) 378 { 379 return ctx->info.opt & BIT(flag); 380 } 381 382 void f2fs_printk(struct f2fs_sb_info *sbi, bool limit_rate, 383 const char *fmt, ...) 384 { 385 struct va_format vaf; 386 va_list args; 387 int level; 388 389 va_start(args, fmt); 390 391 level = printk_get_level(fmt); 392 vaf.fmt = printk_skip_level(fmt); 393 vaf.va = &args; 394 if (limit_rate) 395 if (sbi) 396 printk_ratelimited("%c%cF2FS-fs (%s): %pV\n", 397 KERN_SOH_ASCII, level, sbi->sb->s_id, &vaf); 398 else 399 printk_ratelimited("%c%cF2FS-fs: %pV\n", 400 KERN_SOH_ASCII, level, &vaf); 401 else 402 if (sbi) 403 printk("%c%cF2FS-fs (%s): %pV\n", 404 KERN_SOH_ASCII, level, sbi->sb->s_id, &vaf); 405 else 406 printk("%c%cF2FS-fs: %pV\n", 407 KERN_SOH_ASCII, level, &vaf); 408 409 va_end(args); 410 } 411 412 #if IS_ENABLED(CONFIG_UNICODE) 413 static const struct f2fs_sb_encodings { 414 __u16 magic; 415 char *name; 416 unsigned int version; 417 } f2fs_sb_encoding_map[] = { 418 {F2FS_ENC_UTF8_12_1, "utf8", UNICODE_AGE(12, 1, 0)}, 419 }; 420 421 static const struct f2fs_sb_encodings * 422 f2fs_sb_read_encoding(const struct f2fs_super_block *sb) 423 { 424 __u16 magic = le16_to_cpu(sb->s_encoding); 425 int i; 426 427 for (i = 0; i < ARRAY_SIZE(f2fs_sb_encoding_map); i++) 428 if (magic == f2fs_sb_encoding_map[i].magic) 429 return &f2fs_sb_encoding_map[i]; 430 431 return NULL; 432 } 433 434 struct kmem_cache *f2fs_cf_name_slab; 435 static int __init f2fs_create_casefold_cache(void) 436 { 437 f2fs_cf_name_slab = f2fs_kmem_cache_create("f2fs_casefolded_name", 438 F2FS_NAME_LEN); 439 return f2fs_cf_name_slab ? 0 : -ENOMEM; 440 } 441 442 static void f2fs_destroy_casefold_cache(void) 443 { 444 kmem_cache_destroy(f2fs_cf_name_slab); 445 } 446 #else 447 static int __init f2fs_create_casefold_cache(void) { return 0; } 448 static void f2fs_destroy_casefold_cache(void) { } 449 #endif 450 451 static inline void limit_reserve_root(struct f2fs_sb_info *sbi) 452 { 453 block_t block_limit = min((sbi->user_block_count >> 3), 454 sbi->user_block_count - sbi->reserved_blocks); 455 block_t node_limit = sbi->total_node_count >> 3; 456 457 /* limit is 12.5% */ 458 if (test_opt(sbi, RESERVE_ROOT) && 459 F2FS_OPTION(sbi).root_reserved_blocks > block_limit) { 460 F2FS_OPTION(sbi).root_reserved_blocks = block_limit; 461 f2fs_info(sbi, "Reduce reserved blocks for root = %u", 462 F2FS_OPTION(sbi).root_reserved_blocks); 463 } 464 if (test_opt(sbi, RESERVE_NODE) && 465 F2FS_OPTION(sbi).root_reserved_nodes > node_limit) { 466 F2FS_OPTION(sbi).root_reserved_nodes = node_limit; 467 f2fs_info(sbi, "Reduce reserved nodes for root = %u", 468 F2FS_OPTION(sbi).root_reserved_nodes); 469 } 470 if (!test_opt(sbi, RESERVE_ROOT) && !test_opt(sbi, RESERVE_NODE) && 471 (!uid_eq(F2FS_OPTION(sbi).s_resuid, 472 make_kuid(&init_user_ns, F2FS_DEF_RESUID)) || 473 !gid_eq(F2FS_OPTION(sbi).s_resgid, 474 make_kgid(&init_user_ns, F2FS_DEF_RESGID)))) 475 f2fs_info(sbi, "Ignore s_resuid=%u, s_resgid=%u w/o reserve_root" 476 " and reserve_node", 477 from_kuid_munged(&init_user_ns, 478 F2FS_OPTION(sbi).s_resuid), 479 from_kgid_munged(&init_user_ns, 480 F2FS_OPTION(sbi).s_resgid)); 481 } 482 483 static inline void adjust_unusable_cap_perc(struct f2fs_sb_info *sbi) 484 { 485 if (!F2FS_OPTION(sbi).unusable_cap_perc) 486 return; 487 488 if (F2FS_OPTION(sbi).unusable_cap_perc == 100) 489 F2FS_OPTION(sbi).unusable_cap = sbi->user_block_count; 490 else 491 F2FS_OPTION(sbi).unusable_cap = (sbi->user_block_count / 100) * 492 F2FS_OPTION(sbi).unusable_cap_perc; 493 494 f2fs_info(sbi, "Adjust unusable cap for checkpoint=disable = %u / %u%%", 495 F2FS_OPTION(sbi).unusable_cap, 496 F2FS_OPTION(sbi).unusable_cap_perc); 497 } 498 499 static void init_once(void *foo) 500 { 501 struct f2fs_inode_info *fi = (struct f2fs_inode_info *) foo; 502 503 inode_init_once(&fi->vfs_inode); 504 #ifdef CONFIG_FS_ENCRYPTION 505 fi->i_crypt_info = NULL; 506 #endif 507 #ifdef CONFIG_FS_VERITY 508 fi->i_verity_info = NULL; 509 #endif 510 } 511 512 #ifdef CONFIG_QUOTA 513 static const char * const quotatypes[] = INITQFNAMES; 514 #define QTYPE2NAME(t) (quotatypes[t]) 515 /* 516 * Note the name of the specified quota file. 517 */ 518 static int f2fs_note_qf_name(struct fs_context *fc, int qtype, 519 struct fs_parameter *param) 520 { 521 struct f2fs_fs_context *ctx = fc->fs_private; 522 char *qname; 523 524 if (param->size < 1) { 525 f2fs_err(NULL, "Missing quota name"); 526 return -EINVAL; 527 } 528 if (strchr(param->string, '/')) { 529 f2fs_err(NULL, "quotafile must be on filesystem root"); 530 return -EINVAL; 531 } 532 if (ctx->info.s_qf_names[qtype]) { 533 if (strcmp(ctx->info.s_qf_names[qtype], param->string) != 0) { 534 f2fs_err(NULL, "Quota file already specified"); 535 return -EINVAL; 536 } 537 return 0; 538 } 539 540 qname = kmemdup_nul(param->string, param->size, GFP_KERNEL); 541 if (!qname) { 542 f2fs_err(NULL, "Not enough memory for storing quotafile name"); 543 return -ENOMEM; 544 } 545 F2FS_CTX_INFO(ctx).s_qf_names[qtype] = qname; 546 ctx->qname_mask |= 1 << qtype; 547 return 0; 548 } 549 550 /* 551 * Clear the name of the specified quota file. 552 */ 553 static int f2fs_unnote_qf_name(struct fs_context *fc, int qtype) 554 { 555 struct f2fs_fs_context *ctx = fc->fs_private; 556 557 kfree(ctx->info.s_qf_names[qtype]); 558 ctx->info.s_qf_names[qtype] = NULL; 559 ctx->qname_mask |= 1 << qtype; 560 return 0; 561 } 562 563 static void f2fs_unnote_qf_name_all(struct fs_context *fc) 564 { 565 int i; 566 567 for (i = 0; i < MAXQUOTAS; i++) 568 f2fs_unnote_qf_name(fc, i); 569 } 570 #endif 571 572 static int f2fs_parse_test_dummy_encryption(const struct fs_parameter *param, 573 struct f2fs_fs_context *ctx) 574 { 575 int err; 576 577 if (!IS_ENABLED(CONFIG_FS_ENCRYPTION)) { 578 f2fs_warn(NULL, "test_dummy_encryption option not supported"); 579 return -EINVAL; 580 } 581 err = fscrypt_parse_test_dummy_encryption(param, 582 &ctx->info.dummy_enc_policy); 583 if (err) { 584 if (err == -EINVAL) 585 f2fs_warn(NULL, "Value of option \"%s\" is unrecognized", 586 param->key); 587 else if (err == -EEXIST) 588 f2fs_warn(NULL, "Conflicting test_dummy_encryption options"); 589 else 590 f2fs_warn(NULL, "Error processing option \"%s\" [%d]", 591 param->key, err); 592 return -EINVAL; 593 } 594 return 0; 595 } 596 597 #ifdef CONFIG_F2FS_FS_COMPRESSION 598 static bool is_compress_extension_exist(struct f2fs_mount_info *info, 599 const char *new_ext, bool is_ext) 600 { 601 unsigned char (*ext)[F2FS_EXTENSION_LEN]; 602 int ext_cnt; 603 int i; 604 605 if (is_ext) { 606 ext = info->extensions; 607 ext_cnt = info->compress_ext_cnt; 608 } else { 609 ext = info->noextensions; 610 ext_cnt = info->nocompress_ext_cnt; 611 } 612 613 for (i = 0; i < ext_cnt; i++) { 614 if (!strcasecmp(new_ext, ext[i])) 615 return true; 616 } 617 618 return false; 619 } 620 621 /* 622 * 1. The same extension name cannot not appear in both compress and non-compress extension 623 * at the same time. 624 * 2. If the compress extension specifies all files, the types specified by the non-compress 625 * extension will be treated as special cases and will not be compressed. 626 * 3. Don't allow the non-compress extension specifies all files. 627 */ 628 static int f2fs_test_compress_extension(unsigned char (*noext)[F2FS_EXTENSION_LEN], 629 int noext_cnt, 630 unsigned char (*ext)[F2FS_EXTENSION_LEN], 631 int ext_cnt) 632 { 633 int index = 0, no_index = 0; 634 635 if (!noext_cnt) 636 return 0; 637 638 for (no_index = 0; no_index < noext_cnt; no_index++) { 639 if (strlen(noext[no_index]) == 0) 640 continue; 641 if (!strcasecmp("*", noext[no_index])) { 642 f2fs_info(NULL, "Don't allow the nocompress extension specifies all files"); 643 return -EINVAL; 644 } 645 for (index = 0; index < ext_cnt; index++) { 646 if (strlen(ext[index]) == 0) 647 continue; 648 if (!strcasecmp(ext[index], noext[no_index])) { 649 f2fs_info(NULL, "Don't allow the same extension %s appear in both compress and nocompress extension", 650 ext[index]); 651 return -EINVAL; 652 } 653 } 654 } 655 return 0; 656 } 657 658 #ifdef CONFIG_F2FS_FS_LZ4 659 static int f2fs_set_lz4hc_level(struct f2fs_fs_context *ctx, const char *str) 660 { 661 #ifdef CONFIG_F2FS_FS_LZ4HC 662 unsigned int level; 663 664 if (strlen(str) == 3) { 665 F2FS_CTX_INFO(ctx).compress_level = 0; 666 ctx->spec_mask |= F2FS_SPEC_compress_level; 667 return 0; 668 } 669 670 str += 3; 671 672 if (str[0] != ':') { 673 f2fs_info(NULL, "wrong format, e.g. <alg_name>:<compr_level>"); 674 return -EINVAL; 675 } 676 if (kstrtouint(str + 1, 10, &level)) 677 return -EINVAL; 678 679 if (!f2fs_is_compress_level_valid(COMPRESS_LZ4, level)) { 680 f2fs_info(NULL, "invalid lz4hc compress level: %d", level); 681 return -EINVAL; 682 } 683 684 F2FS_CTX_INFO(ctx).compress_level = level; 685 ctx->spec_mask |= F2FS_SPEC_compress_level; 686 return 0; 687 #else 688 if (strlen(str) == 3) { 689 F2FS_CTX_INFO(ctx).compress_level = 0; 690 ctx->spec_mask |= F2FS_SPEC_compress_level; 691 return 0; 692 } 693 f2fs_info(NULL, "kernel doesn't support lz4hc compression"); 694 return -EINVAL; 695 #endif 696 } 697 #endif 698 699 #ifdef CONFIG_F2FS_FS_ZSTD 700 static int f2fs_set_zstd_level(struct f2fs_fs_context *ctx, const char *str) 701 { 702 int level; 703 int len = 4; 704 705 if (strlen(str) == len) { 706 F2FS_CTX_INFO(ctx).compress_level = F2FS_ZSTD_DEFAULT_CLEVEL; 707 ctx->spec_mask |= F2FS_SPEC_compress_level; 708 return 0; 709 } 710 711 str += len; 712 713 if (str[0] != ':') { 714 f2fs_info(NULL, "wrong format, e.g. <alg_name>:<compr_level>"); 715 return -EINVAL; 716 } 717 if (kstrtoint(str + 1, 10, &level)) 718 return -EINVAL; 719 720 /* f2fs does not support negative compress level now */ 721 if (level < 0) { 722 f2fs_info(NULL, "do not support negative compress level: %d", level); 723 return -ERANGE; 724 } 725 726 if (!f2fs_is_compress_level_valid(COMPRESS_ZSTD, level)) { 727 f2fs_info(NULL, "invalid zstd compress level: %d", level); 728 return -EINVAL; 729 } 730 731 F2FS_CTX_INFO(ctx).compress_level = level; 732 ctx->spec_mask |= F2FS_SPEC_compress_level; 733 return 0; 734 } 735 #endif 736 #endif 737 738 static int f2fs_parse_param(struct fs_context *fc, struct fs_parameter *param) 739 { 740 struct f2fs_fs_context *ctx = fc->fs_private; 741 #ifdef CONFIG_F2FS_FS_COMPRESSION 742 unsigned char (*ext)[F2FS_EXTENSION_LEN]; 743 unsigned char (*noext)[F2FS_EXTENSION_LEN]; 744 int ext_cnt, noext_cnt; 745 char *name; 746 #endif 747 substring_t args[MAX_OPT_ARGS]; 748 struct fs_parse_result result; 749 int token, ret, arg; 750 751 token = fs_parse(fc, f2fs_param_specs, param, &result); 752 if (token < 0) 753 return token; 754 755 switch (token) { 756 case Opt_gc_background: 757 F2FS_CTX_INFO(ctx).bggc_mode = result.uint_32; 758 ctx->spec_mask |= F2FS_SPEC_background_gc; 759 break; 760 case Opt_disable_roll_forward: 761 ctx_set_opt(ctx, F2FS_MOUNT_DISABLE_ROLL_FORWARD); 762 break; 763 case Opt_norecovery: 764 /* requires ro mount, checked in f2fs_validate_options */ 765 ctx_set_opt(ctx, F2FS_MOUNT_NORECOVERY); 766 break; 767 case Opt_discard: 768 if (result.negated) 769 ctx_clear_opt(ctx, F2FS_MOUNT_DISCARD); 770 else 771 ctx_set_opt(ctx, F2FS_MOUNT_DISCARD); 772 break; 773 case Opt_noheap: 774 case Opt_heap: 775 f2fs_warn(NULL, "heap/no_heap options were deprecated"); 776 break; 777 #ifdef CONFIG_F2FS_FS_XATTR 778 case Opt_user_xattr: 779 if (result.negated) 780 ctx_clear_opt(ctx, F2FS_MOUNT_XATTR_USER); 781 else 782 ctx_set_opt(ctx, F2FS_MOUNT_XATTR_USER); 783 break; 784 case Opt_inline_xattr: 785 if (result.negated) 786 ctx_clear_opt(ctx, F2FS_MOUNT_INLINE_XATTR); 787 else 788 ctx_set_opt(ctx, F2FS_MOUNT_INLINE_XATTR); 789 break; 790 case Opt_inline_xattr_size: 791 if (result.int_32 < MIN_INLINE_XATTR_SIZE || 792 result.int_32 > MAX_INLINE_XATTR_SIZE) { 793 f2fs_err(NULL, "inline xattr size is out of range: %u ~ %u", 794 (u32)MIN_INLINE_XATTR_SIZE, (u32)MAX_INLINE_XATTR_SIZE); 795 return -EINVAL; 796 } 797 ctx_set_opt(ctx, F2FS_MOUNT_INLINE_XATTR_SIZE); 798 F2FS_CTX_INFO(ctx).inline_xattr_size = result.int_32; 799 ctx->spec_mask |= F2FS_SPEC_inline_xattr_size; 800 break; 801 #else 802 case Opt_user_xattr: 803 case Opt_inline_xattr: 804 case Opt_inline_xattr_size: 805 f2fs_info(NULL, "%s options not supported", param->key); 806 break; 807 #endif 808 #ifdef CONFIG_F2FS_FS_POSIX_ACL 809 case Opt_acl: 810 if (result.negated) 811 ctx_clear_opt(ctx, F2FS_MOUNT_POSIX_ACL); 812 else 813 ctx_set_opt(ctx, F2FS_MOUNT_POSIX_ACL); 814 break; 815 #else 816 case Opt_acl: 817 f2fs_info(NULL, "%s options not supported", param->key); 818 break; 819 #endif 820 case Opt_active_logs: 821 if (result.int_32 != 2 && result.int_32 != 4 && 822 result.int_32 != NR_CURSEG_PERSIST_TYPE) 823 return -EINVAL; 824 ctx->spec_mask |= F2FS_SPEC_active_logs; 825 F2FS_CTX_INFO(ctx).active_logs = result.int_32; 826 break; 827 case Opt_disable_ext_identify: 828 ctx_set_opt(ctx, F2FS_MOUNT_DISABLE_EXT_IDENTIFY); 829 break; 830 case Opt_inline_data: 831 if (result.negated) 832 ctx_clear_opt(ctx, F2FS_MOUNT_INLINE_DATA); 833 else 834 ctx_set_opt(ctx, F2FS_MOUNT_INLINE_DATA); 835 break; 836 case Opt_inline_dentry: 837 if (result.negated) 838 ctx_clear_opt(ctx, F2FS_MOUNT_INLINE_DENTRY); 839 else 840 ctx_set_opt(ctx, F2FS_MOUNT_INLINE_DENTRY); 841 break; 842 case Opt_flush_merge: 843 if (result.negated) 844 ctx_clear_opt(ctx, F2FS_MOUNT_FLUSH_MERGE); 845 else 846 ctx_set_opt(ctx, F2FS_MOUNT_FLUSH_MERGE); 847 break; 848 case Opt_barrier: 849 if (result.negated) 850 ctx_set_opt(ctx, F2FS_MOUNT_NOBARRIER); 851 else 852 ctx_clear_opt(ctx, F2FS_MOUNT_NOBARRIER); 853 break; 854 case Opt_fastboot: 855 ctx_set_opt(ctx, F2FS_MOUNT_FASTBOOT); 856 break; 857 case Opt_extent_cache: 858 if (result.negated) 859 ctx_clear_opt(ctx, F2FS_MOUNT_READ_EXTENT_CACHE); 860 else 861 ctx_set_opt(ctx, F2FS_MOUNT_READ_EXTENT_CACHE); 862 break; 863 case Opt_data_flush: 864 ctx_set_opt(ctx, F2FS_MOUNT_DATA_FLUSH); 865 break; 866 case Opt_reserve_root: 867 ctx_set_opt(ctx, F2FS_MOUNT_RESERVE_ROOT); 868 F2FS_CTX_INFO(ctx).root_reserved_blocks = result.uint_32; 869 ctx->spec_mask |= F2FS_SPEC_reserve_root; 870 break; 871 case Opt_reserve_node: 872 ctx_set_opt(ctx, F2FS_MOUNT_RESERVE_NODE); 873 F2FS_CTX_INFO(ctx).root_reserved_nodes = result.uint_32; 874 ctx->spec_mask |= F2FS_SPEC_reserve_node; 875 break; 876 case Opt_resuid: 877 F2FS_CTX_INFO(ctx).s_resuid = result.uid; 878 ctx->spec_mask |= F2FS_SPEC_resuid; 879 break; 880 case Opt_resgid: 881 F2FS_CTX_INFO(ctx).s_resgid = result.gid; 882 ctx->spec_mask |= F2FS_SPEC_resgid; 883 break; 884 case Opt_mode: 885 F2FS_CTX_INFO(ctx).fs_mode = result.uint_32; 886 ctx->spec_mask |= F2FS_SPEC_mode; 887 break; 888 #ifdef CONFIG_F2FS_FAULT_INJECTION 889 case Opt_fault_injection: 890 F2FS_CTX_INFO(ctx).fault_info.inject_rate = result.int_32; 891 ctx->spec_mask |= F2FS_SPEC_fault_injection; 892 ctx_set_opt(ctx, F2FS_MOUNT_FAULT_INJECTION); 893 break; 894 895 case Opt_fault_type: 896 if (result.uint_32 > BIT(FAULT_MAX)) 897 return -EINVAL; 898 F2FS_CTX_INFO(ctx).fault_info.inject_type = result.uint_32; 899 ctx->spec_mask |= F2FS_SPEC_fault_type; 900 ctx_set_opt(ctx, F2FS_MOUNT_FAULT_INJECTION); 901 break; 902 #else 903 case Opt_fault_injection: 904 case Opt_fault_type: 905 f2fs_info(NULL, "%s options not supported", param->key); 906 break; 907 #endif 908 case Opt_lazytime: 909 if (result.negated) 910 ctx_clear_opt(ctx, F2FS_MOUNT_LAZYTIME); 911 else 912 ctx_set_opt(ctx, F2FS_MOUNT_LAZYTIME); 913 break; 914 #ifdef CONFIG_QUOTA 915 case Opt_quota: 916 if (result.negated) { 917 ctx_clear_opt(ctx, F2FS_MOUNT_QUOTA); 918 ctx_clear_opt(ctx, F2FS_MOUNT_USRQUOTA); 919 ctx_clear_opt(ctx, F2FS_MOUNT_GRPQUOTA); 920 ctx_clear_opt(ctx, F2FS_MOUNT_PRJQUOTA); 921 } else 922 ctx_set_opt(ctx, F2FS_MOUNT_USRQUOTA); 923 break; 924 case Opt_usrquota: 925 ctx_set_opt(ctx, F2FS_MOUNT_USRQUOTA); 926 break; 927 case Opt_grpquota: 928 ctx_set_opt(ctx, F2FS_MOUNT_GRPQUOTA); 929 break; 930 case Opt_prjquota: 931 ctx_set_opt(ctx, F2FS_MOUNT_PRJQUOTA); 932 break; 933 case Opt_usrjquota: 934 if (!*param->string) 935 ret = f2fs_unnote_qf_name(fc, USRQUOTA); 936 else 937 ret = f2fs_note_qf_name(fc, USRQUOTA, param); 938 if (ret) 939 return ret; 940 break; 941 case Opt_grpjquota: 942 if (!*param->string) 943 ret = f2fs_unnote_qf_name(fc, GRPQUOTA); 944 else 945 ret = f2fs_note_qf_name(fc, GRPQUOTA, param); 946 if (ret) 947 return ret; 948 break; 949 case Opt_prjjquota: 950 if (!*param->string) 951 ret = f2fs_unnote_qf_name(fc, PRJQUOTA); 952 else 953 ret = f2fs_note_qf_name(fc, PRJQUOTA, param); 954 if (ret) 955 return ret; 956 break; 957 case Opt_jqfmt: 958 F2FS_CTX_INFO(ctx).s_jquota_fmt = result.int_32; 959 ctx->spec_mask |= F2FS_SPEC_jqfmt; 960 break; 961 #else 962 case Opt_quota: 963 case Opt_usrquota: 964 case Opt_grpquota: 965 case Opt_prjquota: 966 case Opt_usrjquota: 967 case Opt_grpjquota: 968 case Opt_prjjquota: 969 f2fs_info(NULL, "quota operations not supported"); 970 break; 971 #endif 972 case Opt_alloc: 973 F2FS_CTX_INFO(ctx).alloc_mode = result.uint_32; 974 ctx->spec_mask |= F2FS_SPEC_alloc_mode; 975 break; 976 case Opt_fsync: 977 F2FS_CTX_INFO(ctx).fsync_mode = result.uint_32; 978 ctx->spec_mask |= F2FS_SPEC_fsync_mode; 979 break; 980 case Opt_test_dummy_encryption: 981 ret = f2fs_parse_test_dummy_encryption(param, ctx); 982 if (ret) 983 return ret; 984 break; 985 case Opt_inlinecrypt: 986 #ifdef CONFIG_FS_ENCRYPTION_INLINE_CRYPT 987 ctx_set_opt(ctx, F2FS_MOUNT_INLINECRYPT); 988 #else 989 f2fs_info(NULL, "inline encryption not supported"); 990 #endif 991 break; 992 case Opt_checkpoint: 993 /* 994 * Initialize args struct so we know whether arg was 995 * found; some options take optional arguments. 996 */ 997 args[0].from = args[0].to = NULL; 998 arg = 0; 999 1000 /* revert to match_table for checkpoint= options */ 1001 token = match_token(param->string, f2fs_checkpoint_tokens, args); 1002 switch (token) { 1003 case Opt_checkpoint_disable_cap_perc: 1004 if (args->from && match_int(args, &arg)) 1005 return -EINVAL; 1006 if (arg < 0 || arg > 100) 1007 return -EINVAL; 1008 F2FS_CTX_INFO(ctx).unusable_cap_perc = arg; 1009 ctx->spec_mask |= F2FS_SPEC_checkpoint_disable_cap_perc; 1010 ctx_set_opt(ctx, F2FS_MOUNT_DISABLE_CHECKPOINT); 1011 break; 1012 case Opt_checkpoint_disable_cap: 1013 if (args->from && match_int(args, &arg)) 1014 return -EINVAL; 1015 F2FS_CTX_INFO(ctx).unusable_cap = arg; 1016 ctx->spec_mask |= F2FS_SPEC_checkpoint_disable_cap; 1017 ctx_set_opt(ctx, F2FS_MOUNT_DISABLE_CHECKPOINT); 1018 break; 1019 case Opt_checkpoint_disable: 1020 ctx_set_opt(ctx, F2FS_MOUNT_DISABLE_CHECKPOINT); 1021 break; 1022 case Opt_checkpoint_enable: 1023 F2FS_CTX_INFO(ctx).unusable_cap_perc = 0; 1024 ctx->spec_mask |= F2FS_SPEC_checkpoint_disable_cap_perc; 1025 F2FS_CTX_INFO(ctx).unusable_cap = 0; 1026 ctx->spec_mask |= F2FS_SPEC_checkpoint_disable_cap; 1027 ctx_clear_opt(ctx, F2FS_MOUNT_DISABLE_CHECKPOINT); 1028 break; 1029 default: 1030 return -EINVAL; 1031 } 1032 break; 1033 case Opt_checkpoint_merge: 1034 if (result.negated) 1035 ctx_clear_opt(ctx, F2FS_MOUNT_MERGE_CHECKPOINT); 1036 else 1037 ctx_set_opt(ctx, F2FS_MOUNT_MERGE_CHECKPOINT); 1038 break; 1039 #ifdef CONFIG_F2FS_FS_COMPRESSION 1040 case Opt_compress_algorithm: 1041 name = param->string; 1042 if (!strcmp(name, "lzo")) { 1043 #ifdef CONFIG_F2FS_FS_LZO 1044 F2FS_CTX_INFO(ctx).compress_level = 0; 1045 F2FS_CTX_INFO(ctx).compress_algorithm = COMPRESS_LZO; 1046 ctx->spec_mask |= F2FS_SPEC_compress_level; 1047 ctx->spec_mask |= F2FS_SPEC_compress_algorithm; 1048 #else 1049 f2fs_info(NULL, "kernel doesn't support lzo compression"); 1050 #endif 1051 } else if (!strncmp(name, "lz4", 3)) { 1052 #ifdef CONFIG_F2FS_FS_LZ4 1053 ret = f2fs_set_lz4hc_level(ctx, name); 1054 if (ret) 1055 return -EINVAL; 1056 F2FS_CTX_INFO(ctx).compress_algorithm = COMPRESS_LZ4; 1057 ctx->spec_mask |= F2FS_SPEC_compress_algorithm; 1058 #else 1059 f2fs_info(NULL, "kernel doesn't support lz4 compression"); 1060 #endif 1061 } else if (!strncmp(name, "zstd", 4)) { 1062 #ifdef CONFIG_F2FS_FS_ZSTD 1063 ret = f2fs_set_zstd_level(ctx, name); 1064 if (ret) 1065 return -EINVAL; 1066 F2FS_CTX_INFO(ctx).compress_algorithm = COMPRESS_ZSTD; 1067 ctx->spec_mask |= F2FS_SPEC_compress_algorithm; 1068 #else 1069 f2fs_info(NULL, "kernel doesn't support zstd compression"); 1070 #endif 1071 } else if (!strcmp(name, "lzo-rle")) { 1072 #ifdef CONFIG_F2FS_FS_LZORLE 1073 F2FS_CTX_INFO(ctx).compress_level = 0; 1074 F2FS_CTX_INFO(ctx).compress_algorithm = COMPRESS_LZORLE; 1075 ctx->spec_mask |= F2FS_SPEC_compress_level; 1076 ctx->spec_mask |= F2FS_SPEC_compress_algorithm; 1077 #else 1078 f2fs_info(NULL, "kernel doesn't support lzorle compression"); 1079 #endif 1080 } else 1081 return -EINVAL; 1082 break; 1083 case Opt_compress_log_size: 1084 if (result.uint_32 < MIN_COMPRESS_LOG_SIZE || 1085 result.uint_32 > MAX_COMPRESS_LOG_SIZE) { 1086 f2fs_err(NULL, 1087 "Compress cluster log size is out of range"); 1088 return -EINVAL; 1089 } 1090 F2FS_CTX_INFO(ctx).compress_log_size = result.uint_32; 1091 ctx->spec_mask |= F2FS_SPEC_compress_log_size; 1092 break; 1093 case Opt_compress_extension: 1094 name = param->string; 1095 ext = F2FS_CTX_INFO(ctx).extensions; 1096 ext_cnt = F2FS_CTX_INFO(ctx).compress_ext_cnt; 1097 1098 if (strlen(name) >= F2FS_EXTENSION_LEN || 1099 ext_cnt >= COMPRESS_EXT_NUM) { 1100 f2fs_err(NULL, "invalid extension length/number"); 1101 return -EINVAL; 1102 } 1103 1104 if (is_compress_extension_exist(&ctx->info, name, true)) 1105 break; 1106 1107 ret = strscpy(ext[ext_cnt], name, F2FS_EXTENSION_LEN); 1108 if (ret < 0) 1109 return ret; 1110 F2FS_CTX_INFO(ctx).compress_ext_cnt++; 1111 ctx->spec_mask |= F2FS_SPEC_compress_extension; 1112 break; 1113 case Opt_nocompress_extension: 1114 name = param->string; 1115 noext = F2FS_CTX_INFO(ctx).noextensions; 1116 noext_cnt = F2FS_CTX_INFO(ctx).nocompress_ext_cnt; 1117 1118 if (strlen(name) >= F2FS_EXTENSION_LEN || 1119 noext_cnt >= COMPRESS_EXT_NUM) { 1120 f2fs_err(NULL, "invalid extension length/number"); 1121 return -EINVAL; 1122 } 1123 1124 if (is_compress_extension_exist(&ctx->info, name, false)) 1125 break; 1126 1127 ret = strscpy(noext[noext_cnt], name, F2FS_EXTENSION_LEN); 1128 if (ret < 0) 1129 return ret; 1130 F2FS_CTX_INFO(ctx).nocompress_ext_cnt++; 1131 ctx->spec_mask |= F2FS_SPEC_nocompress_extension; 1132 break; 1133 case Opt_compress_chksum: 1134 F2FS_CTX_INFO(ctx).compress_chksum = true; 1135 ctx->spec_mask |= F2FS_SPEC_compress_chksum; 1136 break; 1137 case Opt_compress_mode: 1138 F2FS_CTX_INFO(ctx).compress_mode = result.uint_32; 1139 ctx->spec_mask |= F2FS_SPEC_compress_mode; 1140 break; 1141 case Opt_compress_cache: 1142 ctx_set_opt(ctx, F2FS_MOUNT_COMPRESS_CACHE); 1143 break; 1144 #else 1145 case Opt_compress_algorithm: 1146 case Opt_compress_log_size: 1147 case Opt_compress_extension: 1148 case Opt_nocompress_extension: 1149 case Opt_compress_chksum: 1150 case Opt_compress_mode: 1151 case Opt_compress_cache: 1152 f2fs_info(NULL, "compression options not supported"); 1153 break; 1154 #endif 1155 case Opt_atgc: 1156 ctx_set_opt(ctx, F2FS_MOUNT_ATGC); 1157 break; 1158 case Opt_gc_merge: 1159 if (result.negated) 1160 ctx_clear_opt(ctx, F2FS_MOUNT_GC_MERGE); 1161 else 1162 ctx_set_opt(ctx, F2FS_MOUNT_GC_MERGE); 1163 break; 1164 case Opt_discard_unit: 1165 F2FS_CTX_INFO(ctx).discard_unit = result.uint_32; 1166 ctx->spec_mask |= F2FS_SPEC_discard_unit; 1167 break; 1168 case Opt_memory_mode: 1169 F2FS_CTX_INFO(ctx).memory_mode = result.uint_32; 1170 ctx->spec_mask |= F2FS_SPEC_memory_mode; 1171 break; 1172 case Opt_age_extent_cache: 1173 ctx_set_opt(ctx, F2FS_MOUNT_AGE_EXTENT_CACHE); 1174 break; 1175 case Opt_errors: 1176 F2FS_CTX_INFO(ctx).errors = result.uint_32; 1177 ctx->spec_mask |= F2FS_SPEC_errors; 1178 break; 1179 case Opt_nat_bits: 1180 ctx_set_opt(ctx, F2FS_MOUNT_NAT_BITS); 1181 break; 1182 case Opt_lookup_mode: 1183 F2FS_CTX_INFO(ctx).lookup_mode = result.uint_32; 1184 ctx->spec_mask |= F2FS_SPEC_lookup_mode; 1185 break; 1186 } 1187 return 0; 1188 } 1189 1190 /* 1191 * Check quota settings consistency. 1192 */ 1193 static int f2fs_check_quota_consistency(struct fs_context *fc, 1194 struct super_block *sb) 1195 { 1196 struct f2fs_sb_info *sbi = F2FS_SB(sb); 1197 #ifdef CONFIG_QUOTA 1198 struct f2fs_fs_context *ctx = fc->fs_private; 1199 bool quota_feature = f2fs_sb_has_quota_ino(sbi); 1200 bool quota_turnon = sb_any_quota_loaded(sb); 1201 char *old_qname, *new_qname; 1202 bool usr_qf_name, grp_qf_name, prj_qf_name, usrquota, grpquota, prjquota; 1203 int i; 1204 1205 /* 1206 * We do the test below only for project quotas. 'usrquota' and 1207 * 'grpquota' mount options are allowed even without quota feature 1208 * to support legacy quotas in quota files. 1209 */ 1210 if (ctx_test_opt(ctx, F2FS_MOUNT_PRJQUOTA) && 1211 !f2fs_sb_has_project_quota(sbi)) { 1212 f2fs_err(sbi, "Project quota feature not enabled. Cannot enable project quota enforcement."); 1213 return -EINVAL; 1214 } 1215 1216 if (ctx->qname_mask) { 1217 for (i = 0; i < MAXQUOTAS; i++) { 1218 if (!(ctx->qname_mask & (1 << i))) 1219 continue; 1220 1221 old_qname = F2FS_OPTION(sbi).s_qf_names[i]; 1222 new_qname = F2FS_CTX_INFO(ctx).s_qf_names[i]; 1223 if (quota_turnon && 1224 !!old_qname != !!new_qname) 1225 goto err_jquota_change; 1226 1227 if (old_qname) { 1228 if (!new_qname) { 1229 f2fs_info(sbi, "remove qf_name %s", 1230 old_qname); 1231 continue; 1232 } else if (strcmp(old_qname, new_qname) == 0) { 1233 ctx->qname_mask &= ~(1 << i); 1234 continue; 1235 } 1236 goto err_jquota_specified; 1237 } 1238 1239 if (quota_feature) { 1240 f2fs_info(sbi, "QUOTA feature is enabled, so ignore qf_name"); 1241 ctx->qname_mask &= ~(1 << i); 1242 kfree(F2FS_CTX_INFO(ctx).s_qf_names[i]); 1243 F2FS_CTX_INFO(ctx).s_qf_names[i] = NULL; 1244 } 1245 } 1246 } 1247 1248 /* Make sure we don't mix old and new quota format */ 1249 usr_qf_name = F2FS_OPTION(sbi).s_qf_names[USRQUOTA] || 1250 F2FS_CTX_INFO(ctx).s_qf_names[USRQUOTA]; 1251 grp_qf_name = F2FS_OPTION(sbi).s_qf_names[GRPQUOTA] || 1252 F2FS_CTX_INFO(ctx).s_qf_names[GRPQUOTA]; 1253 prj_qf_name = F2FS_OPTION(sbi).s_qf_names[PRJQUOTA] || 1254 F2FS_CTX_INFO(ctx).s_qf_names[PRJQUOTA]; 1255 usrquota = test_opt(sbi, USRQUOTA) || 1256 ctx_test_opt(ctx, F2FS_MOUNT_USRQUOTA); 1257 grpquota = test_opt(sbi, GRPQUOTA) || 1258 ctx_test_opt(ctx, F2FS_MOUNT_GRPQUOTA); 1259 prjquota = test_opt(sbi, PRJQUOTA) || 1260 ctx_test_opt(ctx, F2FS_MOUNT_PRJQUOTA); 1261 1262 if (usr_qf_name) { 1263 ctx_clear_opt(ctx, F2FS_MOUNT_USRQUOTA); 1264 usrquota = false; 1265 } 1266 if (grp_qf_name) { 1267 ctx_clear_opt(ctx, F2FS_MOUNT_GRPQUOTA); 1268 grpquota = false; 1269 } 1270 if (prj_qf_name) { 1271 ctx_clear_opt(ctx, F2FS_MOUNT_PRJQUOTA); 1272 prjquota = false; 1273 } 1274 if (usr_qf_name || grp_qf_name || prj_qf_name) { 1275 if (grpquota || usrquota || prjquota) { 1276 f2fs_err(sbi, "old and new quota format mixing"); 1277 return -EINVAL; 1278 } 1279 if (!(ctx->spec_mask & F2FS_SPEC_jqfmt || 1280 F2FS_OPTION(sbi).s_jquota_fmt)) { 1281 f2fs_err(sbi, "journaled quota format not specified"); 1282 return -EINVAL; 1283 } 1284 } 1285 return 0; 1286 1287 err_jquota_change: 1288 f2fs_err(sbi, "Cannot change journaled quota options when quota turned on"); 1289 return -EINVAL; 1290 err_jquota_specified: 1291 f2fs_err(sbi, "%s quota file already specified", 1292 QTYPE2NAME(i)); 1293 return -EINVAL; 1294 1295 #else 1296 if (f2fs_readonly(sbi->sb)) 1297 return 0; 1298 if (f2fs_sb_has_quota_ino(sbi)) { 1299 f2fs_info(sbi, "Filesystem with quota feature cannot be mounted RDWR without CONFIG_QUOTA"); 1300 return -EINVAL; 1301 } 1302 if (f2fs_sb_has_project_quota(sbi)) { 1303 f2fs_err(sbi, "Filesystem with project quota feature cannot be mounted RDWR without CONFIG_QUOTA"); 1304 return -EINVAL; 1305 } 1306 1307 return 0; 1308 #endif 1309 } 1310 1311 static int f2fs_check_test_dummy_encryption(struct fs_context *fc, 1312 struct super_block *sb) 1313 { 1314 struct f2fs_fs_context *ctx = fc->fs_private; 1315 struct f2fs_sb_info *sbi = F2FS_SB(sb); 1316 1317 if (!fscrypt_is_dummy_policy_set(&F2FS_CTX_INFO(ctx).dummy_enc_policy)) 1318 return 0; 1319 1320 if (!f2fs_sb_has_encrypt(sbi)) { 1321 f2fs_err(sbi, "Encrypt feature is off"); 1322 return -EINVAL; 1323 } 1324 1325 /* 1326 * This mount option is just for testing, and it's not worthwhile to 1327 * implement the extra complexity (e.g. RCU protection) that would be 1328 * needed to allow it to be set or changed during remount. We do allow 1329 * it to be specified during remount, but only if there is no change. 1330 */ 1331 if (fc->purpose == FS_CONTEXT_FOR_RECONFIGURE) { 1332 if (fscrypt_dummy_policies_equal(&F2FS_OPTION(sbi).dummy_enc_policy, 1333 &F2FS_CTX_INFO(ctx).dummy_enc_policy)) 1334 return 0; 1335 f2fs_warn(sbi, "Can't set or change test_dummy_encryption on remount"); 1336 return -EINVAL; 1337 } 1338 return 0; 1339 } 1340 1341 static inline bool test_compression_spec(unsigned int mask) 1342 { 1343 return mask & (F2FS_SPEC_compress_algorithm 1344 | F2FS_SPEC_compress_log_size 1345 | F2FS_SPEC_compress_extension 1346 | F2FS_SPEC_nocompress_extension 1347 | F2FS_SPEC_compress_chksum 1348 | F2FS_SPEC_compress_mode); 1349 } 1350 1351 static inline void clear_compression_spec(struct f2fs_fs_context *ctx) 1352 { 1353 ctx->spec_mask &= ~(F2FS_SPEC_compress_algorithm 1354 | F2FS_SPEC_compress_log_size 1355 | F2FS_SPEC_compress_extension 1356 | F2FS_SPEC_nocompress_extension 1357 | F2FS_SPEC_compress_chksum 1358 | F2FS_SPEC_compress_mode); 1359 } 1360 1361 static int f2fs_check_compression(struct fs_context *fc, 1362 struct super_block *sb) 1363 { 1364 #ifdef CONFIG_F2FS_FS_COMPRESSION 1365 struct f2fs_fs_context *ctx = fc->fs_private; 1366 struct f2fs_sb_info *sbi = F2FS_SB(sb); 1367 int i, cnt; 1368 1369 if (!f2fs_sb_has_compression(sbi)) { 1370 if (test_compression_spec(ctx->spec_mask) || 1371 ctx_test_opt(ctx, F2FS_MOUNT_COMPRESS_CACHE)) 1372 f2fs_info(sbi, "Image doesn't support compression"); 1373 clear_compression_spec(ctx); 1374 ctx->opt_mask &= ~BIT(F2FS_MOUNT_COMPRESS_CACHE); 1375 return 0; 1376 } 1377 if (ctx->spec_mask & F2FS_SPEC_compress_extension) { 1378 cnt = F2FS_CTX_INFO(ctx).compress_ext_cnt; 1379 for (i = 0; i < F2FS_CTX_INFO(ctx).compress_ext_cnt; i++) { 1380 if (is_compress_extension_exist(&F2FS_OPTION(sbi), 1381 F2FS_CTX_INFO(ctx).extensions[i], true)) { 1382 F2FS_CTX_INFO(ctx).extensions[i][0] = '\0'; 1383 cnt--; 1384 } 1385 } 1386 if (F2FS_OPTION(sbi).compress_ext_cnt + cnt > COMPRESS_EXT_NUM) { 1387 f2fs_err(sbi, "invalid extension length/number"); 1388 return -EINVAL; 1389 } 1390 } 1391 if (ctx->spec_mask & F2FS_SPEC_nocompress_extension) { 1392 cnt = F2FS_CTX_INFO(ctx).nocompress_ext_cnt; 1393 for (i = 0; i < F2FS_CTX_INFO(ctx).nocompress_ext_cnt; i++) { 1394 if (is_compress_extension_exist(&F2FS_OPTION(sbi), 1395 F2FS_CTX_INFO(ctx).noextensions[i], false)) { 1396 F2FS_CTX_INFO(ctx).noextensions[i][0] = '\0'; 1397 cnt--; 1398 } 1399 } 1400 if (F2FS_OPTION(sbi).nocompress_ext_cnt + cnt > COMPRESS_EXT_NUM) { 1401 f2fs_err(sbi, "invalid noextension length/number"); 1402 return -EINVAL; 1403 } 1404 } 1405 1406 if (f2fs_test_compress_extension(F2FS_CTX_INFO(ctx).noextensions, 1407 F2FS_CTX_INFO(ctx).nocompress_ext_cnt, 1408 F2FS_CTX_INFO(ctx).extensions, 1409 F2FS_CTX_INFO(ctx).compress_ext_cnt)) { 1410 f2fs_err(sbi, "new noextensions conflicts with new extensions"); 1411 return -EINVAL; 1412 } 1413 if (f2fs_test_compress_extension(F2FS_CTX_INFO(ctx).noextensions, 1414 F2FS_CTX_INFO(ctx).nocompress_ext_cnt, 1415 F2FS_OPTION(sbi).extensions, 1416 F2FS_OPTION(sbi).compress_ext_cnt)) { 1417 f2fs_err(sbi, "new noextensions conflicts with old extensions"); 1418 return -EINVAL; 1419 } 1420 if (f2fs_test_compress_extension(F2FS_OPTION(sbi).noextensions, 1421 F2FS_OPTION(sbi).nocompress_ext_cnt, 1422 F2FS_CTX_INFO(ctx).extensions, 1423 F2FS_CTX_INFO(ctx).compress_ext_cnt)) { 1424 f2fs_err(sbi, "new extensions conflicts with old noextensions"); 1425 return -EINVAL; 1426 } 1427 #endif 1428 return 0; 1429 } 1430 1431 static int f2fs_check_opt_consistency(struct fs_context *fc, 1432 struct super_block *sb) 1433 { 1434 struct f2fs_fs_context *ctx = fc->fs_private; 1435 struct f2fs_sb_info *sbi = F2FS_SB(sb); 1436 int err; 1437 1438 if (ctx_test_opt(ctx, F2FS_MOUNT_NORECOVERY) && !f2fs_readonly(sb)) 1439 return -EINVAL; 1440 1441 if (f2fs_hw_should_discard(sbi) && 1442 (ctx->opt_mask & BIT(F2FS_MOUNT_DISCARD)) && 1443 !ctx_test_opt(ctx, F2FS_MOUNT_DISCARD)) { 1444 f2fs_warn(sbi, "discard is required for zoned block devices"); 1445 return -EINVAL; 1446 } 1447 1448 if (!f2fs_hw_support_discard(sbi) && 1449 (ctx->opt_mask & BIT(F2FS_MOUNT_DISCARD)) && 1450 ctx_test_opt(ctx, F2FS_MOUNT_DISCARD)) { 1451 f2fs_warn(sbi, "device does not support discard"); 1452 ctx_clear_opt(ctx, F2FS_MOUNT_DISCARD); 1453 ctx->opt_mask &= ~BIT(F2FS_MOUNT_DISCARD); 1454 } 1455 1456 if (f2fs_sb_has_device_alias(sbi) && 1457 (ctx->opt_mask & BIT(F2FS_MOUNT_READ_EXTENT_CACHE)) && 1458 !ctx_test_opt(ctx, F2FS_MOUNT_READ_EXTENT_CACHE)) { 1459 f2fs_err(sbi, "device aliasing requires extent cache"); 1460 return -EINVAL; 1461 } 1462 1463 if (test_opt(sbi, RESERVE_ROOT) && 1464 (ctx->opt_mask & BIT(F2FS_MOUNT_RESERVE_ROOT)) && 1465 ctx_test_opt(ctx, F2FS_MOUNT_RESERVE_ROOT)) { 1466 f2fs_info(sbi, "Preserve previous reserve_root=%u", 1467 F2FS_OPTION(sbi).root_reserved_blocks); 1468 ctx_clear_opt(ctx, F2FS_MOUNT_RESERVE_ROOT); 1469 ctx->opt_mask &= ~BIT(F2FS_MOUNT_RESERVE_ROOT); 1470 } 1471 if (test_opt(sbi, RESERVE_NODE) && 1472 (ctx->opt_mask & BIT(F2FS_MOUNT_RESERVE_NODE)) && 1473 ctx_test_opt(ctx, F2FS_MOUNT_RESERVE_NODE)) { 1474 f2fs_info(sbi, "Preserve previous reserve_node=%u", 1475 F2FS_OPTION(sbi).root_reserved_nodes); 1476 ctx_clear_opt(ctx, F2FS_MOUNT_RESERVE_NODE); 1477 ctx->opt_mask &= ~BIT(F2FS_MOUNT_RESERVE_NODE); 1478 } 1479 1480 err = f2fs_check_test_dummy_encryption(fc, sb); 1481 if (err) 1482 return err; 1483 1484 err = f2fs_check_compression(fc, sb); 1485 if (err) 1486 return err; 1487 1488 err = f2fs_check_quota_consistency(fc, sb); 1489 if (err) 1490 return err; 1491 1492 if (!IS_ENABLED(CONFIG_UNICODE) && f2fs_sb_has_casefold(sbi)) { 1493 f2fs_err(sbi, 1494 "Filesystem with casefold feature cannot be mounted without CONFIG_UNICODE"); 1495 return -EINVAL; 1496 } 1497 1498 /* 1499 * The BLKZONED feature indicates that the drive was formatted with 1500 * zone alignment optimization. This is optional for host-aware 1501 * devices, but mandatory for host-managed zoned block devices. 1502 */ 1503 if (f2fs_sb_has_blkzoned(sbi)) { 1504 if (F2FS_CTX_INFO(ctx).bggc_mode == BGGC_MODE_OFF) { 1505 f2fs_warn(sbi, "zoned devices need bggc"); 1506 return -EINVAL; 1507 } 1508 #ifdef CONFIG_BLK_DEV_ZONED 1509 if ((ctx->spec_mask & F2FS_SPEC_discard_unit) && 1510 F2FS_CTX_INFO(ctx).discard_unit != DISCARD_UNIT_SECTION) { 1511 f2fs_info(sbi, "Zoned block device doesn't need small discard, set discard_unit=section by default"); 1512 F2FS_CTX_INFO(ctx).discard_unit = DISCARD_UNIT_SECTION; 1513 } 1514 1515 if ((ctx->spec_mask & F2FS_SPEC_mode) && 1516 F2FS_CTX_INFO(ctx).fs_mode != FS_MODE_LFS) { 1517 f2fs_info(sbi, "Only lfs mode is allowed with zoned block device feature"); 1518 return -EINVAL; 1519 } 1520 #else 1521 f2fs_err(sbi, "Zoned block device support is not enabled"); 1522 return -EINVAL; 1523 #endif 1524 } 1525 1526 if (ctx_test_opt(ctx, F2FS_MOUNT_INLINE_XATTR_SIZE)) { 1527 if (!f2fs_sb_has_extra_attr(sbi) || 1528 !f2fs_sb_has_flexible_inline_xattr(sbi)) { 1529 f2fs_err(sbi, "extra_attr or flexible_inline_xattr feature is off"); 1530 return -EINVAL; 1531 } 1532 if (!ctx_test_opt(ctx, F2FS_MOUNT_INLINE_XATTR) && !test_opt(sbi, INLINE_XATTR)) { 1533 f2fs_err(sbi, "inline_xattr_size option should be set with inline_xattr option"); 1534 return -EINVAL; 1535 } 1536 } 1537 1538 if (ctx_test_opt(ctx, F2FS_MOUNT_ATGC) && 1539 F2FS_CTX_INFO(ctx).fs_mode == FS_MODE_LFS) { 1540 f2fs_err(sbi, "LFS is not compatible with ATGC"); 1541 return -EINVAL; 1542 } 1543 1544 if (f2fs_is_readonly(sbi) && ctx_test_opt(ctx, F2FS_MOUNT_FLUSH_MERGE)) { 1545 f2fs_err(sbi, "FLUSH_MERGE not compatible with readonly mode"); 1546 return -EINVAL; 1547 } 1548 1549 if (f2fs_sb_has_readonly(sbi) && !f2fs_readonly(sbi->sb)) { 1550 f2fs_err(sbi, "Allow to mount readonly mode only"); 1551 return -EROFS; 1552 } 1553 return 0; 1554 } 1555 1556 static void f2fs_apply_quota_options(struct fs_context *fc, 1557 struct super_block *sb) 1558 { 1559 #ifdef CONFIG_QUOTA 1560 struct f2fs_fs_context *ctx = fc->fs_private; 1561 struct f2fs_sb_info *sbi = F2FS_SB(sb); 1562 bool quota_feature = f2fs_sb_has_quota_ino(sbi); 1563 char *qname; 1564 int i; 1565 1566 if (quota_feature) 1567 return; 1568 1569 for (i = 0; i < MAXQUOTAS; i++) { 1570 if (!(ctx->qname_mask & (1 << i))) 1571 continue; 1572 1573 qname = F2FS_CTX_INFO(ctx).s_qf_names[i]; 1574 if (qname) { 1575 qname = kstrdup(F2FS_CTX_INFO(ctx).s_qf_names[i], 1576 GFP_KERNEL | __GFP_NOFAIL); 1577 set_opt(sbi, QUOTA); 1578 } 1579 F2FS_OPTION(sbi).s_qf_names[i] = qname; 1580 } 1581 1582 if (ctx->spec_mask & F2FS_SPEC_jqfmt) 1583 F2FS_OPTION(sbi).s_jquota_fmt = F2FS_CTX_INFO(ctx).s_jquota_fmt; 1584 1585 if (quota_feature && F2FS_OPTION(sbi).s_jquota_fmt) { 1586 f2fs_info(sbi, "QUOTA feature is enabled, so ignore jquota_fmt"); 1587 F2FS_OPTION(sbi).s_jquota_fmt = 0; 1588 } 1589 #endif 1590 } 1591 1592 static void f2fs_apply_test_dummy_encryption(struct fs_context *fc, 1593 struct super_block *sb) 1594 { 1595 struct f2fs_fs_context *ctx = fc->fs_private; 1596 struct f2fs_sb_info *sbi = F2FS_SB(sb); 1597 1598 if (!fscrypt_is_dummy_policy_set(&F2FS_CTX_INFO(ctx).dummy_enc_policy) || 1599 /* if already set, it was already verified to be the same */ 1600 fscrypt_is_dummy_policy_set(&F2FS_OPTION(sbi).dummy_enc_policy)) 1601 return; 1602 swap(F2FS_OPTION(sbi).dummy_enc_policy, F2FS_CTX_INFO(ctx).dummy_enc_policy); 1603 f2fs_warn(sbi, "Test dummy encryption mode enabled"); 1604 } 1605 1606 static void f2fs_apply_compression(struct fs_context *fc, 1607 struct super_block *sb) 1608 { 1609 #ifdef CONFIG_F2FS_FS_COMPRESSION 1610 struct f2fs_fs_context *ctx = fc->fs_private; 1611 struct f2fs_sb_info *sbi = F2FS_SB(sb); 1612 unsigned char (*ctx_ext)[F2FS_EXTENSION_LEN]; 1613 unsigned char (*sbi_ext)[F2FS_EXTENSION_LEN]; 1614 int ctx_cnt, sbi_cnt, i; 1615 1616 if (ctx->spec_mask & F2FS_SPEC_compress_level) 1617 F2FS_OPTION(sbi).compress_level = 1618 F2FS_CTX_INFO(ctx).compress_level; 1619 if (ctx->spec_mask & F2FS_SPEC_compress_algorithm) 1620 F2FS_OPTION(sbi).compress_algorithm = 1621 F2FS_CTX_INFO(ctx).compress_algorithm; 1622 if (ctx->spec_mask & F2FS_SPEC_compress_log_size) 1623 F2FS_OPTION(sbi).compress_log_size = 1624 F2FS_CTX_INFO(ctx).compress_log_size; 1625 if (ctx->spec_mask & F2FS_SPEC_compress_chksum) 1626 F2FS_OPTION(sbi).compress_chksum = 1627 F2FS_CTX_INFO(ctx).compress_chksum; 1628 if (ctx->spec_mask & F2FS_SPEC_compress_mode) 1629 F2FS_OPTION(sbi).compress_mode = 1630 F2FS_CTX_INFO(ctx).compress_mode; 1631 if (ctx->spec_mask & F2FS_SPEC_compress_extension) { 1632 ctx_ext = F2FS_CTX_INFO(ctx).extensions; 1633 ctx_cnt = F2FS_CTX_INFO(ctx).compress_ext_cnt; 1634 sbi_ext = F2FS_OPTION(sbi).extensions; 1635 sbi_cnt = F2FS_OPTION(sbi).compress_ext_cnt; 1636 for (i = 0; i < ctx_cnt; i++) { 1637 if (strlen(ctx_ext[i]) == 0) 1638 continue; 1639 strscpy(sbi_ext[sbi_cnt], ctx_ext[i]); 1640 sbi_cnt++; 1641 } 1642 F2FS_OPTION(sbi).compress_ext_cnt = sbi_cnt; 1643 } 1644 if (ctx->spec_mask & F2FS_SPEC_nocompress_extension) { 1645 ctx_ext = F2FS_CTX_INFO(ctx).noextensions; 1646 ctx_cnt = F2FS_CTX_INFO(ctx).nocompress_ext_cnt; 1647 sbi_ext = F2FS_OPTION(sbi).noextensions; 1648 sbi_cnt = F2FS_OPTION(sbi).nocompress_ext_cnt; 1649 for (i = 0; i < ctx_cnt; i++) { 1650 if (strlen(ctx_ext[i]) == 0) 1651 continue; 1652 strscpy(sbi_ext[sbi_cnt], ctx_ext[i]); 1653 sbi_cnt++; 1654 } 1655 F2FS_OPTION(sbi).nocompress_ext_cnt = sbi_cnt; 1656 } 1657 #endif 1658 } 1659 1660 static void f2fs_apply_options(struct fs_context *fc, struct super_block *sb) 1661 { 1662 struct f2fs_fs_context *ctx = fc->fs_private; 1663 struct f2fs_sb_info *sbi = F2FS_SB(sb); 1664 1665 F2FS_OPTION(sbi).opt &= ~ctx->opt_mask; 1666 F2FS_OPTION(sbi).opt |= F2FS_CTX_INFO(ctx).opt; 1667 1668 if (ctx->spec_mask & F2FS_SPEC_background_gc) 1669 F2FS_OPTION(sbi).bggc_mode = F2FS_CTX_INFO(ctx).bggc_mode; 1670 if (ctx->spec_mask & F2FS_SPEC_inline_xattr_size) 1671 F2FS_OPTION(sbi).inline_xattr_size = 1672 F2FS_CTX_INFO(ctx).inline_xattr_size; 1673 if (ctx->spec_mask & F2FS_SPEC_active_logs) 1674 F2FS_OPTION(sbi).active_logs = F2FS_CTX_INFO(ctx).active_logs; 1675 if (ctx->spec_mask & F2FS_SPEC_reserve_root) 1676 F2FS_OPTION(sbi).root_reserved_blocks = 1677 F2FS_CTX_INFO(ctx).root_reserved_blocks; 1678 if (ctx->spec_mask & F2FS_SPEC_reserve_node) 1679 F2FS_OPTION(sbi).root_reserved_nodes = 1680 F2FS_CTX_INFO(ctx).root_reserved_nodes; 1681 if (ctx->spec_mask & F2FS_SPEC_resgid) 1682 F2FS_OPTION(sbi).s_resgid = F2FS_CTX_INFO(ctx).s_resgid; 1683 if (ctx->spec_mask & F2FS_SPEC_resuid) 1684 F2FS_OPTION(sbi).s_resuid = F2FS_CTX_INFO(ctx).s_resuid; 1685 if (ctx->spec_mask & F2FS_SPEC_mode) 1686 F2FS_OPTION(sbi).fs_mode = F2FS_CTX_INFO(ctx).fs_mode; 1687 #ifdef CONFIG_F2FS_FAULT_INJECTION 1688 if (ctx->spec_mask & F2FS_SPEC_fault_injection) 1689 (void)f2fs_build_fault_attr(sbi, 1690 F2FS_CTX_INFO(ctx).fault_info.inject_rate, 0, FAULT_RATE); 1691 if (ctx->spec_mask & F2FS_SPEC_fault_type) 1692 (void)f2fs_build_fault_attr(sbi, 0, 1693 F2FS_CTX_INFO(ctx).fault_info.inject_type, FAULT_TYPE); 1694 #endif 1695 if (ctx->spec_mask & F2FS_SPEC_alloc_mode) 1696 F2FS_OPTION(sbi).alloc_mode = F2FS_CTX_INFO(ctx).alloc_mode; 1697 if (ctx->spec_mask & F2FS_SPEC_fsync_mode) 1698 F2FS_OPTION(sbi).fsync_mode = F2FS_CTX_INFO(ctx).fsync_mode; 1699 if (ctx->spec_mask & F2FS_SPEC_checkpoint_disable_cap) 1700 F2FS_OPTION(sbi).unusable_cap = F2FS_CTX_INFO(ctx).unusable_cap; 1701 if (ctx->spec_mask & F2FS_SPEC_checkpoint_disable_cap_perc) 1702 F2FS_OPTION(sbi).unusable_cap_perc = 1703 F2FS_CTX_INFO(ctx).unusable_cap_perc; 1704 if (ctx->spec_mask & F2FS_SPEC_discard_unit) 1705 F2FS_OPTION(sbi).discard_unit = F2FS_CTX_INFO(ctx).discard_unit; 1706 if (ctx->spec_mask & F2FS_SPEC_memory_mode) 1707 F2FS_OPTION(sbi).memory_mode = F2FS_CTX_INFO(ctx).memory_mode; 1708 if (ctx->spec_mask & F2FS_SPEC_errors) 1709 F2FS_OPTION(sbi).errors = F2FS_CTX_INFO(ctx).errors; 1710 if (ctx->spec_mask & F2FS_SPEC_lookup_mode) 1711 F2FS_OPTION(sbi).lookup_mode = F2FS_CTX_INFO(ctx).lookup_mode; 1712 1713 f2fs_apply_compression(fc, sb); 1714 f2fs_apply_test_dummy_encryption(fc, sb); 1715 f2fs_apply_quota_options(fc, sb); 1716 } 1717 1718 static int f2fs_sanity_check_options(struct f2fs_sb_info *sbi, bool remount) 1719 { 1720 if (f2fs_sb_has_device_alias(sbi) && 1721 !test_opt(sbi, READ_EXTENT_CACHE)) { 1722 f2fs_err(sbi, "device aliasing requires extent cache"); 1723 return -EINVAL; 1724 } 1725 1726 if (!remount) 1727 return 0; 1728 1729 #ifdef CONFIG_BLK_DEV_ZONED 1730 if (f2fs_sb_has_blkzoned(sbi) && 1731 sbi->max_open_zones < F2FS_OPTION(sbi).active_logs) { 1732 f2fs_err(sbi, 1733 "zoned: max open zones %u is too small, need at least %u open zones", 1734 sbi->max_open_zones, F2FS_OPTION(sbi).active_logs); 1735 return -EINVAL; 1736 } 1737 #endif 1738 if (f2fs_lfs_mode(sbi) && !IS_F2FS_IPU_DISABLE(sbi)) { 1739 f2fs_warn(sbi, "LFS is not compatible with IPU"); 1740 return -EINVAL; 1741 } 1742 return 0; 1743 } 1744 1745 static struct inode *f2fs_alloc_inode(struct super_block *sb) 1746 { 1747 struct f2fs_inode_info *fi; 1748 1749 if (time_to_inject(F2FS_SB(sb), FAULT_SLAB_ALLOC)) 1750 return NULL; 1751 1752 fi = alloc_inode_sb(sb, f2fs_inode_cachep, GFP_F2FS_ZERO); 1753 if (!fi) 1754 return NULL; 1755 1756 init_once((void *) fi); 1757 1758 /* Initialize f2fs-specific inode info */ 1759 atomic_set(&fi->dirty_pages, 0); 1760 atomic_set(&fi->i_compr_blocks, 0); 1761 atomic_set(&fi->open_count, 0); 1762 atomic_set(&fi->writeback, 0); 1763 init_f2fs_rwsem(&fi->i_sem); 1764 spin_lock_init(&fi->i_size_lock); 1765 INIT_LIST_HEAD(&fi->dirty_list); 1766 INIT_LIST_HEAD(&fi->gdirty_list); 1767 INIT_LIST_HEAD(&fi->gdonate_list); 1768 init_f2fs_rwsem(&fi->i_gc_rwsem[READ]); 1769 init_f2fs_rwsem(&fi->i_gc_rwsem[WRITE]); 1770 init_f2fs_rwsem(&fi->i_xattr_sem); 1771 1772 /* Will be used by directory only */ 1773 fi->i_dir_level = F2FS_SB(sb)->dir_level; 1774 1775 return &fi->vfs_inode; 1776 } 1777 1778 static int f2fs_drop_inode(struct inode *inode) 1779 { 1780 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1781 int ret; 1782 1783 /* 1784 * during filesystem shutdown, if checkpoint is disabled, 1785 * drop useless meta/node dirty pages. 1786 */ 1787 if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED))) { 1788 if (inode->i_ino == F2FS_NODE_INO(sbi) || 1789 inode->i_ino == F2FS_META_INO(sbi)) { 1790 trace_f2fs_drop_inode(inode, 1); 1791 return 1; 1792 } 1793 } 1794 1795 /* 1796 * This is to avoid a deadlock condition like below. 1797 * writeback_single_inode(inode) 1798 * - f2fs_write_data_page 1799 * - f2fs_gc -> iput -> evict 1800 * - inode_wait_for_writeback(inode) 1801 */ 1802 if ((!inode_unhashed(inode) && inode_state_read(inode) & I_SYNC)) { 1803 if (!inode->i_nlink && !is_bad_inode(inode)) { 1804 /* to avoid evict_inode call simultaneously */ 1805 __iget(inode); 1806 spin_unlock(&inode->i_lock); 1807 1808 /* should remain fi->extent_tree for writepage */ 1809 f2fs_destroy_extent_node(inode); 1810 1811 sb_start_intwrite(inode->i_sb); 1812 f2fs_i_size_write(inode, 0); 1813 1814 f2fs_submit_merged_write_cond(F2FS_I_SB(inode), 1815 inode, NULL, 0, DATA); 1816 truncate_inode_pages_final(inode->i_mapping); 1817 1818 if (F2FS_HAS_BLOCKS(inode)) 1819 f2fs_truncate(inode); 1820 1821 sb_end_intwrite(inode->i_sb); 1822 1823 spin_lock(&inode->i_lock); 1824 atomic_dec(&inode->i_count); 1825 } 1826 trace_f2fs_drop_inode(inode, 0); 1827 return 0; 1828 } 1829 ret = inode_generic_drop(inode); 1830 if (!ret) 1831 ret = fscrypt_drop_inode(inode); 1832 trace_f2fs_drop_inode(inode, ret); 1833 return ret; 1834 } 1835 1836 int f2fs_inode_dirtied(struct inode *inode, bool sync) 1837 { 1838 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1839 int ret = 0; 1840 1841 spin_lock(&sbi->inode_lock[DIRTY_META]); 1842 if (is_inode_flag_set(inode, FI_DIRTY_INODE)) { 1843 ret = 1; 1844 } else { 1845 set_inode_flag(inode, FI_DIRTY_INODE); 1846 stat_inc_dirty_inode(sbi, DIRTY_META); 1847 } 1848 if (sync && list_empty(&F2FS_I(inode)->gdirty_list)) { 1849 list_add_tail(&F2FS_I(inode)->gdirty_list, 1850 &sbi->inode_list[DIRTY_META]); 1851 inc_page_count(sbi, F2FS_DIRTY_IMETA); 1852 } 1853 spin_unlock(&sbi->inode_lock[DIRTY_META]); 1854 1855 /* if atomic write is not committed, set inode w/ atomic dirty */ 1856 if (!ret && f2fs_is_atomic_file(inode) && 1857 !is_inode_flag_set(inode, FI_ATOMIC_COMMITTED)) 1858 set_inode_flag(inode, FI_ATOMIC_DIRTIED); 1859 1860 return ret; 1861 } 1862 1863 void f2fs_inode_synced(struct inode *inode) 1864 { 1865 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1866 1867 spin_lock(&sbi->inode_lock[DIRTY_META]); 1868 if (!is_inode_flag_set(inode, FI_DIRTY_INODE)) { 1869 spin_unlock(&sbi->inode_lock[DIRTY_META]); 1870 return; 1871 } 1872 if (!list_empty(&F2FS_I(inode)->gdirty_list)) { 1873 list_del_init(&F2FS_I(inode)->gdirty_list); 1874 dec_page_count(sbi, F2FS_DIRTY_IMETA); 1875 } 1876 clear_inode_flag(inode, FI_DIRTY_INODE); 1877 clear_inode_flag(inode, FI_AUTO_RECOVER); 1878 stat_dec_dirty_inode(F2FS_I_SB(inode), DIRTY_META); 1879 spin_unlock(&sbi->inode_lock[DIRTY_META]); 1880 } 1881 1882 /* 1883 * f2fs_dirty_inode() is called from __mark_inode_dirty() 1884 * 1885 * We should call set_dirty_inode to write the dirty inode through write_inode. 1886 */ 1887 static void f2fs_dirty_inode(struct inode *inode, int flags) 1888 { 1889 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1890 1891 if (inode->i_ino == F2FS_NODE_INO(sbi) || 1892 inode->i_ino == F2FS_META_INO(sbi)) 1893 return; 1894 1895 if (is_inode_flag_set(inode, FI_AUTO_RECOVER)) 1896 clear_inode_flag(inode, FI_AUTO_RECOVER); 1897 1898 f2fs_inode_dirtied(inode, false); 1899 } 1900 1901 static void f2fs_free_inode(struct inode *inode) 1902 { 1903 fscrypt_free_inode(inode); 1904 kmem_cache_free(f2fs_inode_cachep, F2FS_I(inode)); 1905 } 1906 1907 static void destroy_percpu_info(struct f2fs_sb_info *sbi) 1908 { 1909 percpu_counter_destroy(&sbi->total_valid_inode_count); 1910 percpu_counter_destroy(&sbi->rf_node_block_count); 1911 percpu_counter_destroy(&sbi->alloc_valid_block_count); 1912 } 1913 1914 static void destroy_device_list(struct f2fs_sb_info *sbi) 1915 { 1916 int i; 1917 1918 for (i = 0; i < sbi->s_ndevs; i++) { 1919 if (i > 0) 1920 bdev_fput(FDEV(i).bdev_file); 1921 #ifdef CONFIG_BLK_DEV_ZONED 1922 kvfree(FDEV(i).blkz_seq); 1923 #endif 1924 } 1925 kvfree(sbi->devs); 1926 } 1927 1928 static void f2fs_put_super(struct super_block *sb) 1929 { 1930 struct f2fs_sb_info *sbi = F2FS_SB(sb); 1931 int i; 1932 int err = 0; 1933 bool done; 1934 1935 /* unregister procfs/sysfs entries in advance to avoid race case */ 1936 f2fs_unregister_sysfs(sbi); 1937 1938 f2fs_quota_off_umount(sb); 1939 1940 /* prevent remaining shrinker jobs */ 1941 mutex_lock(&sbi->umount_mutex); 1942 1943 /* 1944 * flush all issued checkpoints and stop checkpoint issue thread. 1945 * after then, all checkpoints should be done by each process context. 1946 */ 1947 f2fs_stop_ckpt_thread(sbi); 1948 1949 /* 1950 * We don't need to do checkpoint when superblock is clean. 1951 * But, the previous checkpoint was not done by umount, it needs to do 1952 * clean checkpoint again. 1953 */ 1954 if ((is_sbi_flag_set(sbi, SBI_IS_DIRTY) || 1955 !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG))) { 1956 struct cp_control cpc = { 1957 .reason = CP_UMOUNT, 1958 }; 1959 stat_inc_cp_call_count(sbi, TOTAL_CALL); 1960 err = f2fs_write_checkpoint(sbi, &cpc); 1961 } 1962 1963 /* be sure to wait for any on-going discard commands */ 1964 done = f2fs_issue_discard_timeout(sbi); 1965 if (f2fs_realtime_discard_enable(sbi) && !sbi->discard_blks && done) { 1966 struct cp_control cpc = { 1967 .reason = CP_UMOUNT | CP_TRIMMED, 1968 }; 1969 stat_inc_cp_call_count(sbi, TOTAL_CALL); 1970 err = f2fs_write_checkpoint(sbi, &cpc); 1971 } 1972 1973 /* 1974 * normally superblock is clean, so we need to release this. 1975 * In addition, EIO will skip do checkpoint, we need this as well. 1976 */ 1977 f2fs_release_ino_entry(sbi, true); 1978 1979 f2fs_leave_shrinker(sbi); 1980 mutex_unlock(&sbi->umount_mutex); 1981 1982 /* our cp_error case, we can wait for any writeback page */ 1983 f2fs_flush_merged_writes(sbi); 1984 1985 f2fs_wait_on_all_pages(sbi, F2FS_WB_CP_DATA); 1986 1987 if (err || f2fs_cp_error(sbi)) { 1988 truncate_inode_pages_final(NODE_MAPPING(sbi)); 1989 truncate_inode_pages_final(META_MAPPING(sbi)); 1990 } 1991 1992 f2fs_bug_on(sbi, sbi->fsync_node_num); 1993 1994 f2fs_destroy_compress_inode(sbi); 1995 1996 iput(sbi->node_inode); 1997 sbi->node_inode = NULL; 1998 1999 iput(sbi->meta_inode); 2000 sbi->meta_inode = NULL; 2001 2002 /* Should check the page counts after dropping all node/meta pages */ 2003 for (i = 0; i < NR_COUNT_TYPE; i++) { 2004 if (!get_pages(sbi, i)) 2005 continue; 2006 f2fs_err(sbi, "detect filesystem reference count leak during " 2007 "umount, type: %d, count: %lld", i, get_pages(sbi, i)); 2008 f2fs_bug_on(sbi, 1); 2009 } 2010 2011 /* 2012 * iput() can update stat information, if f2fs_write_checkpoint() 2013 * above failed with error. 2014 */ 2015 f2fs_destroy_stats(sbi); 2016 2017 /* destroy f2fs internal modules */ 2018 f2fs_destroy_node_manager(sbi); 2019 f2fs_destroy_segment_manager(sbi); 2020 2021 /* flush s_error_work before sbi destroy */ 2022 flush_work(&sbi->s_error_work); 2023 2024 f2fs_destroy_post_read_wq(sbi); 2025 2026 kvfree(sbi->ckpt); 2027 2028 kfree(sbi->raw_super); 2029 2030 f2fs_destroy_page_array_cache(sbi); 2031 #ifdef CONFIG_QUOTA 2032 for (i = 0; i < MAXQUOTAS; i++) 2033 kfree(F2FS_OPTION(sbi).s_qf_names[i]); 2034 #endif 2035 fscrypt_free_dummy_policy(&F2FS_OPTION(sbi).dummy_enc_policy); 2036 destroy_percpu_info(sbi); 2037 f2fs_destroy_iostat(sbi); 2038 for (i = 0; i < NR_PAGE_TYPE; i++) 2039 kfree(sbi->write_io[i]); 2040 #if IS_ENABLED(CONFIG_UNICODE) 2041 utf8_unload(sb->s_encoding); 2042 #endif 2043 } 2044 2045 int f2fs_sync_fs(struct super_block *sb, int sync) 2046 { 2047 struct f2fs_sb_info *sbi = F2FS_SB(sb); 2048 int err = 0; 2049 2050 if (unlikely(f2fs_cp_error(sbi))) 2051 return 0; 2052 if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED))) 2053 return 0; 2054 2055 trace_f2fs_sync_fs(sb, sync); 2056 2057 if (unlikely(is_sbi_flag_set(sbi, SBI_POR_DOING))) 2058 return -EAGAIN; 2059 2060 if (sync) { 2061 stat_inc_cp_call_count(sbi, TOTAL_CALL); 2062 err = f2fs_issue_checkpoint(sbi); 2063 } 2064 2065 return err; 2066 } 2067 2068 static int f2fs_freeze(struct super_block *sb) 2069 { 2070 struct f2fs_sb_info *sbi = F2FS_SB(sb); 2071 2072 if (f2fs_readonly(sb)) 2073 return 0; 2074 2075 /* IO error happened before */ 2076 if (unlikely(f2fs_cp_error(sbi))) 2077 return -EIO; 2078 2079 /* must be clean, since sync_filesystem() was already called */ 2080 if (is_sbi_flag_set(sbi, SBI_IS_DIRTY)) 2081 return -EINVAL; 2082 2083 sbi->umount_lock_holder = current; 2084 2085 /* Let's flush checkpoints and stop the thread. */ 2086 f2fs_flush_ckpt_thread(sbi); 2087 2088 sbi->umount_lock_holder = NULL; 2089 2090 /* to avoid deadlock on f2fs_evict_inode->SB_FREEZE_FS */ 2091 set_sbi_flag(sbi, SBI_IS_FREEZING); 2092 return 0; 2093 } 2094 2095 static int f2fs_unfreeze(struct super_block *sb) 2096 { 2097 struct f2fs_sb_info *sbi = F2FS_SB(sb); 2098 2099 /* 2100 * It will update discard_max_bytes of mounted lvm device to zero 2101 * after creating snapshot on this lvm device, let's drop all 2102 * remained discards. 2103 * We don't need to disable real-time discard because discard_max_bytes 2104 * will recover after removal of snapshot. 2105 */ 2106 if (test_opt(sbi, DISCARD) && !f2fs_hw_support_discard(sbi)) 2107 f2fs_issue_discard_timeout(sbi); 2108 2109 clear_sbi_flag(F2FS_SB(sb), SBI_IS_FREEZING); 2110 return 0; 2111 } 2112 2113 #ifdef CONFIG_QUOTA 2114 static int f2fs_statfs_project(struct super_block *sb, 2115 kprojid_t projid, struct kstatfs *buf) 2116 { 2117 struct kqid qid; 2118 struct dquot *dquot; 2119 u64 limit; 2120 u64 curblock; 2121 2122 qid = make_kqid_projid(projid); 2123 dquot = dqget(sb, qid); 2124 if (IS_ERR(dquot)) 2125 return PTR_ERR(dquot); 2126 spin_lock(&dquot->dq_dqb_lock); 2127 2128 limit = min_not_zero(dquot->dq_dqb.dqb_bsoftlimit, 2129 dquot->dq_dqb.dqb_bhardlimit); 2130 limit >>= sb->s_blocksize_bits; 2131 2132 if (limit) { 2133 uint64_t remaining = 0; 2134 2135 curblock = (dquot->dq_dqb.dqb_curspace + 2136 dquot->dq_dqb.dqb_rsvspace) >> sb->s_blocksize_bits; 2137 if (limit > curblock) 2138 remaining = limit - curblock; 2139 2140 buf->f_blocks = min(buf->f_blocks, limit); 2141 buf->f_bfree = min(buf->f_bfree, remaining); 2142 buf->f_bavail = min(buf->f_bavail, remaining); 2143 } 2144 2145 limit = min_not_zero(dquot->dq_dqb.dqb_isoftlimit, 2146 dquot->dq_dqb.dqb_ihardlimit); 2147 2148 if (limit) { 2149 uint64_t remaining = 0; 2150 2151 if (limit > dquot->dq_dqb.dqb_curinodes) 2152 remaining = limit - dquot->dq_dqb.dqb_curinodes; 2153 2154 buf->f_files = min(buf->f_files, limit); 2155 buf->f_ffree = min(buf->f_ffree, remaining); 2156 } 2157 2158 spin_unlock(&dquot->dq_dqb_lock); 2159 dqput(dquot); 2160 return 0; 2161 } 2162 #endif 2163 2164 static int f2fs_statfs(struct dentry *dentry, struct kstatfs *buf) 2165 { 2166 struct super_block *sb = dentry->d_sb; 2167 struct f2fs_sb_info *sbi = F2FS_SB(sb); 2168 u64 id = huge_encode_dev(sb->s_bdev->bd_dev); 2169 block_t total_count, user_block_count, start_count; 2170 u64 avail_node_count; 2171 unsigned int total_valid_node_count; 2172 2173 total_count = le64_to_cpu(sbi->raw_super->block_count); 2174 start_count = le32_to_cpu(sbi->raw_super->segment0_blkaddr); 2175 buf->f_type = F2FS_SUPER_MAGIC; 2176 buf->f_bsize = sbi->blocksize; 2177 2178 buf->f_blocks = total_count - start_count; 2179 2180 spin_lock(&sbi->stat_lock); 2181 if (sbi->carve_out) 2182 buf->f_blocks -= sbi->current_reserved_blocks; 2183 user_block_count = sbi->user_block_count; 2184 total_valid_node_count = valid_node_count(sbi); 2185 avail_node_count = sbi->total_node_count - F2FS_RESERVED_NODE_NUM; 2186 buf->f_bfree = user_block_count - valid_user_blocks(sbi) - 2187 sbi->current_reserved_blocks; 2188 2189 if (unlikely(buf->f_bfree <= sbi->unusable_block_count)) 2190 buf->f_bfree = 0; 2191 else 2192 buf->f_bfree -= sbi->unusable_block_count; 2193 spin_unlock(&sbi->stat_lock); 2194 2195 if (buf->f_bfree > F2FS_OPTION(sbi).root_reserved_blocks) 2196 buf->f_bavail = buf->f_bfree - 2197 F2FS_OPTION(sbi).root_reserved_blocks; 2198 else 2199 buf->f_bavail = 0; 2200 2201 if (avail_node_count > user_block_count) { 2202 buf->f_files = user_block_count; 2203 buf->f_ffree = buf->f_bavail; 2204 } else { 2205 buf->f_files = avail_node_count; 2206 buf->f_ffree = min(avail_node_count - total_valid_node_count, 2207 buf->f_bavail); 2208 } 2209 2210 buf->f_namelen = F2FS_NAME_LEN; 2211 buf->f_fsid = u64_to_fsid(id); 2212 2213 #ifdef CONFIG_QUOTA 2214 if (is_inode_flag_set(d_inode(dentry), FI_PROJ_INHERIT) && 2215 sb_has_quota_limits_enabled(sb, PRJQUOTA)) { 2216 f2fs_statfs_project(sb, F2FS_I(d_inode(dentry))->i_projid, buf); 2217 } 2218 #endif 2219 return 0; 2220 } 2221 2222 static inline void f2fs_show_quota_options(struct seq_file *seq, 2223 struct super_block *sb) 2224 { 2225 #ifdef CONFIG_QUOTA 2226 struct f2fs_sb_info *sbi = F2FS_SB(sb); 2227 2228 if (F2FS_OPTION(sbi).s_jquota_fmt) { 2229 char *fmtname = ""; 2230 2231 switch (F2FS_OPTION(sbi).s_jquota_fmt) { 2232 case QFMT_VFS_OLD: 2233 fmtname = "vfsold"; 2234 break; 2235 case QFMT_VFS_V0: 2236 fmtname = "vfsv0"; 2237 break; 2238 case QFMT_VFS_V1: 2239 fmtname = "vfsv1"; 2240 break; 2241 } 2242 seq_printf(seq, ",jqfmt=%s", fmtname); 2243 } 2244 2245 if (F2FS_OPTION(sbi).s_qf_names[USRQUOTA]) 2246 seq_show_option(seq, "usrjquota", 2247 F2FS_OPTION(sbi).s_qf_names[USRQUOTA]); 2248 2249 if (F2FS_OPTION(sbi).s_qf_names[GRPQUOTA]) 2250 seq_show_option(seq, "grpjquota", 2251 F2FS_OPTION(sbi).s_qf_names[GRPQUOTA]); 2252 2253 if (F2FS_OPTION(sbi).s_qf_names[PRJQUOTA]) 2254 seq_show_option(seq, "prjjquota", 2255 F2FS_OPTION(sbi).s_qf_names[PRJQUOTA]); 2256 #endif 2257 } 2258 2259 #ifdef CONFIG_F2FS_FS_COMPRESSION 2260 static inline void f2fs_show_compress_options(struct seq_file *seq, 2261 struct super_block *sb) 2262 { 2263 struct f2fs_sb_info *sbi = F2FS_SB(sb); 2264 char *algtype = ""; 2265 int i; 2266 2267 if (!f2fs_sb_has_compression(sbi)) 2268 return; 2269 2270 switch (F2FS_OPTION(sbi).compress_algorithm) { 2271 case COMPRESS_LZO: 2272 algtype = "lzo"; 2273 break; 2274 case COMPRESS_LZ4: 2275 algtype = "lz4"; 2276 break; 2277 case COMPRESS_ZSTD: 2278 algtype = "zstd"; 2279 break; 2280 case COMPRESS_LZORLE: 2281 algtype = "lzo-rle"; 2282 break; 2283 } 2284 seq_printf(seq, ",compress_algorithm=%s", algtype); 2285 2286 if (F2FS_OPTION(sbi).compress_level) 2287 seq_printf(seq, ":%d", F2FS_OPTION(sbi).compress_level); 2288 2289 seq_printf(seq, ",compress_log_size=%u", 2290 F2FS_OPTION(sbi).compress_log_size); 2291 2292 for (i = 0; i < F2FS_OPTION(sbi).compress_ext_cnt; i++) { 2293 seq_printf(seq, ",compress_extension=%s", 2294 F2FS_OPTION(sbi).extensions[i]); 2295 } 2296 2297 for (i = 0; i < F2FS_OPTION(sbi).nocompress_ext_cnt; i++) { 2298 seq_printf(seq, ",nocompress_extension=%s", 2299 F2FS_OPTION(sbi).noextensions[i]); 2300 } 2301 2302 if (F2FS_OPTION(sbi).compress_chksum) 2303 seq_puts(seq, ",compress_chksum"); 2304 2305 if (F2FS_OPTION(sbi).compress_mode == COMPR_MODE_FS) 2306 seq_printf(seq, ",compress_mode=%s", "fs"); 2307 else if (F2FS_OPTION(sbi).compress_mode == COMPR_MODE_USER) 2308 seq_printf(seq, ",compress_mode=%s", "user"); 2309 2310 if (test_opt(sbi, COMPRESS_CACHE)) 2311 seq_puts(seq, ",compress_cache"); 2312 } 2313 #endif 2314 2315 static int f2fs_show_options(struct seq_file *seq, struct dentry *root) 2316 { 2317 struct f2fs_sb_info *sbi = F2FS_SB(root->d_sb); 2318 2319 if (F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_SYNC) 2320 seq_printf(seq, ",background_gc=%s", "sync"); 2321 else if (F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_ON) 2322 seq_printf(seq, ",background_gc=%s", "on"); 2323 else if (F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_OFF) 2324 seq_printf(seq, ",background_gc=%s", "off"); 2325 2326 if (test_opt(sbi, GC_MERGE)) 2327 seq_puts(seq, ",gc_merge"); 2328 else 2329 seq_puts(seq, ",nogc_merge"); 2330 2331 if (test_opt(sbi, DISABLE_ROLL_FORWARD)) 2332 seq_puts(seq, ",disable_roll_forward"); 2333 if (test_opt(sbi, NORECOVERY)) 2334 seq_puts(seq, ",norecovery"); 2335 if (test_opt(sbi, DISCARD)) { 2336 seq_puts(seq, ",discard"); 2337 if (F2FS_OPTION(sbi).discard_unit == DISCARD_UNIT_BLOCK) 2338 seq_printf(seq, ",discard_unit=%s", "block"); 2339 else if (F2FS_OPTION(sbi).discard_unit == DISCARD_UNIT_SEGMENT) 2340 seq_printf(seq, ",discard_unit=%s", "segment"); 2341 else if (F2FS_OPTION(sbi).discard_unit == DISCARD_UNIT_SECTION) 2342 seq_printf(seq, ",discard_unit=%s", "section"); 2343 } else { 2344 seq_puts(seq, ",nodiscard"); 2345 } 2346 #ifdef CONFIG_F2FS_FS_XATTR 2347 if (test_opt(sbi, XATTR_USER)) 2348 seq_puts(seq, ",user_xattr"); 2349 else 2350 seq_puts(seq, ",nouser_xattr"); 2351 if (test_opt(sbi, INLINE_XATTR)) 2352 seq_puts(seq, ",inline_xattr"); 2353 else 2354 seq_puts(seq, ",noinline_xattr"); 2355 if (test_opt(sbi, INLINE_XATTR_SIZE)) 2356 seq_printf(seq, ",inline_xattr_size=%u", 2357 F2FS_OPTION(sbi).inline_xattr_size); 2358 #endif 2359 #ifdef CONFIG_F2FS_FS_POSIX_ACL 2360 if (test_opt(sbi, POSIX_ACL)) 2361 seq_puts(seq, ",acl"); 2362 else 2363 seq_puts(seq, ",noacl"); 2364 #endif 2365 if (test_opt(sbi, DISABLE_EXT_IDENTIFY)) 2366 seq_puts(seq, ",disable_ext_identify"); 2367 if (test_opt(sbi, INLINE_DATA)) 2368 seq_puts(seq, ",inline_data"); 2369 else 2370 seq_puts(seq, ",noinline_data"); 2371 if (test_opt(sbi, INLINE_DENTRY)) 2372 seq_puts(seq, ",inline_dentry"); 2373 else 2374 seq_puts(seq, ",noinline_dentry"); 2375 if (test_opt(sbi, FLUSH_MERGE)) 2376 seq_puts(seq, ",flush_merge"); 2377 else 2378 seq_puts(seq, ",noflush_merge"); 2379 if (test_opt(sbi, NOBARRIER)) 2380 seq_puts(seq, ",nobarrier"); 2381 else 2382 seq_puts(seq, ",barrier"); 2383 if (test_opt(sbi, FASTBOOT)) 2384 seq_puts(seq, ",fastboot"); 2385 if (test_opt(sbi, READ_EXTENT_CACHE)) 2386 seq_puts(seq, ",extent_cache"); 2387 else 2388 seq_puts(seq, ",noextent_cache"); 2389 if (test_opt(sbi, AGE_EXTENT_CACHE)) 2390 seq_puts(seq, ",age_extent_cache"); 2391 if (test_opt(sbi, DATA_FLUSH)) 2392 seq_puts(seq, ",data_flush"); 2393 2394 seq_puts(seq, ",mode="); 2395 if (F2FS_OPTION(sbi).fs_mode == FS_MODE_ADAPTIVE) 2396 seq_puts(seq, "adaptive"); 2397 else if (F2FS_OPTION(sbi).fs_mode == FS_MODE_LFS) 2398 seq_puts(seq, "lfs"); 2399 else if (F2FS_OPTION(sbi).fs_mode == FS_MODE_FRAGMENT_SEG) 2400 seq_puts(seq, "fragment:segment"); 2401 else if (F2FS_OPTION(sbi).fs_mode == FS_MODE_FRAGMENT_BLK) 2402 seq_puts(seq, "fragment:block"); 2403 seq_printf(seq, ",active_logs=%u", F2FS_OPTION(sbi).active_logs); 2404 if (test_opt(sbi, RESERVE_ROOT) || test_opt(sbi, RESERVE_NODE)) 2405 seq_printf(seq, ",reserve_root=%u,reserve_node=%u,resuid=%u," 2406 "resgid=%u", 2407 F2FS_OPTION(sbi).root_reserved_blocks, 2408 F2FS_OPTION(sbi).root_reserved_nodes, 2409 from_kuid_munged(&init_user_ns, 2410 F2FS_OPTION(sbi).s_resuid), 2411 from_kgid_munged(&init_user_ns, 2412 F2FS_OPTION(sbi).s_resgid)); 2413 #ifdef CONFIG_F2FS_FAULT_INJECTION 2414 if (test_opt(sbi, FAULT_INJECTION)) { 2415 seq_printf(seq, ",fault_injection=%u", 2416 F2FS_OPTION(sbi).fault_info.inject_rate); 2417 seq_printf(seq, ",fault_type=%u", 2418 F2FS_OPTION(sbi).fault_info.inject_type); 2419 } 2420 #endif 2421 #ifdef CONFIG_QUOTA 2422 if (test_opt(sbi, QUOTA)) 2423 seq_puts(seq, ",quota"); 2424 if (test_opt(sbi, USRQUOTA)) 2425 seq_puts(seq, ",usrquota"); 2426 if (test_opt(sbi, GRPQUOTA)) 2427 seq_puts(seq, ",grpquota"); 2428 if (test_opt(sbi, PRJQUOTA)) 2429 seq_puts(seq, ",prjquota"); 2430 #endif 2431 f2fs_show_quota_options(seq, sbi->sb); 2432 2433 fscrypt_show_test_dummy_encryption(seq, ',', sbi->sb); 2434 2435 if (sbi->sb->s_flags & SB_INLINECRYPT) 2436 seq_puts(seq, ",inlinecrypt"); 2437 2438 if (F2FS_OPTION(sbi).alloc_mode == ALLOC_MODE_DEFAULT) 2439 seq_printf(seq, ",alloc_mode=%s", "default"); 2440 else if (F2FS_OPTION(sbi).alloc_mode == ALLOC_MODE_REUSE) 2441 seq_printf(seq, ",alloc_mode=%s", "reuse"); 2442 2443 if (test_opt(sbi, DISABLE_CHECKPOINT)) 2444 seq_printf(seq, ",checkpoint=disable:%u", 2445 F2FS_OPTION(sbi).unusable_cap); 2446 if (test_opt(sbi, MERGE_CHECKPOINT)) 2447 seq_puts(seq, ",checkpoint_merge"); 2448 else 2449 seq_puts(seq, ",nocheckpoint_merge"); 2450 if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_POSIX) 2451 seq_printf(seq, ",fsync_mode=%s", "posix"); 2452 else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_STRICT) 2453 seq_printf(seq, ",fsync_mode=%s", "strict"); 2454 else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_NOBARRIER) 2455 seq_printf(seq, ",fsync_mode=%s", "nobarrier"); 2456 2457 #ifdef CONFIG_F2FS_FS_COMPRESSION 2458 f2fs_show_compress_options(seq, sbi->sb); 2459 #endif 2460 2461 if (test_opt(sbi, ATGC)) 2462 seq_puts(seq, ",atgc"); 2463 2464 if (F2FS_OPTION(sbi).memory_mode == MEMORY_MODE_NORMAL) 2465 seq_printf(seq, ",memory=%s", "normal"); 2466 else if (F2FS_OPTION(sbi).memory_mode == MEMORY_MODE_LOW) 2467 seq_printf(seq, ",memory=%s", "low"); 2468 2469 if (F2FS_OPTION(sbi).errors == MOUNT_ERRORS_READONLY) 2470 seq_printf(seq, ",errors=%s", "remount-ro"); 2471 else if (F2FS_OPTION(sbi).errors == MOUNT_ERRORS_CONTINUE) 2472 seq_printf(seq, ",errors=%s", "continue"); 2473 else if (F2FS_OPTION(sbi).errors == MOUNT_ERRORS_PANIC) 2474 seq_printf(seq, ",errors=%s", "panic"); 2475 2476 if (test_opt(sbi, NAT_BITS)) 2477 seq_puts(seq, ",nat_bits"); 2478 2479 if (F2FS_OPTION(sbi).lookup_mode == LOOKUP_PERF) 2480 seq_show_option(seq, "lookup_mode", "perf"); 2481 else if (F2FS_OPTION(sbi).lookup_mode == LOOKUP_COMPAT) 2482 seq_show_option(seq, "lookup_mode", "compat"); 2483 else if (F2FS_OPTION(sbi).lookup_mode == LOOKUP_AUTO) 2484 seq_show_option(seq, "lookup_mode", "auto"); 2485 2486 return 0; 2487 } 2488 2489 static void default_options(struct f2fs_sb_info *sbi, bool remount) 2490 { 2491 /* init some FS parameters */ 2492 if (!remount) { 2493 set_opt(sbi, READ_EXTENT_CACHE); 2494 clear_opt(sbi, DISABLE_CHECKPOINT); 2495 2496 if (f2fs_hw_support_discard(sbi) || f2fs_hw_should_discard(sbi)) 2497 set_opt(sbi, DISCARD); 2498 2499 if (f2fs_sb_has_blkzoned(sbi)) 2500 F2FS_OPTION(sbi).discard_unit = DISCARD_UNIT_SECTION; 2501 else 2502 F2FS_OPTION(sbi).discard_unit = DISCARD_UNIT_BLOCK; 2503 } 2504 2505 if (f2fs_sb_has_readonly(sbi)) 2506 F2FS_OPTION(sbi).active_logs = NR_CURSEG_RO_TYPE; 2507 else 2508 F2FS_OPTION(sbi).active_logs = NR_CURSEG_PERSIST_TYPE; 2509 2510 F2FS_OPTION(sbi).inline_xattr_size = DEFAULT_INLINE_XATTR_ADDRS; 2511 if (le32_to_cpu(F2FS_RAW_SUPER(sbi)->segment_count_main) <= 2512 SMALL_VOLUME_SEGMENTS) 2513 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_REUSE; 2514 else 2515 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_DEFAULT; 2516 F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_POSIX; 2517 F2FS_OPTION(sbi).s_resuid = make_kuid(&init_user_ns, F2FS_DEF_RESUID); 2518 F2FS_OPTION(sbi).s_resgid = make_kgid(&init_user_ns, F2FS_DEF_RESGID); 2519 if (f2fs_sb_has_compression(sbi)) { 2520 F2FS_OPTION(sbi).compress_algorithm = COMPRESS_LZ4; 2521 F2FS_OPTION(sbi).compress_log_size = MIN_COMPRESS_LOG_SIZE; 2522 F2FS_OPTION(sbi).compress_ext_cnt = 0; 2523 F2FS_OPTION(sbi).compress_mode = COMPR_MODE_FS; 2524 } 2525 F2FS_OPTION(sbi).bggc_mode = BGGC_MODE_ON; 2526 F2FS_OPTION(sbi).memory_mode = MEMORY_MODE_NORMAL; 2527 F2FS_OPTION(sbi).errors = MOUNT_ERRORS_CONTINUE; 2528 2529 set_opt(sbi, INLINE_XATTR); 2530 set_opt(sbi, INLINE_DATA); 2531 set_opt(sbi, INLINE_DENTRY); 2532 set_opt(sbi, MERGE_CHECKPOINT); 2533 set_opt(sbi, LAZYTIME); 2534 F2FS_OPTION(sbi).unusable_cap = 0; 2535 if (!f2fs_is_readonly(sbi)) 2536 set_opt(sbi, FLUSH_MERGE); 2537 if (f2fs_sb_has_blkzoned(sbi)) 2538 F2FS_OPTION(sbi).fs_mode = FS_MODE_LFS; 2539 else 2540 F2FS_OPTION(sbi).fs_mode = FS_MODE_ADAPTIVE; 2541 2542 #ifdef CONFIG_F2FS_FS_XATTR 2543 set_opt(sbi, XATTR_USER); 2544 #endif 2545 #ifdef CONFIG_F2FS_FS_POSIX_ACL 2546 set_opt(sbi, POSIX_ACL); 2547 #endif 2548 2549 f2fs_build_fault_attr(sbi, 0, 0, FAULT_ALL); 2550 2551 F2FS_OPTION(sbi).lookup_mode = LOOKUP_PERF; 2552 } 2553 2554 #ifdef CONFIG_QUOTA 2555 static int f2fs_enable_quotas(struct super_block *sb); 2556 #endif 2557 2558 static int f2fs_disable_checkpoint(struct f2fs_sb_info *sbi) 2559 { 2560 unsigned int s_flags = sbi->sb->s_flags; 2561 struct cp_control cpc; 2562 unsigned int gc_mode = sbi->gc_mode; 2563 int err = 0; 2564 int ret; 2565 block_t unusable; 2566 2567 if (s_flags & SB_RDONLY) { 2568 f2fs_err(sbi, "checkpoint=disable on readonly fs"); 2569 return -EINVAL; 2570 } 2571 sbi->sb->s_flags |= SB_ACTIVE; 2572 2573 /* check if we need more GC first */ 2574 unusable = f2fs_get_unusable_blocks(sbi); 2575 if (!f2fs_disable_cp_again(sbi, unusable)) 2576 goto skip_gc; 2577 2578 f2fs_update_time(sbi, DISABLE_TIME); 2579 2580 sbi->gc_mode = GC_URGENT_HIGH; 2581 2582 while (!f2fs_time_over(sbi, DISABLE_TIME)) { 2583 struct f2fs_gc_control gc_control = { 2584 .victim_segno = NULL_SEGNO, 2585 .init_gc_type = FG_GC, 2586 .should_migrate_blocks = false, 2587 .err_gc_skipped = true, 2588 .no_bg_gc = true, 2589 .nr_free_secs = 1 }; 2590 2591 f2fs_down_write(&sbi->gc_lock); 2592 stat_inc_gc_call_count(sbi, FOREGROUND); 2593 err = f2fs_gc(sbi, &gc_control); 2594 if (err == -ENODATA) { 2595 err = 0; 2596 break; 2597 } 2598 if (err && err != -EAGAIN) 2599 break; 2600 } 2601 2602 ret = sync_filesystem(sbi->sb); 2603 if (ret || err) { 2604 err = ret ? ret : err; 2605 goto restore_flag; 2606 } 2607 2608 unusable = f2fs_get_unusable_blocks(sbi); 2609 if (f2fs_disable_cp_again(sbi, unusable)) { 2610 err = -EAGAIN; 2611 goto restore_flag; 2612 } 2613 2614 skip_gc: 2615 f2fs_down_write(&sbi->gc_lock); 2616 cpc.reason = CP_PAUSE; 2617 set_sbi_flag(sbi, SBI_CP_DISABLED); 2618 stat_inc_cp_call_count(sbi, TOTAL_CALL); 2619 err = f2fs_write_checkpoint(sbi, &cpc); 2620 if (err) 2621 goto out_unlock; 2622 2623 spin_lock(&sbi->stat_lock); 2624 sbi->unusable_block_count = unusable; 2625 spin_unlock(&sbi->stat_lock); 2626 2627 out_unlock: 2628 f2fs_up_write(&sbi->gc_lock); 2629 restore_flag: 2630 sbi->gc_mode = gc_mode; 2631 sbi->sb->s_flags = s_flags; /* Restore SB_RDONLY status */ 2632 f2fs_info(sbi, "f2fs_disable_checkpoint() finish, err:%d", err); 2633 return err; 2634 } 2635 2636 static int f2fs_enable_checkpoint(struct f2fs_sb_info *sbi) 2637 { 2638 unsigned int nr_pages = get_pages(sbi, F2FS_DIRTY_DATA) / 16; 2639 long long start, writeback, lock, sync_inode, end; 2640 int ret; 2641 2642 f2fs_info(sbi, "%s start, meta: %lld, node: %lld, data: %lld", 2643 __func__, 2644 get_pages(sbi, F2FS_DIRTY_META), 2645 get_pages(sbi, F2FS_DIRTY_NODES), 2646 get_pages(sbi, F2FS_DIRTY_DATA)); 2647 2648 f2fs_update_time(sbi, ENABLE_TIME); 2649 2650 start = ktime_get(); 2651 2652 /* we should flush all the data to keep data consistency */ 2653 while (get_pages(sbi, F2FS_DIRTY_DATA)) { 2654 writeback_inodes_sb_nr(sbi->sb, nr_pages, WB_REASON_SYNC); 2655 f2fs_io_schedule_timeout(DEFAULT_SCHEDULE_TIMEOUT); 2656 2657 if (f2fs_time_over(sbi, ENABLE_TIME)) 2658 break; 2659 } 2660 writeback = ktime_get(); 2661 2662 f2fs_down_write(&sbi->cp_enable_rwsem); 2663 2664 lock = ktime_get(); 2665 2666 if (get_pages(sbi, F2FS_DIRTY_DATA)) 2667 sync_inodes_sb(sbi->sb); 2668 2669 if (unlikely(get_pages(sbi, F2FS_DIRTY_DATA))) 2670 f2fs_warn(sbi, "%s: has some unwritten data: %lld", 2671 __func__, get_pages(sbi, F2FS_DIRTY_DATA)); 2672 2673 sync_inode = ktime_get(); 2674 2675 f2fs_down_write(&sbi->gc_lock); 2676 f2fs_dirty_to_prefree(sbi); 2677 2678 clear_sbi_flag(sbi, SBI_CP_DISABLED); 2679 set_sbi_flag(sbi, SBI_IS_DIRTY); 2680 f2fs_up_write(&sbi->gc_lock); 2681 2682 f2fs_info(sbi, "%s sync_fs, meta: %lld, imeta: %lld, node: %lld, dents: %lld, qdata: %lld", 2683 __func__, 2684 get_pages(sbi, F2FS_DIRTY_META), 2685 get_pages(sbi, F2FS_DIRTY_IMETA), 2686 get_pages(sbi, F2FS_DIRTY_NODES), 2687 get_pages(sbi, F2FS_DIRTY_DENTS), 2688 get_pages(sbi, F2FS_DIRTY_QDATA)); 2689 ret = f2fs_sync_fs(sbi->sb, 1); 2690 if (ret) 2691 f2fs_err(sbi, "%s sync_fs failed, ret: %d", __func__, ret); 2692 2693 /* Let's ensure there's no pending checkpoint anymore */ 2694 f2fs_flush_ckpt_thread(sbi); 2695 2696 f2fs_up_write(&sbi->cp_enable_rwsem); 2697 2698 end = ktime_get(); 2699 2700 f2fs_info(sbi, "%s end, writeback:%llu, " 2701 "lock:%llu, sync_inode:%llu, sync_fs:%llu", 2702 __func__, 2703 ktime_ms_delta(writeback, start), 2704 ktime_ms_delta(lock, writeback), 2705 ktime_ms_delta(sync_inode, lock), 2706 ktime_ms_delta(end, sync_inode)); 2707 return ret; 2708 } 2709 2710 static int __f2fs_remount(struct fs_context *fc, struct super_block *sb) 2711 { 2712 struct f2fs_sb_info *sbi = F2FS_SB(sb); 2713 struct f2fs_mount_info org_mount_opt; 2714 unsigned long old_sb_flags; 2715 unsigned int flags = fc->sb_flags; 2716 int err; 2717 bool need_restart_gc = false, need_stop_gc = false; 2718 bool need_restart_flush = false, need_stop_flush = false; 2719 bool need_restart_discard = false, need_stop_discard = false; 2720 bool need_enable_checkpoint = false, need_disable_checkpoint = false; 2721 bool no_read_extent_cache = !test_opt(sbi, READ_EXTENT_CACHE); 2722 bool no_age_extent_cache = !test_opt(sbi, AGE_EXTENT_CACHE); 2723 bool enable_checkpoint = !test_opt(sbi, DISABLE_CHECKPOINT); 2724 bool no_atgc = !test_opt(sbi, ATGC); 2725 bool no_discard = !test_opt(sbi, DISCARD); 2726 bool no_compress_cache = !test_opt(sbi, COMPRESS_CACHE); 2727 bool block_unit_discard = f2fs_block_unit_discard(sbi); 2728 bool no_nat_bits = !test_opt(sbi, NAT_BITS); 2729 #ifdef CONFIG_QUOTA 2730 int i, j; 2731 #endif 2732 2733 /* 2734 * Save the old mount options in case we 2735 * need to restore them. 2736 */ 2737 org_mount_opt = sbi->mount_opt; 2738 old_sb_flags = sb->s_flags; 2739 2740 sbi->umount_lock_holder = current; 2741 2742 #ifdef CONFIG_QUOTA 2743 org_mount_opt.s_jquota_fmt = F2FS_OPTION(sbi).s_jquota_fmt; 2744 for (i = 0; i < MAXQUOTAS; i++) { 2745 if (F2FS_OPTION(sbi).s_qf_names[i]) { 2746 org_mount_opt.s_qf_names[i] = 2747 kstrdup(F2FS_OPTION(sbi).s_qf_names[i], 2748 GFP_KERNEL); 2749 if (!org_mount_opt.s_qf_names[i]) { 2750 for (j = 0; j < i; j++) 2751 kfree(org_mount_opt.s_qf_names[j]); 2752 return -ENOMEM; 2753 } 2754 } else { 2755 org_mount_opt.s_qf_names[i] = NULL; 2756 } 2757 } 2758 #endif 2759 2760 /* recover superblocks we couldn't write due to previous RO mount */ 2761 if (!(flags & SB_RDONLY) && is_sbi_flag_set(sbi, SBI_NEED_SB_WRITE)) { 2762 err = f2fs_commit_super(sbi, false); 2763 f2fs_info(sbi, "Try to recover all the superblocks, ret: %d", 2764 err); 2765 if (!err) 2766 clear_sbi_flag(sbi, SBI_NEED_SB_WRITE); 2767 } 2768 2769 default_options(sbi, true); 2770 2771 err = f2fs_check_opt_consistency(fc, sb); 2772 if (err) 2773 goto restore_opts; 2774 2775 f2fs_apply_options(fc, sb); 2776 2777 err = f2fs_sanity_check_options(sbi, true); 2778 if (err) 2779 goto restore_opts; 2780 2781 /* flush outstanding errors before changing fs state */ 2782 flush_work(&sbi->s_error_work); 2783 2784 /* 2785 * Previous and new state of filesystem is RO, 2786 * so skip checking GC and FLUSH_MERGE conditions. 2787 */ 2788 if (f2fs_readonly(sb) && (flags & SB_RDONLY)) 2789 goto skip; 2790 2791 if (f2fs_dev_is_readonly(sbi) && !(flags & SB_RDONLY)) { 2792 err = -EROFS; 2793 goto restore_opts; 2794 } 2795 2796 #ifdef CONFIG_QUOTA 2797 if (!f2fs_readonly(sb) && (flags & SB_RDONLY)) { 2798 err = dquot_suspend(sb, -1); 2799 if (err < 0) 2800 goto restore_opts; 2801 } else if (f2fs_readonly(sb) && !(flags & SB_RDONLY)) { 2802 /* dquot_resume needs RW */ 2803 sb->s_flags &= ~SB_RDONLY; 2804 if (sb_any_quota_suspended(sb)) { 2805 dquot_resume(sb, -1); 2806 } else if (f2fs_sb_has_quota_ino(sbi)) { 2807 err = f2fs_enable_quotas(sb); 2808 if (err) 2809 goto restore_opts; 2810 } 2811 } 2812 #endif 2813 /* disallow enable atgc dynamically */ 2814 if (no_atgc == !!test_opt(sbi, ATGC)) { 2815 err = -EINVAL; 2816 f2fs_warn(sbi, "switch atgc option is not allowed"); 2817 goto restore_opts; 2818 } 2819 2820 /* disallow enable/disable extent_cache dynamically */ 2821 if (no_read_extent_cache == !!test_opt(sbi, READ_EXTENT_CACHE)) { 2822 err = -EINVAL; 2823 f2fs_warn(sbi, "switch extent_cache option is not allowed"); 2824 goto restore_opts; 2825 } 2826 /* disallow enable/disable age extent_cache dynamically */ 2827 if (no_age_extent_cache == !!test_opt(sbi, AGE_EXTENT_CACHE)) { 2828 err = -EINVAL; 2829 f2fs_warn(sbi, "switch age_extent_cache option is not allowed"); 2830 goto restore_opts; 2831 } 2832 2833 if (no_compress_cache == !!test_opt(sbi, COMPRESS_CACHE)) { 2834 err = -EINVAL; 2835 f2fs_warn(sbi, "switch compress_cache option is not allowed"); 2836 goto restore_opts; 2837 } 2838 2839 if (block_unit_discard != f2fs_block_unit_discard(sbi)) { 2840 err = -EINVAL; 2841 f2fs_warn(sbi, "switch discard_unit option is not allowed"); 2842 goto restore_opts; 2843 } 2844 2845 if (no_nat_bits == !!test_opt(sbi, NAT_BITS)) { 2846 err = -EINVAL; 2847 f2fs_warn(sbi, "switch nat_bits option is not allowed"); 2848 goto restore_opts; 2849 } 2850 2851 if ((flags & SB_RDONLY) && test_opt(sbi, DISABLE_CHECKPOINT)) { 2852 err = -EINVAL; 2853 f2fs_warn(sbi, "disabling checkpoint not compatible with read-only"); 2854 goto restore_opts; 2855 } 2856 2857 /* 2858 * We stop the GC thread if FS is mounted as RO 2859 * or if background_gc = off is passed in mount 2860 * option. Also sync the filesystem. 2861 */ 2862 if ((flags & SB_RDONLY) || 2863 (F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_OFF && 2864 !test_opt(sbi, GC_MERGE))) { 2865 if (sbi->gc_thread) { 2866 f2fs_stop_gc_thread(sbi); 2867 need_restart_gc = true; 2868 } 2869 } else if (!sbi->gc_thread) { 2870 err = f2fs_start_gc_thread(sbi); 2871 if (err) 2872 goto restore_opts; 2873 need_stop_gc = true; 2874 } 2875 2876 if (flags & SB_RDONLY) { 2877 sync_inodes_sb(sb); 2878 2879 set_sbi_flag(sbi, SBI_IS_DIRTY); 2880 set_sbi_flag(sbi, SBI_IS_CLOSE); 2881 f2fs_sync_fs(sb, 1); 2882 clear_sbi_flag(sbi, SBI_IS_CLOSE); 2883 } 2884 2885 /* 2886 * We stop issue flush thread if FS is mounted as RO 2887 * or if flush_merge is not passed in mount option. 2888 */ 2889 if ((flags & SB_RDONLY) || !test_opt(sbi, FLUSH_MERGE)) { 2890 clear_opt(sbi, FLUSH_MERGE); 2891 f2fs_destroy_flush_cmd_control(sbi, false); 2892 need_restart_flush = true; 2893 } else { 2894 err = f2fs_create_flush_cmd_control(sbi); 2895 if (err) 2896 goto restore_gc; 2897 need_stop_flush = true; 2898 } 2899 2900 if (no_discard == !!test_opt(sbi, DISCARD)) { 2901 if (test_opt(sbi, DISCARD)) { 2902 err = f2fs_start_discard_thread(sbi); 2903 if (err) 2904 goto restore_flush; 2905 need_stop_discard = true; 2906 } else { 2907 f2fs_stop_discard_thread(sbi); 2908 f2fs_issue_discard_timeout(sbi); 2909 need_restart_discard = true; 2910 } 2911 } 2912 2913 adjust_unusable_cap_perc(sbi); 2914 if (enable_checkpoint == !!test_opt(sbi, DISABLE_CHECKPOINT)) { 2915 if (test_opt(sbi, DISABLE_CHECKPOINT)) { 2916 err = f2fs_disable_checkpoint(sbi); 2917 if (err) 2918 goto restore_discard; 2919 need_enable_checkpoint = true; 2920 } else { 2921 err = f2fs_enable_checkpoint(sbi); 2922 if (err) 2923 goto restore_discard; 2924 need_disable_checkpoint = true; 2925 } 2926 } 2927 2928 /* 2929 * Place this routine at the end, since a new checkpoint would be 2930 * triggered while remount and we need to take care of it before 2931 * returning from remount. 2932 */ 2933 if ((flags & SB_RDONLY) || test_opt(sbi, DISABLE_CHECKPOINT) || 2934 !test_opt(sbi, MERGE_CHECKPOINT)) { 2935 f2fs_stop_ckpt_thread(sbi); 2936 } else { 2937 /* Flush if the previous checkpoint, if exists. */ 2938 f2fs_flush_ckpt_thread(sbi); 2939 2940 err = f2fs_start_ckpt_thread(sbi); 2941 if (err) { 2942 f2fs_err(sbi, 2943 "Failed to start F2FS issue_checkpoint_thread (%d)", 2944 err); 2945 goto restore_checkpoint; 2946 } 2947 } 2948 2949 skip: 2950 #ifdef CONFIG_QUOTA 2951 /* Release old quota file names */ 2952 for (i = 0; i < MAXQUOTAS; i++) 2953 kfree(org_mount_opt.s_qf_names[i]); 2954 #endif 2955 /* Update the POSIXACL Flag */ 2956 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) | 2957 (test_opt(sbi, POSIX_ACL) ? SB_POSIXACL : 0); 2958 2959 limit_reserve_root(sbi); 2960 fc->sb_flags = (flags & ~SB_LAZYTIME) | (sb->s_flags & SB_LAZYTIME); 2961 2962 sbi->umount_lock_holder = NULL; 2963 return 0; 2964 restore_checkpoint: 2965 if (need_enable_checkpoint) { 2966 if (f2fs_enable_checkpoint(sbi)) 2967 f2fs_warn(sbi, "checkpoint has not been enabled"); 2968 } else if (need_disable_checkpoint) { 2969 if (f2fs_disable_checkpoint(sbi)) 2970 f2fs_warn(sbi, "checkpoint has not been disabled"); 2971 } 2972 restore_discard: 2973 if (need_restart_discard) { 2974 if (f2fs_start_discard_thread(sbi)) 2975 f2fs_warn(sbi, "discard has been stopped"); 2976 } else if (need_stop_discard) { 2977 f2fs_stop_discard_thread(sbi); 2978 } 2979 restore_flush: 2980 if (need_restart_flush) { 2981 if (f2fs_create_flush_cmd_control(sbi)) 2982 f2fs_warn(sbi, "background flush thread has stopped"); 2983 } else if (need_stop_flush) { 2984 clear_opt(sbi, FLUSH_MERGE); 2985 f2fs_destroy_flush_cmd_control(sbi, false); 2986 } 2987 restore_gc: 2988 if (need_restart_gc) { 2989 if (f2fs_start_gc_thread(sbi)) 2990 f2fs_warn(sbi, "background gc thread has stopped"); 2991 } else if (need_stop_gc) { 2992 f2fs_stop_gc_thread(sbi); 2993 } 2994 restore_opts: 2995 #ifdef CONFIG_QUOTA 2996 F2FS_OPTION(sbi).s_jquota_fmt = org_mount_opt.s_jquota_fmt; 2997 for (i = 0; i < MAXQUOTAS; i++) { 2998 kfree(F2FS_OPTION(sbi).s_qf_names[i]); 2999 F2FS_OPTION(sbi).s_qf_names[i] = org_mount_opt.s_qf_names[i]; 3000 } 3001 #endif 3002 sbi->mount_opt = org_mount_opt; 3003 sb->s_flags = old_sb_flags; 3004 3005 sbi->umount_lock_holder = NULL; 3006 return err; 3007 } 3008 3009 static void f2fs_shutdown(struct super_block *sb) 3010 { 3011 f2fs_do_shutdown(F2FS_SB(sb), F2FS_GOING_DOWN_NOSYNC, false, false); 3012 } 3013 3014 #ifdef CONFIG_QUOTA 3015 static bool f2fs_need_recovery(struct f2fs_sb_info *sbi) 3016 { 3017 /* need to recovery orphan */ 3018 if (is_set_ckpt_flags(sbi, CP_ORPHAN_PRESENT_FLAG)) 3019 return true; 3020 /* need to recovery data */ 3021 if (test_opt(sbi, DISABLE_ROLL_FORWARD)) 3022 return false; 3023 if (test_opt(sbi, NORECOVERY)) 3024 return false; 3025 return !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG); 3026 } 3027 3028 static bool f2fs_recover_quota_begin(struct f2fs_sb_info *sbi) 3029 { 3030 bool readonly = f2fs_readonly(sbi->sb); 3031 3032 if (!f2fs_need_recovery(sbi)) 3033 return false; 3034 3035 /* it doesn't need to check f2fs_sb_has_readonly() */ 3036 if (f2fs_hw_is_readonly(sbi)) 3037 return false; 3038 3039 if (readonly) { 3040 sbi->sb->s_flags &= ~SB_RDONLY; 3041 set_sbi_flag(sbi, SBI_IS_WRITABLE); 3042 } 3043 3044 /* 3045 * Turn on quotas which were not enabled for read-only mounts if 3046 * filesystem has quota feature, so that they are updated correctly. 3047 */ 3048 return f2fs_enable_quota_files(sbi, readonly); 3049 } 3050 3051 static void f2fs_recover_quota_end(struct f2fs_sb_info *sbi, 3052 bool quota_enabled) 3053 { 3054 if (quota_enabled) 3055 f2fs_quota_off_umount(sbi->sb); 3056 3057 if (is_sbi_flag_set(sbi, SBI_IS_WRITABLE)) { 3058 clear_sbi_flag(sbi, SBI_IS_WRITABLE); 3059 sbi->sb->s_flags |= SB_RDONLY; 3060 } 3061 } 3062 3063 /* Read data from quotafile */ 3064 static ssize_t f2fs_quota_read(struct super_block *sb, int type, char *data, 3065 size_t len, loff_t off) 3066 { 3067 struct inode *inode = sb_dqopt(sb)->files[type]; 3068 struct address_space *mapping = inode->i_mapping; 3069 int tocopy; 3070 size_t toread; 3071 loff_t i_size = i_size_read(inode); 3072 3073 if (off > i_size) 3074 return 0; 3075 3076 if (off + len > i_size) 3077 len = i_size - off; 3078 toread = len; 3079 while (toread > 0) { 3080 struct folio *folio; 3081 size_t offset; 3082 3083 repeat: 3084 folio = mapping_read_folio_gfp(mapping, off >> PAGE_SHIFT, 3085 GFP_NOFS); 3086 if (IS_ERR(folio)) { 3087 if (PTR_ERR(folio) == -ENOMEM) { 3088 memalloc_retry_wait(GFP_NOFS); 3089 goto repeat; 3090 } 3091 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR); 3092 return PTR_ERR(folio); 3093 } 3094 offset = offset_in_folio(folio, off); 3095 tocopy = min(folio_size(folio) - offset, toread); 3096 3097 folio_lock(folio); 3098 3099 if (unlikely(folio->mapping != mapping)) { 3100 f2fs_folio_put(folio, true); 3101 goto repeat; 3102 } 3103 3104 /* 3105 * should never happen, just leave f2fs_bug_on() here to catch 3106 * any potential bug. 3107 */ 3108 f2fs_bug_on(F2FS_SB(sb), !folio_test_uptodate(folio)); 3109 3110 memcpy_from_folio(data, folio, offset, tocopy); 3111 f2fs_folio_put(folio, true); 3112 3113 toread -= tocopy; 3114 data += tocopy; 3115 off += tocopy; 3116 } 3117 return len; 3118 } 3119 3120 /* Write to quotafile */ 3121 static ssize_t f2fs_quota_write(struct super_block *sb, int type, 3122 const char *data, size_t len, loff_t off) 3123 { 3124 struct inode *inode = sb_dqopt(sb)->files[type]; 3125 struct address_space *mapping = inode->i_mapping; 3126 const struct address_space_operations *a_ops = mapping->a_ops; 3127 int offset = off & (sb->s_blocksize - 1); 3128 size_t towrite = len; 3129 struct folio *folio; 3130 void *fsdata = NULL; 3131 int err = 0; 3132 int tocopy; 3133 3134 while (towrite > 0) { 3135 tocopy = min_t(unsigned long, sb->s_blocksize - offset, 3136 towrite); 3137 retry: 3138 err = a_ops->write_begin(NULL, mapping, off, tocopy, 3139 &folio, &fsdata); 3140 if (unlikely(err)) { 3141 if (err == -ENOMEM) { 3142 memalloc_retry_wait(GFP_NOFS); 3143 goto retry; 3144 } 3145 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR); 3146 break; 3147 } 3148 3149 memcpy_to_folio(folio, offset_in_folio(folio, off), data, tocopy); 3150 3151 a_ops->write_end(NULL, mapping, off, tocopy, tocopy, 3152 folio, fsdata); 3153 offset = 0; 3154 towrite -= tocopy; 3155 off += tocopy; 3156 data += tocopy; 3157 cond_resched(); 3158 } 3159 3160 if (len == towrite) 3161 return err; 3162 inode_set_mtime_to_ts(inode, inode_set_ctime_current(inode)); 3163 f2fs_mark_inode_dirty_sync(inode, false); 3164 return len - towrite; 3165 } 3166 3167 int f2fs_dquot_initialize(struct inode *inode) 3168 { 3169 if (time_to_inject(F2FS_I_SB(inode), FAULT_DQUOT_INIT)) 3170 return -ESRCH; 3171 3172 return dquot_initialize(inode); 3173 } 3174 3175 static struct dquot __rcu **f2fs_get_dquots(struct inode *inode) 3176 { 3177 return F2FS_I(inode)->i_dquot; 3178 } 3179 3180 static qsize_t *f2fs_get_reserved_space(struct inode *inode) 3181 { 3182 return &F2FS_I(inode)->i_reserved_quota; 3183 } 3184 3185 static int f2fs_quota_on_mount(struct f2fs_sb_info *sbi, int type) 3186 { 3187 if (is_set_ckpt_flags(sbi, CP_QUOTA_NEED_FSCK_FLAG)) { 3188 f2fs_err(sbi, "quota sysfile may be corrupted, skip loading it"); 3189 return 0; 3190 } 3191 3192 return dquot_quota_on_mount(sbi->sb, F2FS_OPTION(sbi).s_qf_names[type], 3193 F2FS_OPTION(sbi).s_jquota_fmt, type); 3194 } 3195 3196 int f2fs_enable_quota_files(struct f2fs_sb_info *sbi, bool rdonly) 3197 { 3198 int enabled = 0; 3199 int i, err; 3200 3201 if (f2fs_sb_has_quota_ino(sbi) && rdonly) { 3202 err = f2fs_enable_quotas(sbi->sb); 3203 if (err) { 3204 f2fs_err(sbi, "Cannot turn on quota_ino: %d", err); 3205 return 0; 3206 } 3207 return 1; 3208 } 3209 3210 for (i = 0; i < MAXQUOTAS; i++) { 3211 if (F2FS_OPTION(sbi).s_qf_names[i]) { 3212 err = f2fs_quota_on_mount(sbi, i); 3213 if (!err) { 3214 enabled = 1; 3215 continue; 3216 } 3217 f2fs_err(sbi, "Cannot turn on quotas: %d on %d", 3218 err, i); 3219 } 3220 } 3221 return enabled; 3222 } 3223 3224 static int f2fs_quota_enable(struct super_block *sb, int type, int format_id, 3225 unsigned int flags) 3226 { 3227 struct inode *qf_inode; 3228 unsigned long qf_inum; 3229 unsigned long qf_flag = F2FS_QUOTA_DEFAULT_FL; 3230 int err; 3231 3232 BUG_ON(!f2fs_sb_has_quota_ino(F2FS_SB(sb))); 3233 3234 qf_inum = f2fs_qf_ino(sb, type); 3235 if (!qf_inum) 3236 return -EPERM; 3237 3238 qf_inode = f2fs_iget(sb, qf_inum); 3239 if (IS_ERR(qf_inode)) { 3240 f2fs_err(F2FS_SB(sb), "Bad quota inode %u:%lu", type, qf_inum); 3241 return PTR_ERR(qf_inode); 3242 } 3243 3244 /* Don't account quota for quota files to avoid recursion */ 3245 inode_lock(qf_inode); 3246 qf_inode->i_flags |= S_NOQUOTA; 3247 3248 if ((F2FS_I(qf_inode)->i_flags & qf_flag) != qf_flag) { 3249 F2FS_I(qf_inode)->i_flags |= qf_flag; 3250 f2fs_set_inode_flags(qf_inode); 3251 } 3252 inode_unlock(qf_inode); 3253 3254 err = dquot_load_quota_inode(qf_inode, type, format_id, flags); 3255 iput(qf_inode); 3256 return err; 3257 } 3258 3259 static int f2fs_enable_quotas(struct super_block *sb) 3260 { 3261 struct f2fs_sb_info *sbi = F2FS_SB(sb); 3262 int type, err = 0; 3263 unsigned long qf_inum; 3264 bool quota_mopt[MAXQUOTAS] = { 3265 test_opt(sbi, USRQUOTA), 3266 test_opt(sbi, GRPQUOTA), 3267 test_opt(sbi, PRJQUOTA), 3268 }; 3269 3270 if (is_set_ckpt_flags(F2FS_SB(sb), CP_QUOTA_NEED_FSCK_FLAG)) { 3271 f2fs_err(sbi, "quota file may be corrupted, skip loading it"); 3272 return 0; 3273 } 3274 3275 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE; 3276 3277 for (type = 0; type < MAXQUOTAS; type++) { 3278 qf_inum = f2fs_qf_ino(sb, type); 3279 if (qf_inum) { 3280 err = f2fs_quota_enable(sb, type, QFMT_VFS_V1, 3281 DQUOT_USAGE_ENABLED | 3282 (quota_mopt[type] ? DQUOT_LIMITS_ENABLED : 0)); 3283 if (err) { 3284 f2fs_err(sbi, "Failed to enable quota tracking (type=%d, err=%d). Please run fsck to fix.", 3285 type, err); 3286 for (type--; type >= 0; type--) 3287 dquot_quota_off(sb, type); 3288 set_sbi_flag(F2FS_SB(sb), 3289 SBI_QUOTA_NEED_REPAIR); 3290 return err; 3291 } 3292 } 3293 } 3294 return 0; 3295 } 3296 3297 static int f2fs_quota_sync_file(struct f2fs_sb_info *sbi, int type) 3298 { 3299 struct quota_info *dqopt = sb_dqopt(sbi->sb); 3300 struct address_space *mapping = dqopt->files[type]->i_mapping; 3301 int ret = 0; 3302 3303 ret = dquot_writeback_dquots(sbi->sb, type); 3304 if (ret) 3305 goto out; 3306 3307 ret = filemap_fdatawrite(mapping); 3308 if (ret) 3309 goto out; 3310 3311 /* if we are using journalled quota */ 3312 if (is_journalled_quota(sbi)) 3313 goto out; 3314 3315 ret = filemap_fdatawait(mapping); 3316 3317 truncate_inode_pages(&dqopt->files[type]->i_data, 0); 3318 out: 3319 if (ret) 3320 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR); 3321 return ret; 3322 } 3323 3324 int f2fs_do_quota_sync(struct super_block *sb, int type) 3325 { 3326 struct f2fs_sb_info *sbi = F2FS_SB(sb); 3327 struct quota_info *dqopt = sb_dqopt(sb); 3328 int cnt; 3329 int ret = 0; 3330 3331 /* 3332 * Now when everything is written we can discard the pagecache so 3333 * that userspace sees the changes. 3334 */ 3335 for (cnt = 0; cnt < MAXQUOTAS; cnt++) { 3336 3337 if (type != -1 && cnt != type) 3338 continue; 3339 3340 if (!sb_has_quota_active(sb, cnt)) 3341 continue; 3342 3343 if (!f2fs_sb_has_quota_ino(sbi)) 3344 inode_lock(dqopt->files[cnt]); 3345 3346 /* 3347 * do_quotactl 3348 * f2fs_quota_sync 3349 * f2fs_down_read(quota_sem) 3350 * dquot_writeback_dquots() 3351 * f2fs_dquot_commit 3352 * block_operation 3353 * f2fs_down_read(quota_sem) 3354 */ 3355 f2fs_lock_op(sbi); 3356 f2fs_down_read(&sbi->quota_sem); 3357 3358 ret = f2fs_quota_sync_file(sbi, cnt); 3359 3360 f2fs_up_read(&sbi->quota_sem); 3361 f2fs_unlock_op(sbi); 3362 3363 if (!f2fs_sb_has_quota_ino(sbi)) 3364 inode_unlock(dqopt->files[cnt]); 3365 3366 if (ret) 3367 break; 3368 } 3369 return ret; 3370 } 3371 3372 static int f2fs_quota_sync(struct super_block *sb, int type) 3373 { 3374 int ret; 3375 3376 F2FS_SB(sb)->umount_lock_holder = current; 3377 ret = f2fs_do_quota_sync(sb, type); 3378 F2FS_SB(sb)->umount_lock_holder = NULL; 3379 return ret; 3380 } 3381 3382 static int f2fs_quota_on(struct super_block *sb, int type, int format_id, 3383 const struct path *path) 3384 { 3385 struct inode *inode; 3386 int err = 0; 3387 3388 /* if quota sysfile exists, deny enabling quota with specific file */ 3389 if (f2fs_sb_has_quota_ino(F2FS_SB(sb))) { 3390 f2fs_err(F2FS_SB(sb), "quota sysfile already exists"); 3391 return -EBUSY; 3392 } 3393 3394 if (path->dentry->d_sb != sb) 3395 return -EXDEV; 3396 3397 F2FS_SB(sb)->umount_lock_holder = current; 3398 3399 err = f2fs_do_quota_sync(sb, type); 3400 if (err) 3401 goto out; 3402 3403 inode = d_inode(path->dentry); 3404 3405 err = filemap_fdatawrite(inode->i_mapping); 3406 if (err) 3407 goto out; 3408 3409 err = filemap_fdatawait(inode->i_mapping); 3410 if (err) 3411 goto out; 3412 3413 err = dquot_quota_on(sb, type, format_id, path); 3414 if (err) 3415 goto out; 3416 3417 inode_lock(inode); 3418 F2FS_I(inode)->i_flags |= F2FS_QUOTA_DEFAULT_FL; 3419 f2fs_set_inode_flags(inode); 3420 inode_unlock(inode); 3421 f2fs_mark_inode_dirty_sync(inode, false); 3422 out: 3423 F2FS_SB(sb)->umount_lock_holder = NULL; 3424 return err; 3425 } 3426 3427 static int __f2fs_quota_off(struct super_block *sb, int type) 3428 { 3429 struct inode *inode = sb_dqopt(sb)->files[type]; 3430 int err; 3431 3432 if (!inode || !igrab(inode)) 3433 return dquot_quota_off(sb, type); 3434 3435 err = f2fs_do_quota_sync(sb, type); 3436 if (err) 3437 goto out_put; 3438 3439 err = dquot_quota_off(sb, type); 3440 if (err || f2fs_sb_has_quota_ino(F2FS_SB(sb))) 3441 goto out_put; 3442 3443 inode_lock(inode); 3444 F2FS_I(inode)->i_flags &= ~F2FS_QUOTA_DEFAULT_FL; 3445 f2fs_set_inode_flags(inode); 3446 inode_unlock(inode); 3447 f2fs_mark_inode_dirty_sync(inode, false); 3448 out_put: 3449 iput(inode); 3450 return err; 3451 } 3452 3453 static int f2fs_quota_off(struct super_block *sb, int type) 3454 { 3455 struct f2fs_sb_info *sbi = F2FS_SB(sb); 3456 int err; 3457 3458 F2FS_SB(sb)->umount_lock_holder = current; 3459 3460 err = __f2fs_quota_off(sb, type); 3461 3462 /* 3463 * quotactl can shutdown journalled quota, result in inconsistence 3464 * between quota record and fs data by following updates, tag the 3465 * flag to let fsck be aware of it. 3466 */ 3467 if (is_journalled_quota(sbi)) 3468 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR); 3469 3470 F2FS_SB(sb)->umount_lock_holder = NULL; 3471 3472 return err; 3473 } 3474 3475 void f2fs_quota_off_umount(struct super_block *sb) 3476 { 3477 int type; 3478 int err; 3479 3480 for (type = 0; type < MAXQUOTAS; type++) { 3481 err = __f2fs_quota_off(sb, type); 3482 if (err) { 3483 int ret = dquot_quota_off(sb, type); 3484 3485 f2fs_err(F2FS_SB(sb), "Fail to turn off disk quota (type: %d, err: %d, ret:%d), Please run fsck to fix it.", 3486 type, err, ret); 3487 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR); 3488 } 3489 } 3490 /* 3491 * In case of checkpoint=disable, we must flush quota blocks. 3492 * This can cause NULL exception for node_inode in end_io, since 3493 * put_super already dropped it. 3494 */ 3495 sync_filesystem(sb); 3496 } 3497 3498 static void f2fs_truncate_quota_inode_pages(struct super_block *sb) 3499 { 3500 struct quota_info *dqopt = sb_dqopt(sb); 3501 int type; 3502 3503 for (type = 0; type < MAXQUOTAS; type++) { 3504 if (!dqopt->files[type]) 3505 continue; 3506 f2fs_inode_synced(dqopt->files[type]); 3507 } 3508 } 3509 3510 static int f2fs_dquot_commit(struct dquot *dquot) 3511 { 3512 struct f2fs_sb_info *sbi = F2FS_SB(dquot->dq_sb); 3513 int ret; 3514 3515 f2fs_down_read_nested(&sbi->quota_sem, SINGLE_DEPTH_NESTING); 3516 ret = dquot_commit(dquot); 3517 if (ret < 0) 3518 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR); 3519 f2fs_up_read(&sbi->quota_sem); 3520 return ret; 3521 } 3522 3523 static int f2fs_dquot_acquire(struct dquot *dquot) 3524 { 3525 struct f2fs_sb_info *sbi = F2FS_SB(dquot->dq_sb); 3526 int ret; 3527 3528 f2fs_down_read(&sbi->quota_sem); 3529 ret = dquot_acquire(dquot); 3530 if (ret < 0) 3531 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR); 3532 f2fs_up_read(&sbi->quota_sem); 3533 return ret; 3534 } 3535 3536 static int f2fs_dquot_release(struct dquot *dquot) 3537 { 3538 struct f2fs_sb_info *sbi = F2FS_SB(dquot->dq_sb); 3539 int ret = dquot_release(dquot); 3540 3541 if (ret < 0) 3542 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR); 3543 return ret; 3544 } 3545 3546 static int f2fs_dquot_mark_dquot_dirty(struct dquot *dquot) 3547 { 3548 struct super_block *sb = dquot->dq_sb; 3549 struct f2fs_sb_info *sbi = F2FS_SB(sb); 3550 int ret = dquot_mark_dquot_dirty(dquot); 3551 3552 /* if we are using journalled quota */ 3553 if (is_journalled_quota(sbi)) 3554 set_sbi_flag(sbi, SBI_QUOTA_NEED_FLUSH); 3555 3556 return ret; 3557 } 3558 3559 static int f2fs_dquot_commit_info(struct super_block *sb, int type) 3560 { 3561 struct f2fs_sb_info *sbi = F2FS_SB(sb); 3562 int ret = dquot_commit_info(sb, type); 3563 3564 if (ret < 0) 3565 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR); 3566 return ret; 3567 } 3568 3569 static int f2fs_get_projid(struct inode *inode, kprojid_t *projid) 3570 { 3571 *projid = F2FS_I(inode)->i_projid; 3572 return 0; 3573 } 3574 3575 static const struct dquot_operations f2fs_quota_operations = { 3576 .get_reserved_space = f2fs_get_reserved_space, 3577 .write_dquot = f2fs_dquot_commit, 3578 .acquire_dquot = f2fs_dquot_acquire, 3579 .release_dquot = f2fs_dquot_release, 3580 .mark_dirty = f2fs_dquot_mark_dquot_dirty, 3581 .write_info = f2fs_dquot_commit_info, 3582 .alloc_dquot = dquot_alloc, 3583 .destroy_dquot = dquot_destroy, 3584 .get_projid = f2fs_get_projid, 3585 .get_next_id = dquot_get_next_id, 3586 }; 3587 3588 static const struct quotactl_ops f2fs_quotactl_ops = { 3589 .quota_on = f2fs_quota_on, 3590 .quota_off = f2fs_quota_off, 3591 .quota_sync = f2fs_quota_sync, 3592 .get_state = dquot_get_state, 3593 .set_info = dquot_set_dqinfo, 3594 .get_dqblk = dquot_get_dqblk, 3595 .set_dqblk = dquot_set_dqblk, 3596 .get_nextdqblk = dquot_get_next_dqblk, 3597 }; 3598 #else 3599 int f2fs_dquot_initialize(struct inode *inode) 3600 { 3601 return 0; 3602 } 3603 3604 int f2fs_do_quota_sync(struct super_block *sb, int type) 3605 { 3606 return 0; 3607 } 3608 3609 void f2fs_quota_off_umount(struct super_block *sb) 3610 { 3611 } 3612 #endif 3613 3614 static const struct super_operations f2fs_sops = { 3615 .alloc_inode = f2fs_alloc_inode, 3616 .free_inode = f2fs_free_inode, 3617 .drop_inode = f2fs_drop_inode, 3618 .write_inode = f2fs_write_inode, 3619 .dirty_inode = f2fs_dirty_inode, 3620 .show_options = f2fs_show_options, 3621 #ifdef CONFIG_QUOTA 3622 .quota_read = f2fs_quota_read, 3623 .quota_write = f2fs_quota_write, 3624 .get_dquots = f2fs_get_dquots, 3625 #endif 3626 .evict_inode = f2fs_evict_inode, 3627 .put_super = f2fs_put_super, 3628 .sync_fs = f2fs_sync_fs, 3629 .freeze_fs = f2fs_freeze, 3630 .unfreeze_fs = f2fs_unfreeze, 3631 .statfs = f2fs_statfs, 3632 .shutdown = f2fs_shutdown, 3633 }; 3634 3635 #ifdef CONFIG_FS_ENCRYPTION 3636 static int f2fs_get_context(struct inode *inode, void *ctx, size_t len) 3637 { 3638 return f2fs_getxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION, 3639 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT, 3640 ctx, len, NULL); 3641 } 3642 3643 static int f2fs_set_context(struct inode *inode, const void *ctx, size_t len, 3644 void *fs_data) 3645 { 3646 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 3647 3648 /* 3649 * Encrypting the root directory is not allowed because fsck 3650 * expects lost+found directory to exist and remain unencrypted 3651 * if LOST_FOUND feature is enabled. 3652 * 3653 */ 3654 if (f2fs_sb_has_lost_found(sbi) && 3655 inode->i_ino == F2FS_ROOT_INO(sbi)) 3656 return -EPERM; 3657 3658 return f2fs_setxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION, 3659 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT, 3660 ctx, len, fs_data, XATTR_CREATE); 3661 } 3662 3663 static const union fscrypt_policy *f2fs_get_dummy_policy(struct super_block *sb) 3664 { 3665 return F2FS_OPTION(F2FS_SB(sb)).dummy_enc_policy.policy; 3666 } 3667 3668 static bool f2fs_has_stable_inodes(struct super_block *sb) 3669 { 3670 return true; 3671 } 3672 3673 static struct block_device **f2fs_get_devices(struct super_block *sb, 3674 unsigned int *num_devs) 3675 { 3676 struct f2fs_sb_info *sbi = F2FS_SB(sb); 3677 struct block_device **devs; 3678 int i; 3679 3680 if (!f2fs_is_multi_device(sbi)) 3681 return NULL; 3682 3683 devs = kmalloc_array(sbi->s_ndevs, sizeof(*devs), GFP_KERNEL); 3684 if (!devs) 3685 return ERR_PTR(-ENOMEM); 3686 3687 for (i = 0; i < sbi->s_ndevs; i++) 3688 devs[i] = FDEV(i).bdev; 3689 *num_devs = sbi->s_ndevs; 3690 return devs; 3691 } 3692 3693 static const struct fscrypt_operations f2fs_cryptops = { 3694 .inode_info_offs = (int)offsetof(struct f2fs_inode_info, i_crypt_info) - 3695 (int)offsetof(struct f2fs_inode_info, vfs_inode), 3696 .needs_bounce_pages = 1, 3697 .has_32bit_inodes = 1, 3698 .supports_subblock_data_units = 1, 3699 .legacy_key_prefix = "f2fs:", 3700 .get_context = f2fs_get_context, 3701 .set_context = f2fs_set_context, 3702 .get_dummy_policy = f2fs_get_dummy_policy, 3703 .empty_dir = f2fs_empty_dir, 3704 .has_stable_inodes = f2fs_has_stable_inodes, 3705 .get_devices = f2fs_get_devices, 3706 }; 3707 #endif /* CONFIG_FS_ENCRYPTION */ 3708 3709 static struct inode *f2fs_nfs_get_inode(struct super_block *sb, 3710 u64 ino, u32 generation) 3711 { 3712 struct f2fs_sb_info *sbi = F2FS_SB(sb); 3713 struct inode *inode; 3714 3715 if (f2fs_check_nid_range(sbi, ino)) 3716 return ERR_PTR(-ESTALE); 3717 3718 /* 3719 * f2fs_iget isn't quite right if the inode is currently unallocated! 3720 * However f2fs_iget currently does appropriate checks to handle stale 3721 * inodes so everything is OK. 3722 */ 3723 inode = f2fs_iget(sb, ino); 3724 if (IS_ERR(inode)) 3725 return ERR_CAST(inode); 3726 if (unlikely(generation && inode->i_generation != generation)) { 3727 /* we didn't find the right inode.. */ 3728 iput(inode); 3729 return ERR_PTR(-ESTALE); 3730 } 3731 return inode; 3732 } 3733 3734 static struct dentry *f2fs_fh_to_dentry(struct super_block *sb, struct fid *fid, 3735 int fh_len, int fh_type) 3736 { 3737 return generic_fh_to_dentry(sb, fid, fh_len, fh_type, 3738 f2fs_nfs_get_inode); 3739 } 3740 3741 static struct dentry *f2fs_fh_to_parent(struct super_block *sb, struct fid *fid, 3742 int fh_len, int fh_type) 3743 { 3744 return generic_fh_to_parent(sb, fid, fh_len, fh_type, 3745 f2fs_nfs_get_inode); 3746 } 3747 3748 static const struct export_operations f2fs_export_ops = { 3749 .encode_fh = generic_encode_ino32_fh, 3750 .fh_to_dentry = f2fs_fh_to_dentry, 3751 .fh_to_parent = f2fs_fh_to_parent, 3752 .get_parent = f2fs_get_parent, 3753 }; 3754 3755 loff_t max_file_blocks(struct inode *inode) 3756 { 3757 loff_t result = 0; 3758 loff_t leaf_count; 3759 3760 /* 3761 * note: previously, result is equal to (DEF_ADDRS_PER_INODE - 3762 * DEFAULT_INLINE_XATTR_ADDRS), but now f2fs try to reserve more 3763 * space in inode.i_addr, it will be more safe to reassign 3764 * result as zero. 3765 */ 3766 3767 if (inode && f2fs_compressed_file(inode)) 3768 leaf_count = ADDRS_PER_BLOCK(inode); 3769 else 3770 leaf_count = DEF_ADDRS_PER_BLOCK; 3771 3772 /* two direct node blocks */ 3773 result += (leaf_count * 2); 3774 3775 /* two indirect node blocks */ 3776 leaf_count *= NIDS_PER_BLOCK; 3777 result += (leaf_count * 2); 3778 3779 /* one double indirect node block */ 3780 leaf_count *= NIDS_PER_BLOCK; 3781 result += leaf_count; 3782 3783 /* 3784 * For compatibility with FSCRYPT_POLICY_FLAG_IV_INO_LBLK_{64,32} with 3785 * a 4K crypto data unit, we must restrict the max filesize to what can 3786 * fit within U32_MAX + 1 data units. 3787 */ 3788 3789 result = umin(result, F2FS_BYTES_TO_BLK(((loff_t)U32_MAX + 1) * 4096)); 3790 3791 return result; 3792 } 3793 3794 static int __f2fs_commit_super(struct f2fs_sb_info *sbi, struct folio *folio, 3795 pgoff_t index, bool update) 3796 { 3797 struct bio *bio; 3798 /* it's rare case, we can do fua all the time */ 3799 blk_opf_t opf = REQ_OP_WRITE | REQ_SYNC | REQ_PREFLUSH | REQ_FUA; 3800 int ret; 3801 3802 folio_lock(folio); 3803 folio_wait_writeback(folio); 3804 if (update) 3805 memcpy(F2FS_SUPER_BLOCK(folio, index), F2FS_RAW_SUPER(sbi), 3806 sizeof(struct f2fs_super_block)); 3807 folio_mark_dirty(folio); 3808 folio_clear_dirty_for_io(folio); 3809 folio_start_writeback(folio); 3810 folio_unlock(folio); 3811 3812 bio = bio_alloc(sbi->sb->s_bdev, 1, opf, GFP_NOFS); 3813 3814 /* it doesn't need to set crypto context for superblock update */ 3815 bio->bi_iter.bi_sector = SECTOR_FROM_BLOCK(folio->index); 3816 3817 if (!bio_add_folio(bio, folio, folio_size(folio), 0)) 3818 f2fs_bug_on(sbi, 1); 3819 3820 ret = submit_bio_wait(bio); 3821 bio_put(bio); 3822 folio_end_writeback(folio); 3823 3824 return ret; 3825 } 3826 3827 static inline bool sanity_check_area_boundary(struct f2fs_sb_info *sbi, 3828 struct folio *folio, pgoff_t index) 3829 { 3830 struct f2fs_super_block *raw_super = F2FS_SUPER_BLOCK(folio, index); 3831 struct super_block *sb = sbi->sb; 3832 u32 segment0_blkaddr = le32_to_cpu(raw_super->segment0_blkaddr); 3833 u32 cp_blkaddr = le32_to_cpu(raw_super->cp_blkaddr); 3834 u32 sit_blkaddr = le32_to_cpu(raw_super->sit_blkaddr); 3835 u32 nat_blkaddr = le32_to_cpu(raw_super->nat_blkaddr); 3836 u32 ssa_blkaddr = le32_to_cpu(raw_super->ssa_blkaddr); 3837 u32 main_blkaddr = le32_to_cpu(raw_super->main_blkaddr); 3838 u32 segment_count_ckpt = le32_to_cpu(raw_super->segment_count_ckpt); 3839 u32 segment_count_sit = le32_to_cpu(raw_super->segment_count_sit); 3840 u32 segment_count_nat = le32_to_cpu(raw_super->segment_count_nat); 3841 u32 segment_count_ssa = le32_to_cpu(raw_super->segment_count_ssa); 3842 u32 segment_count_main = le32_to_cpu(raw_super->segment_count_main); 3843 u32 segment_count = le32_to_cpu(raw_super->segment_count); 3844 u32 log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg); 3845 u64 main_end_blkaddr = main_blkaddr + 3846 ((u64)segment_count_main << log_blocks_per_seg); 3847 u64 seg_end_blkaddr = segment0_blkaddr + 3848 ((u64)segment_count << log_blocks_per_seg); 3849 3850 if (segment0_blkaddr != cp_blkaddr) { 3851 f2fs_info(sbi, "Mismatch start address, segment0(%u) cp_blkaddr(%u)", 3852 segment0_blkaddr, cp_blkaddr); 3853 return true; 3854 } 3855 3856 if (cp_blkaddr + (segment_count_ckpt << log_blocks_per_seg) != 3857 sit_blkaddr) { 3858 f2fs_info(sbi, "Wrong CP boundary, start(%u) end(%u) blocks(%u)", 3859 cp_blkaddr, sit_blkaddr, 3860 segment_count_ckpt << log_blocks_per_seg); 3861 return true; 3862 } 3863 3864 if (sit_blkaddr + (segment_count_sit << log_blocks_per_seg) != 3865 nat_blkaddr) { 3866 f2fs_info(sbi, "Wrong SIT boundary, start(%u) end(%u) blocks(%u)", 3867 sit_blkaddr, nat_blkaddr, 3868 segment_count_sit << log_blocks_per_seg); 3869 return true; 3870 } 3871 3872 if (nat_blkaddr + (segment_count_nat << log_blocks_per_seg) != 3873 ssa_blkaddr) { 3874 f2fs_info(sbi, "Wrong NAT boundary, start(%u) end(%u) blocks(%u)", 3875 nat_blkaddr, ssa_blkaddr, 3876 segment_count_nat << log_blocks_per_seg); 3877 return true; 3878 } 3879 3880 if (ssa_blkaddr + (segment_count_ssa << log_blocks_per_seg) != 3881 main_blkaddr) { 3882 f2fs_info(sbi, "Wrong SSA boundary, start(%u) end(%u) blocks(%u)", 3883 ssa_blkaddr, main_blkaddr, 3884 segment_count_ssa << log_blocks_per_seg); 3885 return true; 3886 } 3887 3888 if (main_end_blkaddr > seg_end_blkaddr) { 3889 f2fs_info(sbi, "Wrong MAIN_AREA boundary, start(%u) end(%llu) block(%u)", 3890 main_blkaddr, seg_end_blkaddr, 3891 segment_count_main << log_blocks_per_seg); 3892 return true; 3893 } else if (main_end_blkaddr < seg_end_blkaddr) { 3894 int err = 0; 3895 char *res; 3896 3897 /* fix in-memory information all the time */ 3898 raw_super->segment_count = cpu_to_le32((main_end_blkaddr - 3899 segment0_blkaddr) >> log_blocks_per_seg); 3900 3901 if (f2fs_readonly(sb) || f2fs_hw_is_readonly(sbi)) { 3902 set_sbi_flag(sbi, SBI_NEED_SB_WRITE); 3903 res = "internally"; 3904 } else { 3905 err = __f2fs_commit_super(sbi, folio, index, false); 3906 res = err ? "failed" : "done"; 3907 } 3908 f2fs_info(sbi, "Fix alignment : %s, start(%u) end(%llu) block(%u)", 3909 res, main_blkaddr, seg_end_blkaddr, 3910 segment_count_main << log_blocks_per_seg); 3911 if (err) 3912 return true; 3913 } 3914 return false; 3915 } 3916 3917 static int sanity_check_raw_super(struct f2fs_sb_info *sbi, 3918 struct folio *folio, pgoff_t index) 3919 { 3920 block_t segment_count, segs_per_sec, secs_per_zone, segment_count_main; 3921 block_t total_sections, blocks_per_seg; 3922 struct f2fs_super_block *raw_super = F2FS_SUPER_BLOCK(folio, index); 3923 size_t crc_offset = 0; 3924 __u32 crc = 0; 3925 3926 if (le32_to_cpu(raw_super->magic) != F2FS_SUPER_MAGIC) { 3927 f2fs_info(sbi, "Magic Mismatch, valid(0x%x) - read(0x%x)", 3928 F2FS_SUPER_MAGIC, le32_to_cpu(raw_super->magic)); 3929 return -EINVAL; 3930 } 3931 3932 /* Check checksum_offset and crc in superblock */ 3933 if (__F2FS_HAS_FEATURE(raw_super, F2FS_FEATURE_SB_CHKSUM)) { 3934 crc_offset = le32_to_cpu(raw_super->checksum_offset); 3935 if (crc_offset != 3936 offsetof(struct f2fs_super_block, crc)) { 3937 f2fs_info(sbi, "Invalid SB checksum offset: %zu", 3938 crc_offset); 3939 return -EFSCORRUPTED; 3940 } 3941 crc = le32_to_cpu(raw_super->crc); 3942 if (crc != f2fs_crc32(raw_super, crc_offset)) { 3943 f2fs_info(sbi, "Invalid SB checksum value: %u", crc); 3944 return -EFSCORRUPTED; 3945 } 3946 } 3947 3948 /* only support block_size equals to PAGE_SIZE */ 3949 if (le32_to_cpu(raw_super->log_blocksize) != F2FS_BLKSIZE_BITS) { 3950 f2fs_info(sbi, "Invalid log_blocksize (%u), supports only %u", 3951 le32_to_cpu(raw_super->log_blocksize), 3952 F2FS_BLKSIZE_BITS); 3953 return -EFSCORRUPTED; 3954 } 3955 3956 /* check log blocks per segment */ 3957 if (le32_to_cpu(raw_super->log_blocks_per_seg) != 9) { 3958 f2fs_info(sbi, "Invalid log blocks per segment (%u)", 3959 le32_to_cpu(raw_super->log_blocks_per_seg)); 3960 return -EFSCORRUPTED; 3961 } 3962 3963 /* Currently, support 512/1024/2048/4096/16K bytes sector size */ 3964 if (le32_to_cpu(raw_super->log_sectorsize) > 3965 F2FS_MAX_LOG_SECTOR_SIZE || 3966 le32_to_cpu(raw_super->log_sectorsize) < 3967 F2FS_MIN_LOG_SECTOR_SIZE) { 3968 f2fs_info(sbi, "Invalid log sectorsize (%u)", 3969 le32_to_cpu(raw_super->log_sectorsize)); 3970 return -EFSCORRUPTED; 3971 } 3972 if (le32_to_cpu(raw_super->log_sectors_per_block) + 3973 le32_to_cpu(raw_super->log_sectorsize) != 3974 F2FS_MAX_LOG_SECTOR_SIZE) { 3975 f2fs_info(sbi, "Invalid log sectors per block(%u) log sectorsize(%u)", 3976 le32_to_cpu(raw_super->log_sectors_per_block), 3977 le32_to_cpu(raw_super->log_sectorsize)); 3978 return -EFSCORRUPTED; 3979 } 3980 3981 segment_count = le32_to_cpu(raw_super->segment_count); 3982 segment_count_main = le32_to_cpu(raw_super->segment_count_main); 3983 segs_per_sec = le32_to_cpu(raw_super->segs_per_sec); 3984 secs_per_zone = le32_to_cpu(raw_super->secs_per_zone); 3985 total_sections = le32_to_cpu(raw_super->section_count); 3986 3987 /* blocks_per_seg should be 512, given the above check */ 3988 blocks_per_seg = BIT(le32_to_cpu(raw_super->log_blocks_per_seg)); 3989 3990 if (segment_count > F2FS_MAX_SEGMENT || 3991 segment_count < F2FS_MIN_SEGMENTS) { 3992 f2fs_info(sbi, "Invalid segment count (%u)", segment_count); 3993 return -EFSCORRUPTED; 3994 } 3995 3996 if (total_sections > segment_count_main || total_sections < 1 || 3997 segs_per_sec > segment_count || !segs_per_sec) { 3998 f2fs_info(sbi, "Invalid segment/section count (%u, %u x %u)", 3999 segment_count, total_sections, segs_per_sec); 4000 return -EFSCORRUPTED; 4001 } 4002 4003 if (segment_count_main != total_sections * segs_per_sec) { 4004 f2fs_info(sbi, "Invalid segment/section count (%u != %u * %u)", 4005 segment_count_main, total_sections, segs_per_sec); 4006 return -EFSCORRUPTED; 4007 } 4008 4009 if ((segment_count / segs_per_sec) < total_sections) { 4010 f2fs_info(sbi, "Small segment_count (%u < %u * %u)", 4011 segment_count, segs_per_sec, total_sections); 4012 return -EFSCORRUPTED; 4013 } 4014 4015 if (segment_count > (le64_to_cpu(raw_super->block_count) >> 9)) { 4016 f2fs_info(sbi, "Wrong segment_count / block_count (%u > %llu)", 4017 segment_count, le64_to_cpu(raw_super->block_count)); 4018 return -EFSCORRUPTED; 4019 } 4020 4021 if (RDEV(0).path[0]) { 4022 block_t dev_seg_count = le32_to_cpu(RDEV(0).total_segments); 4023 int i = 1; 4024 4025 while (i < MAX_DEVICES && RDEV(i).path[0]) { 4026 dev_seg_count += le32_to_cpu(RDEV(i).total_segments); 4027 i++; 4028 } 4029 if (segment_count != dev_seg_count) { 4030 f2fs_info(sbi, "Segment count (%u) mismatch with total segments from devices (%u)", 4031 segment_count, dev_seg_count); 4032 return -EFSCORRUPTED; 4033 } 4034 } else { 4035 if (__F2FS_HAS_FEATURE(raw_super, F2FS_FEATURE_BLKZONED) && 4036 !bdev_is_zoned(sbi->sb->s_bdev)) { 4037 f2fs_info(sbi, "Zoned block device path is missing"); 4038 return -EFSCORRUPTED; 4039 } 4040 } 4041 4042 if (secs_per_zone > total_sections || !secs_per_zone) { 4043 f2fs_info(sbi, "Wrong secs_per_zone / total_sections (%u, %u)", 4044 secs_per_zone, total_sections); 4045 return -EFSCORRUPTED; 4046 } 4047 if (le32_to_cpu(raw_super->extension_count) > F2FS_MAX_EXTENSION || 4048 raw_super->hot_ext_count > F2FS_MAX_EXTENSION || 4049 (le32_to_cpu(raw_super->extension_count) + 4050 raw_super->hot_ext_count) > F2FS_MAX_EXTENSION) { 4051 f2fs_info(sbi, "Corrupted extension count (%u + %u > %u)", 4052 le32_to_cpu(raw_super->extension_count), 4053 raw_super->hot_ext_count, 4054 F2FS_MAX_EXTENSION); 4055 return -EFSCORRUPTED; 4056 } 4057 4058 if (le32_to_cpu(raw_super->cp_payload) >= 4059 (blocks_per_seg - F2FS_CP_PACKS - 4060 NR_CURSEG_PERSIST_TYPE)) { 4061 f2fs_info(sbi, "Insane cp_payload (%u >= %u)", 4062 le32_to_cpu(raw_super->cp_payload), 4063 blocks_per_seg - F2FS_CP_PACKS - 4064 NR_CURSEG_PERSIST_TYPE); 4065 return -EFSCORRUPTED; 4066 } 4067 4068 /* check reserved ino info */ 4069 if (le32_to_cpu(raw_super->node_ino) != 1 || 4070 le32_to_cpu(raw_super->meta_ino) != 2 || 4071 le32_to_cpu(raw_super->root_ino) != 3) { 4072 f2fs_info(sbi, "Invalid Fs Meta Ino: node(%u) meta(%u) root(%u)", 4073 le32_to_cpu(raw_super->node_ino), 4074 le32_to_cpu(raw_super->meta_ino), 4075 le32_to_cpu(raw_super->root_ino)); 4076 return -EFSCORRUPTED; 4077 } 4078 4079 /* check CP/SIT/NAT/SSA/MAIN_AREA area boundary */ 4080 if (sanity_check_area_boundary(sbi, folio, index)) 4081 return -EFSCORRUPTED; 4082 4083 /* 4084 * Check for legacy summary layout on 16KB+ block devices. 4085 * Modern f2fs-tools packs multiple 4KB summary areas into one block, 4086 * whereas legacy versions used one block per summary, leading 4087 * to a much larger SSA. 4088 */ 4089 if (SUMS_PER_BLOCK > 1 && 4090 !(__F2FS_HAS_FEATURE(raw_super, F2FS_FEATURE_PACKED_SSA))) { 4091 f2fs_info(sbi, "Error: Device formatted with a legacy version. " 4092 "Please reformat with a tool supporting the packed ssa " 4093 "feature for block sizes larger than 4kb."); 4094 return -EOPNOTSUPP; 4095 } 4096 4097 return 0; 4098 } 4099 4100 int f2fs_sanity_check_ckpt(struct f2fs_sb_info *sbi) 4101 { 4102 unsigned int total, fsmeta; 4103 struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi); 4104 struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi); 4105 unsigned int ovp_segments, reserved_segments; 4106 unsigned int main_segs, blocks_per_seg; 4107 unsigned int sit_segs, nat_segs; 4108 unsigned int sit_bitmap_size, nat_bitmap_size; 4109 unsigned int log_blocks_per_seg; 4110 unsigned int segment_count_main; 4111 unsigned int cp_pack_start_sum, cp_payload; 4112 block_t user_block_count, valid_user_blocks; 4113 block_t avail_node_count, valid_node_count; 4114 unsigned int nat_blocks, nat_bits_bytes, nat_bits_blocks; 4115 unsigned int sit_blk_cnt; 4116 int i, j; 4117 4118 total = le32_to_cpu(raw_super->segment_count); 4119 fsmeta = le32_to_cpu(raw_super->segment_count_ckpt); 4120 sit_segs = le32_to_cpu(raw_super->segment_count_sit); 4121 fsmeta += sit_segs; 4122 nat_segs = le32_to_cpu(raw_super->segment_count_nat); 4123 fsmeta += nat_segs; 4124 fsmeta += le32_to_cpu(ckpt->rsvd_segment_count); 4125 fsmeta += le32_to_cpu(raw_super->segment_count_ssa); 4126 4127 if (unlikely(fsmeta >= total)) 4128 return 1; 4129 4130 ovp_segments = le32_to_cpu(ckpt->overprov_segment_count); 4131 reserved_segments = le32_to_cpu(ckpt->rsvd_segment_count); 4132 4133 if (!f2fs_sb_has_readonly(sbi) && 4134 unlikely(fsmeta < F2FS_MIN_META_SEGMENTS || 4135 ovp_segments == 0 || reserved_segments == 0)) { 4136 f2fs_err(sbi, "Wrong layout: check mkfs.f2fs version"); 4137 return 1; 4138 } 4139 user_block_count = le64_to_cpu(ckpt->user_block_count); 4140 segment_count_main = le32_to_cpu(raw_super->segment_count_main) + 4141 (f2fs_sb_has_readonly(sbi) ? 1 : 0); 4142 log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg); 4143 if (!user_block_count || user_block_count >= 4144 segment_count_main << log_blocks_per_seg) { 4145 f2fs_err(sbi, "Wrong user_block_count: %u", 4146 user_block_count); 4147 return 1; 4148 } 4149 4150 valid_user_blocks = le64_to_cpu(ckpt->valid_block_count); 4151 if (valid_user_blocks > user_block_count) { 4152 f2fs_err(sbi, "Wrong valid_user_blocks: %u, user_block_count: %u", 4153 valid_user_blocks, user_block_count); 4154 return 1; 4155 } 4156 4157 valid_node_count = le32_to_cpu(ckpt->valid_node_count); 4158 avail_node_count = sbi->total_node_count - F2FS_RESERVED_NODE_NUM; 4159 if (valid_node_count > avail_node_count) { 4160 f2fs_err(sbi, "Wrong valid_node_count: %u, avail_node_count: %u", 4161 valid_node_count, avail_node_count); 4162 return 1; 4163 } 4164 4165 main_segs = le32_to_cpu(raw_super->segment_count_main); 4166 blocks_per_seg = BLKS_PER_SEG(sbi); 4167 4168 for (i = 0; i < NR_CURSEG_NODE_TYPE; i++) { 4169 if (le32_to_cpu(ckpt->cur_node_segno[i]) >= main_segs || 4170 le16_to_cpu(ckpt->cur_node_blkoff[i]) >= blocks_per_seg) 4171 return 1; 4172 4173 if (f2fs_sb_has_readonly(sbi)) 4174 goto check_data; 4175 4176 for (j = i + 1; j < NR_CURSEG_NODE_TYPE; j++) { 4177 if (le32_to_cpu(ckpt->cur_node_segno[i]) == 4178 le32_to_cpu(ckpt->cur_node_segno[j])) { 4179 f2fs_err(sbi, "Node segment (%u, %u) has the same segno: %u", 4180 i, j, 4181 le32_to_cpu(ckpt->cur_node_segno[i])); 4182 return 1; 4183 } 4184 } 4185 } 4186 check_data: 4187 for (i = 0; i < NR_CURSEG_DATA_TYPE; i++) { 4188 if (le32_to_cpu(ckpt->cur_data_segno[i]) >= main_segs || 4189 le16_to_cpu(ckpt->cur_data_blkoff[i]) >= blocks_per_seg) 4190 return 1; 4191 4192 if (f2fs_sb_has_readonly(sbi)) 4193 goto skip_cross; 4194 4195 for (j = i + 1; j < NR_CURSEG_DATA_TYPE; j++) { 4196 if (le32_to_cpu(ckpt->cur_data_segno[i]) == 4197 le32_to_cpu(ckpt->cur_data_segno[j])) { 4198 f2fs_err(sbi, "Data segment (%u, %u) has the same segno: %u", 4199 i, j, 4200 le32_to_cpu(ckpt->cur_data_segno[i])); 4201 return 1; 4202 } 4203 } 4204 } 4205 for (i = 0; i < NR_CURSEG_NODE_TYPE; i++) { 4206 for (j = 0; j < NR_CURSEG_DATA_TYPE; j++) { 4207 if (le32_to_cpu(ckpt->cur_node_segno[i]) == 4208 le32_to_cpu(ckpt->cur_data_segno[j])) { 4209 f2fs_err(sbi, "Node segment (%u) and Data segment (%u) has the same segno: %u", 4210 i, j, 4211 le32_to_cpu(ckpt->cur_node_segno[i])); 4212 return 1; 4213 } 4214 } 4215 } 4216 skip_cross: 4217 sit_bitmap_size = le32_to_cpu(ckpt->sit_ver_bitmap_bytesize); 4218 nat_bitmap_size = le32_to_cpu(ckpt->nat_ver_bitmap_bytesize); 4219 4220 if (sit_bitmap_size != ((sit_segs / 2) << log_blocks_per_seg) / 8 || 4221 nat_bitmap_size != ((nat_segs / 2) << log_blocks_per_seg) / 8) { 4222 f2fs_err(sbi, "Wrong bitmap size: sit: %u, nat:%u", 4223 sit_bitmap_size, nat_bitmap_size); 4224 return 1; 4225 } 4226 4227 sit_blk_cnt = DIV_ROUND_UP(main_segs, SIT_ENTRY_PER_BLOCK); 4228 if (sit_bitmap_size * 8 < sit_blk_cnt) { 4229 f2fs_err(sbi, "Wrong bitmap size: sit: %u, sit_blk_cnt:%u", 4230 sit_bitmap_size, sit_blk_cnt); 4231 return 1; 4232 } 4233 4234 cp_pack_start_sum = __start_sum_addr(sbi); 4235 cp_payload = __cp_payload(sbi); 4236 if (cp_pack_start_sum < cp_payload + 1 || 4237 cp_pack_start_sum > blocks_per_seg - 1 - 4238 NR_CURSEG_PERSIST_TYPE) { 4239 f2fs_err(sbi, "Wrong cp_pack_start_sum: %u", 4240 cp_pack_start_sum); 4241 return 1; 4242 } 4243 4244 if (__is_set_ckpt_flags(ckpt, CP_LARGE_NAT_BITMAP_FLAG) && 4245 le32_to_cpu(ckpt->checksum_offset) != CP_MIN_CHKSUM_OFFSET) { 4246 f2fs_warn(sbi, "using deprecated layout of large_nat_bitmap, " 4247 "please run fsck v1.13.0 or higher to repair, chksum_offset: %u, " 4248 "fixed with patch: \"f2fs-tools: relocate chksum_offset for large_nat_bitmap feature\"", 4249 le32_to_cpu(ckpt->checksum_offset)); 4250 return 1; 4251 } 4252 4253 nat_blocks = nat_segs << log_blocks_per_seg; 4254 nat_bits_bytes = nat_blocks / BITS_PER_BYTE; 4255 nat_bits_blocks = F2FS_BLK_ALIGN((nat_bits_bytes << 1) + 8); 4256 if (__is_set_ckpt_flags(ckpt, CP_NAT_BITS_FLAG) && 4257 (cp_payload + F2FS_CP_PACKS + 4258 NR_CURSEG_PERSIST_TYPE + nat_bits_blocks >= blocks_per_seg)) { 4259 f2fs_warn(sbi, "Insane cp_payload: %u, nat_bits_blocks: %u)", 4260 cp_payload, nat_bits_blocks); 4261 return 1; 4262 } 4263 4264 if (unlikely(f2fs_cp_error(sbi))) { 4265 f2fs_err(sbi, "A bug case: need to run fsck"); 4266 return 1; 4267 } 4268 return 0; 4269 } 4270 4271 static void init_sb_info(struct f2fs_sb_info *sbi) 4272 { 4273 struct f2fs_super_block *raw_super = sbi->raw_super; 4274 int i; 4275 4276 sbi->log_sectors_per_block = 4277 le32_to_cpu(raw_super->log_sectors_per_block); 4278 sbi->log_blocksize = le32_to_cpu(raw_super->log_blocksize); 4279 sbi->blocksize = BIT(sbi->log_blocksize); 4280 sbi->log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg); 4281 sbi->blocks_per_seg = BIT(sbi->log_blocks_per_seg); 4282 sbi->segs_per_sec = le32_to_cpu(raw_super->segs_per_sec); 4283 sbi->secs_per_zone = le32_to_cpu(raw_super->secs_per_zone); 4284 sbi->total_sections = le32_to_cpu(raw_super->section_count); 4285 sbi->total_node_count = SEGS_TO_BLKS(sbi, 4286 ((le32_to_cpu(raw_super->segment_count_nat) / 2) * 4287 NAT_ENTRY_PER_BLOCK)); 4288 sbi->allocate_section_hint = le32_to_cpu(raw_super->section_count); 4289 sbi->allocate_section_policy = ALLOCATE_FORWARD_NOHINT; 4290 F2FS_ROOT_INO(sbi) = le32_to_cpu(raw_super->root_ino); 4291 F2FS_NODE_INO(sbi) = le32_to_cpu(raw_super->node_ino); 4292 F2FS_META_INO(sbi) = le32_to_cpu(raw_super->meta_ino); 4293 sbi->cur_victim_sec = NULL_SECNO; 4294 sbi->gc_mode = GC_NORMAL; 4295 sbi->next_victim_seg[BG_GC] = NULL_SEGNO; 4296 sbi->next_victim_seg[FG_GC] = NULL_SEGNO; 4297 sbi->max_victim_search = DEF_MAX_VICTIM_SEARCH; 4298 sbi->migration_granularity = SEGS_PER_SEC(sbi); 4299 sbi->migration_window_granularity = f2fs_sb_has_blkzoned(sbi) ? 4300 DEF_MIGRATION_WINDOW_GRANULARITY_ZONED : SEGS_PER_SEC(sbi); 4301 sbi->seq_file_ra_mul = MIN_RA_MUL; 4302 sbi->max_fragment_chunk = DEF_FRAGMENT_SIZE; 4303 sbi->max_fragment_hole = DEF_FRAGMENT_SIZE; 4304 spin_lock_init(&sbi->gc_remaining_trials_lock); 4305 atomic64_set(&sbi->current_atomic_write, 0); 4306 4307 sbi->dir_level = DEF_DIR_LEVEL; 4308 sbi->interval_time[CP_TIME] = DEF_CP_INTERVAL; 4309 sbi->interval_time[REQ_TIME] = DEF_IDLE_INTERVAL; 4310 sbi->interval_time[DISCARD_TIME] = DEF_IDLE_INTERVAL; 4311 sbi->interval_time[GC_TIME] = DEF_IDLE_INTERVAL; 4312 sbi->interval_time[DISABLE_TIME] = DEF_DISABLE_INTERVAL; 4313 sbi->interval_time[ENABLE_TIME] = DEF_ENABLE_INTERVAL; 4314 sbi->interval_time[UMOUNT_DISCARD_TIMEOUT] = 4315 DEF_UMOUNT_DISCARD_TIMEOUT; 4316 clear_sbi_flag(sbi, SBI_NEED_FSCK); 4317 4318 for (i = 0; i < NR_COUNT_TYPE; i++) 4319 atomic_set(&sbi->nr_pages[i], 0); 4320 4321 for (i = 0; i < META; i++) 4322 atomic_set(&sbi->wb_sync_req[i], 0); 4323 4324 INIT_LIST_HEAD(&sbi->s_list); 4325 mutex_init(&sbi->umount_mutex); 4326 init_f2fs_rwsem(&sbi->io_order_lock); 4327 spin_lock_init(&sbi->cp_lock); 4328 4329 sbi->dirty_device = 0; 4330 spin_lock_init(&sbi->dev_lock); 4331 4332 init_f2fs_rwsem(&sbi->sb_lock); 4333 init_f2fs_rwsem(&sbi->pin_sem); 4334 } 4335 4336 static int init_percpu_info(struct f2fs_sb_info *sbi) 4337 { 4338 int err; 4339 4340 err = percpu_counter_init(&sbi->alloc_valid_block_count, 0, GFP_KERNEL); 4341 if (err) 4342 return err; 4343 4344 err = percpu_counter_init(&sbi->rf_node_block_count, 0, GFP_KERNEL); 4345 if (err) 4346 goto err_valid_block; 4347 4348 err = percpu_counter_init(&sbi->total_valid_inode_count, 0, 4349 GFP_KERNEL); 4350 if (err) 4351 goto err_node_block; 4352 return 0; 4353 4354 err_node_block: 4355 percpu_counter_destroy(&sbi->rf_node_block_count); 4356 err_valid_block: 4357 percpu_counter_destroy(&sbi->alloc_valid_block_count); 4358 return err; 4359 } 4360 4361 #ifdef CONFIG_BLK_DEV_ZONED 4362 4363 struct f2fs_report_zones_args { 4364 struct f2fs_sb_info *sbi; 4365 struct f2fs_dev_info *dev; 4366 }; 4367 4368 static int f2fs_report_zone_cb(struct blk_zone *zone, unsigned int idx, 4369 void *data) 4370 { 4371 struct f2fs_report_zones_args *rz_args = data; 4372 block_t unusable_blocks = (zone->len - zone->capacity) >> 4373 F2FS_LOG_SECTORS_PER_BLOCK; 4374 4375 if (zone->type == BLK_ZONE_TYPE_CONVENTIONAL) 4376 return 0; 4377 4378 set_bit(idx, rz_args->dev->blkz_seq); 4379 if (!rz_args->sbi->unusable_blocks_per_sec) { 4380 rz_args->sbi->unusable_blocks_per_sec = unusable_blocks; 4381 return 0; 4382 } 4383 if (rz_args->sbi->unusable_blocks_per_sec != unusable_blocks) { 4384 f2fs_err(rz_args->sbi, "F2FS supports single zone capacity\n"); 4385 return -EINVAL; 4386 } 4387 return 0; 4388 } 4389 4390 static int init_blkz_info(struct f2fs_sb_info *sbi, int devi) 4391 { 4392 struct block_device *bdev = FDEV(devi).bdev; 4393 sector_t nr_sectors = bdev_nr_sectors(bdev); 4394 struct f2fs_report_zones_args rep_zone_arg; 4395 u64 zone_sectors; 4396 unsigned int max_open_zones; 4397 int ret; 4398 4399 if (!f2fs_sb_has_blkzoned(sbi)) 4400 return 0; 4401 4402 if (bdev_is_zoned(FDEV(devi).bdev)) { 4403 max_open_zones = bdev_max_open_zones(bdev); 4404 if (max_open_zones && (max_open_zones < sbi->max_open_zones)) 4405 sbi->max_open_zones = max_open_zones; 4406 if (sbi->max_open_zones < F2FS_OPTION(sbi).active_logs) { 4407 f2fs_err(sbi, 4408 "zoned: max open zones %u is too small, need at least %u open zones", 4409 sbi->max_open_zones, F2FS_OPTION(sbi).active_logs); 4410 return -EINVAL; 4411 } 4412 } 4413 4414 zone_sectors = bdev_zone_sectors(bdev); 4415 if (sbi->blocks_per_blkz && sbi->blocks_per_blkz != 4416 SECTOR_TO_BLOCK(zone_sectors)) 4417 return -EINVAL; 4418 sbi->blocks_per_blkz = SECTOR_TO_BLOCK(zone_sectors); 4419 FDEV(devi).nr_blkz = div_u64(SECTOR_TO_BLOCK(nr_sectors), 4420 sbi->blocks_per_blkz); 4421 if (nr_sectors & (zone_sectors - 1)) 4422 FDEV(devi).nr_blkz++; 4423 4424 FDEV(devi).blkz_seq = f2fs_kvzalloc(sbi, 4425 BITS_TO_LONGS(FDEV(devi).nr_blkz) 4426 * sizeof(unsigned long), 4427 GFP_KERNEL); 4428 if (!FDEV(devi).blkz_seq) 4429 return -ENOMEM; 4430 4431 rep_zone_arg.sbi = sbi; 4432 rep_zone_arg.dev = &FDEV(devi); 4433 4434 ret = blkdev_report_zones(bdev, 0, BLK_ALL_ZONES, f2fs_report_zone_cb, 4435 &rep_zone_arg); 4436 if (ret < 0) 4437 return ret; 4438 return 0; 4439 } 4440 #endif 4441 4442 /* 4443 * Read f2fs raw super block. 4444 * Because we have two copies of super block, so read both of them 4445 * to get the first valid one. If any one of them is broken, we pass 4446 * them recovery flag back to the caller. 4447 */ 4448 static int read_raw_super_block(struct f2fs_sb_info *sbi, 4449 struct f2fs_super_block **raw_super, 4450 int *valid_super_block, int *recovery) 4451 { 4452 struct super_block *sb = sbi->sb; 4453 int block; 4454 struct folio *folio; 4455 struct f2fs_super_block *super; 4456 int err = 0; 4457 4458 super = kzalloc(sizeof(struct f2fs_super_block), GFP_KERNEL); 4459 if (!super) 4460 return -ENOMEM; 4461 4462 for (block = 0; block < 2; block++) { 4463 folio = read_mapping_folio(sb->s_bdev->bd_mapping, block, NULL); 4464 if (IS_ERR(folio)) { 4465 f2fs_err(sbi, "Unable to read %dth superblock", 4466 block + 1); 4467 err = PTR_ERR(folio); 4468 *recovery = 1; 4469 continue; 4470 } 4471 4472 /* sanity checking of raw super */ 4473 err = sanity_check_raw_super(sbi, folio, block); 4474 if (err) { 4475 f2fs_err(sbi, "Can't find valid F2FS filesystem in %dth superblock", 4476 block + 1); 4477 folio_put(folio); 4478 *recovery = 1; 4479 continue; 4480 } 4481 4482 if (!*raw_super) { 4483 memcpy(super, F2FS_SUPER_BLOCK(folio, block), 4484 sizeof(*super)); 4485 *valid_super_block = block; 4486 *raw_super = super; 4487 } 4488 folio_put(folio); 4489 } 4490 4491 /* No valid superblock */ 4492 if (!*raw_super) 4493 kfree(super); 4494 else 4495 err = 0; 4496 4497 return err; 4498 } 4499 4500 int f2fs_commit_super(struct f2fs_sb_info *sbi, bool recover) 4501 { 4502 struct folio *folio; 4503 pgoff_t index; 4504 __u32 crc = 0; 4505 int err; 4506 4507 if ((recover && f2fs_readonly(sbi->sb)) || 4508 f2fs_hw_is_readonly(sbi)) { 4509 set_sbi_flag(sbi, SBI_NEED_SB_WRITE); 4510 return -EROFS; 4511 } 4512 4513 /* we should update superblock crc here */ 4514 if (!recover && f2fs_sb_has_sb_chksum(sbi)) { 4515 crc = f2fs_crc32(F2FS_RAW_SUPER(sbi), 4516 offsetof(struct f2fs_super_block, crc)); 4517 F2FS_RAW_SUPER(sbi)->crc = cpu_to_le32(crc); 4518 } 4519 4520 /* write back-up superblock first */ 4521 index = sbi->valid_super_block ? 0 : 1; 4522 folio = read_mapping_folio(sbi->sb->s_bdev->bd_mapping, index, NULL); 4523 if (IS_ERR(folio)) 4524 return PTR_ERR(folio); 4525 err = __f2fs_commit_super(sbi, folio, index, true); 4526 folio_put(folio); 4527 4528 /* if we are in recovery path, skip writing valid superblock */ 4529 if (recover || err) 4530 return err; 4531 4532 /* write current valid superblock */ 4533 index = sbi->valid_super_block; 4534 folio = read_mapping_folio(sbi->sb->s_bdev->bd_mapping, index, NULL); 4535 if (IS_ERR(folio)) 4536 return PTR_ERR(folio); 4537 err = __f2fs_commit_super(sbi, folio, index, true); 4538 folio_put(folio); 4539 return err; 4540 } 4541 4542 static void save_stop_reason(struct f2fs_sb_info *sbi, unsigned char reason) 4543 { 4544 unsigned long flags; 4545 4546 spin_lock_irqsave(&sbi->error_lock, flags); 4547 if (sbi->stop_reason[reason] < GENMASK(BITS_PER_BYTE - 1, 0)) 4548 sbi->stop_reason[reason]++; 4549 spin_unlock_irqrestore(&sbi->error_lock, flags); 4550 } 4551 4552 static void f2fs_record_stop_reason(struct f2fs_sb_info *sbi) 4553 { 4554 struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi); 4555 unsigned long flags; 4556 int err; 4557 4558 f2fs_down_write(&sbi->sb_lock); 4559 4560 spin_lock_irqsave(&sbi->error_lock, flags); 4561 if (sbi->error_dirty) { 4562 memcpy(F2FS_RAW_SUPER(sbi)->s_errors, sbi->errors, 4563 MAX_F2FS_ERRORS); 4564 sbi->error_dirty = false; 4565 } 4566 memcpy(raw_super->s_stop_reason, sbi->stop_reason, MAX_STOP_REASON); 4567 spin_unlock_irqrestore(&sbi->error_lock, flags); 4568 4569 err = f2fs_commit_super(sbi, false); 4570 4571 f2fs_up_write(&sbi->sb_lock); 4572 if (err) 4573 f2fs_err_ratelimited(sbi, 4574 "f2fs_commit_super fails to record stop_reason, err:%d", 4575 err); 4576 } 4577 4578 void f2fs_save_errors(struct f2fs_sb_info *sbi, unsigned char flag) 4579 { 4580 unsigned long flags; 4581 4582 spin_lock_irqsave(&sbi->error_lock, flags); 4583 if (!test_bit(flag, (unsigned long *)sbi->errors)) { 4584 set_bit(flag, (unsigned long *)sbi->errors); 4585 sbi->error_dirty = true; 4586 } 4587 spin_unlock_irqrestore(&sbi->error_lock, flags); 4588 } 4589 4590 void f2fs_handle_error(struct f2fs_sb_info *sbi, unsigned char error) 4591 { 4592 f2fs_save_errors(sbi, error); 4593 4594 if (!sbi->error_dirty) 4595 return; 4596 if (!test_bit(error, (unsigned long *)sbi->errors)) 4597 return; 4598 schedule_work(&sbi->s_error_work); 4599 } 4600 4601 static bool system_going_down(void) 4602 { 4603 return system_state == SYSTEM_HALT || system_state == SYSTEM_POWER_OFF 4604 || system_state == SYSTEM_RESTART; 4605 } 4606 4607 void f2fs_handle_critical_error(struct f2fs_sb_info *sbi, unsigned char reason) 4608 { 4609 struct super_block *sb = sbi->sb; 4610 bool shutdown = reason == STOP_CP_REASON_SHUTDOWN; 4611 bool continue_fs = !shutdown && 4612 F2FS_OPTION(sbi).errors == MOUNT_ERRORS_CONTINUE; 4613 4614 set_ckpt_flags(sbi, CP_ERROR_FLAG); 4615 4616 if (!f2fs_hw_is_readonly(sbi)) { 4617 save_stop_reason(sbi, reason); 4618 4619 /* 4620 * always create an asynchronous task to record stop_reason 4621 * in order to avoid potential deadlock when running into 4622 * f2fs_record_stop_reason() synchronously. 4623 */ 4624 schedule_work(&sbi->s_error_work); 4625 } 4626 4627 /* 4628 * We force ERRORS_RO behavior when system is rebooting. Otherwise we 4629 * could panic during 'reboot -f' as the underlying device got already 4630 * disabled. 4631 */ 4632 if (F2FS_OPTION(sbi).errors == MOUNT_ERRORS_PANIC && 4633 !shutdown && !system_going_down() && 4634 !is_sbi_flag_set(sbi, SBI_IS_SHUTDOWN)) 4635 panic("F2FS-fs (device %s): panic forced after error\n", 4636 sb->s_id); 4637 4638 if (shutdown) 4639 set_sbi_flag(sbi, SBI_IS_SHUTDOWN); 4640 else 4641 dump_stack(); 4642 4643 /* 4644 * Continue filesystem operators if errors=continue. Should not set 4645 * RO by shutdown, since RO bypasses thaw_super which can hang the 4646 * system. 4647 */ 4648 if (continue_fs || f2fs_readonly(sb) || shutdown) { 4649 f2fs_warn(sbi, "Stopped filesystem due to reason: %d", reason); 4650 return; 4651 } 4652 4653 f2fs_warn(sbi, "Remounting filesystem read-only"); 4654 4655 /* 4656 * We have already set CP_ERROR_FLAG flag to stop all updates 4657 * to filesystem, so it doesn't need to set SB_RDONLY flag here 4658 * because the flag should be set covered w/ sb->s_umount semaphore 4659 * via remount procedure, otherwise, it will confuse code like 4660 * freeze_super() which will lead to deadlocks and other problems. 4661 */ 4662 } 4663 4664 static void f2fs_record_error_work(struct work_struct *work) 4665 { 4666 struct f2fs_sb_info *sbi = container_of(work, 4667 struct f2fs_sb_info, s_error_work); 4668 4669 f2fs_record_stop_reason(sbi); 4670 } 4671 4672 static inline unsigned int get_first_seq_zone_segno(struct f2fs_sb_info *sbi) 4673 { 4674 #ifdef CONFIG_BLK_DEV_ZONED 4675 unsigned int zoneno, total_zones; 4676 int devi; 4677 4678 if (!f2fs_sb_has_blkzoned(sbi)) 4679 return NULL_SEGNO; 4680 4681 for (devi = 0; devi < sbi->s_ndevs; devi++) { 4682 if (!bdev_is_zoned(FDEV(devi).bdev)) 4683 continue; 4684 4685 total_zones = GET_ZONE_FROM_SEG(sbi, FDEV(devi).total_segments); 4686 4687 for (zoneno = 0; zoneno < total_zones; zoneno++) { 4688 unsigned int segs, blks; 4689 4690 if (!f2fs_zone_is_seq(sbi, devi, zoneno)) 4691 continue; 4692 4693 segs = GET_SEG_FROM_SEC(sbi, 4694 zoneno * sbi->secs_per_zone); 4695 blks = SEGS_TO_BLKS(sbi, segs); 4696 return GET_SEGNO(sbi, FDEV(devi).start_blk + blks); 4697 } 4698 } 4699 #endif 4700 return NULL_SEGNO; 4701 } 4702 4703 static int f2fs_scan_devices(struct f2fs_sb_info *sbi) 4704 { 4705 struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi); 4706 unsigned int max_devices = MAX_DEVICES; 4707 unsigned int logical_blksize; 4708 blk_mode_t mode = sb_open_mode(sbi->sb->s_flags); 4709 int i; 4710 4711 /* Initialize single device information */ 4712 if (!RDEV(0).path[0]) { 4713 if (!bdev_is_zoned(sbi->sb->s_bdev)) 4714 return 0; 4715 max_devices = 1; 4716 } 4717 4718 /* 4719 * Initialize multiple devices information, or single 4720 * zoned block device information. 4721 */ 4722 sbi->devs = f2fs_kzalloc(sbi, 4723 array_size(max_devices, 4724 sizeof(struct f2fs_dev_info)), 4725 GFP_KERNEL); 4726 if (!sbi->devs) 4727 return -ENOMEM; 4728 4729 logical_blksize = bdev_logical_block_size(sbi->sb->s_bdev); 4730 sbi->aligned_blksize = true; 4731 sbi->bggc_io_aware = AWARE_ALL_IO; 4732 #ifdef CONFIG_BLK_DEV_ZONED 4733 sbi->max_open_zones = UINT_MAX; 4734 sbi->blkzone_alloc_policy = BLKZONE_ALLOC_PRIOR_SEQ; 4735 sbi->bggc_io_aware = AWARE_READ_IO; 4736 #endif 4737 4738 for (i = 0; i < max_devices; i++) { 4739 if (max_devices == 1) { 4740 FDEV(i).total_segments = 4741 le32_to_cpu(raw_super->segment_count_main); 4742 FDEV(i).start_blk = 0; 4743 FDEV(i).end_blk = FDEV(i).total_segments * 4744 BLKS_PER_SEG(sbi); 4745 } 4746 4747 if (i == 0) 4748 FDEV(0).bdev_file = sbi->sb->s_bdev_file; 4749 else if (!RDEV(i).path[0]) 4750 break; 4751 4752 if (max_devices > 1) { 4753 /* Multi-device mount */ 4754 memcpy(FDEV(i).path, RDEV(i).path, MAX_PATH_LEN); 4755 FDEV(i).total_segments = 4756 le32_to_cpu(RDEV(i).total_segments); 4757 if (i == 0) { 4758 FDEV(i).start_blk = 0; 4759 FDEV(i).end_blk = FDEV(i).start_blk + 4760 SEGS_TO_BLKS(sbi, 4761 FDEV(i).total_segments) - 1 + 4762 le32_to_cpu(raw_super->segment0_blkaddr); 4763 sbi->allocate_section_hint = FDEV(i).total_segments / 4764 SEGS_PER_SEC(sbi); 4765 } else { 4766 FDEV(i).start_blk = FDEV(i - 1).end_blk + 1; 4767 FDEV(i).end_blk = FDEV(i).start_blk + 4768 SEGS_TO_BLKS(sbi, 4769 FDEV(i).total_segments) - 1; 4770 FDEV(i).bdev_file = bdev_file_open_by_path( 4771 FDEV(i).path, mode, sbi->sb, NULL); 4772 } 4773 } 4774 if (IS_ERR(FDEV(i).bdev_file)) 4775 return PTR_ERR(FDEV(i).bdev_file); 4776 4777 FDEV(i).bdev = file_bdev(FDEV(i).bdev_file); 4778 /* to release errored devices */ 4779 sbi->s_ndevs = i + 1; 4780 4781 if (logical_blksize != bdev_logical_block_size(FDEV(i).bdev)) 4782 sbi->aligned_blksize = false; 4783 4784 #ifdef CONFIG_BLK_DEV_ZONED 4785 if (bdev_is_zoned(FDEV(i).bdev)) { 4786 if (!f2fs_sb_has_blkzoned(sbi)) { 4787 f2fs_err(sbi, "Zoned block device feature not enabled"); 4788 return -EINVAL; 4789 } 4790 if (init_blkz_info(sbi, i)) { 4791 f2fs_err(sbi, "Failed to initialize F2FS blkzone information"); 4792 return -EINVAL; 4793 } 4794 if (max_devices == 1) 4795 break; 4796 f2fs_info(sbi, "Mount Device [%2d]: %20s, %8u, %8x - %8x (zone: Host-managed)", 4797 i, FDEV(i).path, 4798 FDEV(i).total_segments, 4799 FDEV(i).start_blk, FDEV(i).end_blk); 4800 continue; 4801 } 4802 #endif 4803 f2fs_info(sbi, "Mount Device [%2d]: %20s, %8u, %8x - %8x", 4804 i, FDEV(i).path, 4805 FDEV(i).total_segments, 4806 FDEV(i).start_blk, FDEV(i).end_blk); 4807 } 4808 return 0; 4809 } 4810 4811 static int f2fs_setup_casefold(struct f2fs_sb_info *sbi) 4812 { 4813 #if IS_ENABLED(CONFIG_UNICODE) 4814 if (f2fs_sb_has_casefold(sbi) && !sbi->sb->s_encoding) { 4815 const struct f2fs_sb_encodings *encoding_info; 4816 struct unicode_map *encoding; 4817 __u16 encoding_flags; 4818 4819 encoding_info = f2fs_sb_read_encoding(sbi->raw_super); 4820 if (!encoding_info) { 4821 f2fs_err(sbi, 4822 "Encoding requested by superblock is unknown"); 4823 return -EINVAL; 4824 } 4825 4826 encoding_flags = le16_to_cpu(sbi->raw_super->s_encoding_flags); 4827 encoding = utf8_load(encoding_info->version); 4828 if (IS_ERR(encoding)) { 4829 f2fs_err(sbi, 4830 "can't mount with superblock charset: %s-%u.%u.%u " 4831 "not supported by the kernel. flags: 0x%x.", 4832 encoding_info->name, 4833 unicode_major(encoding_info->version), 4834 unicode_minor(encoding_info->version), 4835 unicode_rev(encoding_info->version), 4836 encoding_flags); 4837 return PTR_ERR(encoding); 4838 } 4839 f2fs_info(sbi, "Using encoding defined by superblock: " 4840 "%s-%u.%u.%u with flags 0x%hx", encoding_info->name, 4841 unicode_major(encoding_info->version), 4842 unicode_minor(encoding_info->version), 4843 unicode_rev(encoding_info->version), 4844 encoding_flags); 4845 4846 sbi->sb->s_encoding = encoding; 4847 sbi->sb->s_encoding_flags = encoding_flags; 4848 } 4849 #else 4850 if (f2fs_sb_has_casefold(sbi)) { 4851 f2fs_err(sbi, "Filesystem with casefold feature cannot be mounted without CONFIG_UNICODE"); 4852 return -EINVAL; 4853 } 4854 #endif 4855 return 0; 4856 } 4857 4858 static void f2fs_tuning_parameters(struct f2fs_sb_info *sbi) 4859 { 4860 /* adjust parameters according to the volume size */ 4861 if (MAIN_SEGS(sbi) <= SMALL_VOLUME_SEGMENTS) { 4862 if (f2fs_block_unit_discard(sbi)) 4863 SM_I(sbi)->dcc_info->discard_granularity = 4864 MIN_DISCARD_GRANULARITY; 4865 if (!f2fs_lfs_mode(sbi)) 4866 SM_I(sbi)->ipu_policy = BIT(F2FS_IPU_FORCE) | 4867 BIT(F2FS_IPU_HONOR_OPU_WRITE); 4868 } 4869 4870 sbi->readdir_ra = true; 4871 } 4872 4873 static int f2fs_fill_super(struct super_block *sb, struct fs_context *fc) 4874 { 4875 struct f2fs_fs_context *ctx = fc->fs_private; 4876 struct f2fs_sb_info *sbi; 4877 struct f2fs_super_block *raw_super; 4878 struct inode *root; 4879 int err; 4880 bool skip_recovery = false, need_fsck = false; 4881 int recovery, i, valid_super_block; 4882 struct curseg_info *seg_i; 4883 int retry_cnt = 1; 4884 #ifdef CONFIG_QUOTA 4885 bool quota_enabled = false; 4886 #endif 4887 4888 try_onemore: 4889 err = -EINVAL; 4890 raw_super = NULL; 4891 valid_super_block = -1; 4892 recovery = 0; 4893 4894 /* allocate memory for f2fs-specific super block info */ 4895 sbi = kzalloc(sizeof(struct f2fs_sb_info), GFP_KERNEL); 4896 if (!sbi) 4897 return -ENOMEM; 4898 4899 sbi->sb = sb; 4900 4901 /* initialize locks within allocated memory */ 4902 init_f2fs_rwsem(&sbi->gc_lock); 4903 mutex_init(&sbi->writepages); 4904 init_f2fs_rwsem(&sbi->cp_global_sem); 4905 init_f2fs_rwsem(&sbi->node_write); 4906 init_f2fs_rwsem(&sbi->node_change); 4907 spin_lock_init(&sbi->stat_lock); 4908 init_f2fs_rwsem(&sbi->cp_rwsem); 4909 init_f2fs_rwsem(&sbi->cp_enable_rwsem); 4910 init_f2fs_rwsem(&sbi->quota_sem); 4911 init_waitqueue_head(&sbi->cp_wait); 4912 spin_lock_init(&sbi->error_lock); 4913 4914 for (i = 0; i < NR_INODE_TYPE; i++) { 4915 INIT_LIST_HEAD(&sbi->inode_list[i]); 4916 spin_lock_init(&sbi->inode_lock[i]); 4917 } 4918 mutex_init(&sbi->flush_lock); 4919 4920 /* set a block size */ 4921 if (unlikely(!sb_set_blocksize(sb, F2FS_BLKSIZE))) { 4922 f2fs_err(sbi, "unable to set blocksize"); 4923 goto free_sbi; 4924 } 4925 4926 err = read_raw_super_block(sbi, &raw_super, &valid_super_block, 4927 &recovery); 4928 if (err) 4929 goto free_sbi; 4930 4931 sb->s_fs_info = sbi; 4932 sbi->raw_super = raw_super; 4933 4934 INIT_WORK(&sbi->s_error_work, f2fs_record_error_work); 4935 memcpy(sbi->errors, raw_super->s_errors, MAX_F2FS_ERRORS); 4936 memcpy(sbi->stop_reason, raw_super->s_stop_reason, MAX_STOP_REASON); 4937 4938 /* precompute checksum seed for metadata */ 4939 if (f2fs_sb_has_inode_chksum(sbi)) 4940 sbi->s_chksum_seed = f2fs_chksum(~0, raw_super->uuid, 4941 sizeof(raw_super->uuid)); 4942 4943 default_options(sbi, false); 4944 4945 err = f2fs_check_opt_consistency(fc, sb); 4946 if (err) 4947 goto free_sb_buf; 4948 4949 f2fs_apply_options(fc, sb); 4950 4951 err = f2fs_sanity_check_options(sbi, false); 4952 if (err) 4953 goto free_options; 4954 4955 sb->s_maxbytes = max_file_blocks(NULL) << 4956 le32_to_cpu(raw_super->log_blocksize); 4957 sb->s_max_links = F2FS_LINK_MAX; 4958 4959 err = f2fs_setup_casefold(sbi); 4960 if (err) 4961 goto free_options; 4962 4963 #ifdef CONFIG_QUOTA 4964 sb->dq_op = &f2fs_quota_operations; 4965 sb->s_qcop = &f2fs_quotactl_ops; 4966 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ; 4967 4968 if (f2fs_sb_has_quota_ino(sbi)) { 4969 for (i = 0; i < MAXQUOTAS; i++) { 4970 if (f2fs_qf_ino(sbi->sb, i)) 4971 sbi->nquota_files++; 4972 } 4973 } 4974 #endif 4975 4976 sb->s_op = &f2fs_sops; 4977 #ifdef CONFIG_FS_ENCRYPTION 4978 sb->s_cop = &f2fs_cryptops; 4979 #endif 4980 #ifdef CONFIG_FS_VERITY 4981 sb->s_vop = &f2fs_verityops; 4982 #endif 4983 sb->s_xattr = f2fs_xattr_handlers; 4984 sb->s_export_op = &f2fs_export_ops; 4985 sb->s_magic = F2FS_SUPER_MAGIC; 4986 sb->s_time_gran = 1; 4987 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) | 4988 (test_opt(sbi, POSIX_ACL) ? SB_POSIXACL : 0); 4989 if (test_opt(sbi, INLINECRYPT)) 4990 sb->s_flags |= SB_INLINECRYPT; 4991 4992 if (test_opt(sbi, LAZYTIME)) 4993 sb->s_flags |= SB_LAZYTIME; 4994 else 4995 sb->s_flags &= ~SB_LAZYTIME; 4996 4997 super_set_uuid(sb, (void *) raw_super->uuid, sizeof(raw_super->uuid)); 4998 super_set_sysfs_name_bdev(sb); 4999 sb->s_iflags |= SB_I_CGROUPWB; 5000 5001 /* init f2fs-specific super block info */ 5002 sbi->valid_super_block = valid_super_block; 5003 5004 /* disallow all the data/node/meta page writes */ 5005 set_sbi_flag(sbi, SBI_POR_DOING); 5006 5007 err = f2fs_init_write_merge_io(sbi); 5008 if (err) 5009 goto free_bio_info; 5010 5011 init_sb_info(sbi); 5012 5013 err = f2fs_init_iostat(sbi); 5014 if (err) 5015 goto free_bio_info; 5016 5017 err = init_percpu_info(sbi); 5018 if (err) 5019 goto free_iostat; 5020 5021 err = f2fs_init_page_array_cache(sbi); 5022 if (err) 5023 goto free_percpu; 5024 5025 /* get an inode for meta space */ 5026 sbi->meta_inode = f2fs_iget(sb, F2FS_META_INO(sbi)); 5027 if (IS_ERR(sbi->meta_inode)) { 5028 f2fs_err(sbi, "Failed to read F2FS meta data inode"); 5029 err = PTR_ERR(sbi->meta_inode); 5030 goto free_page_array_cache; 5031 } 5032 5033 err = f2fs_get_valid_checkpoint(sbi); 5034 if (err) { 5035 f2fs_err(sbi, "Failed to get valid F2FS checkpoint"); 5036 goto free_meta_inode; 5037 } 5038 5039 if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_QUOTA_NEED_FSCK_FLAG)) 5040 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR); 5041 if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_DISABLED_QUICK_FLAG)) { 5042 set_sbi_flag(sbi, SBI_CP_DISABLED_QUICK); 5043 sbi->interval_time[DISABLE_TIME] = DEF_DISABLE_QUICK_INTERVAL; 5044 } 5045 5046 if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_FSCK_FLAG)) 5047 set_sbi_flag(sbi, SBI_NEED_FSCK); 5048 5049 /* Initialize device list */ 5050 err = f2fs_scan_devices(sbi); 5051 if (err) { 5052 f2fs_err(sbi, "Failed to find devices"); 5053 goto free_devices; 5054 } 5055 5056 err = f2fs_init_post_read_wq(sbi); 5057 if (err) { 5058 f2fs_err(sbi, "Failed to initialize post read workqueue"); 5059 goto free_devices; 5060 } 5061 5062 sbi->total_valid_node_count = 5063 le32_to_cpu(sbi->ckpt->valid_node_count); 5064 percpu_counter_set(&sbi->total_valid_inode_count, 5065 le32_to_cpu(sbi->ckpt->valid_inode_count)); 5066 sbi->user_block_count = le64_to_cpu(sbi->ckpt->user_block_count); 5067 sbi->total_valid_block_count = 5068 le64_to_cpu(sbi->ckpt->valid_block_count); 5069 sbi->last_valid_block_count = sbi->total_valid_block_count; 5070 sbi->reserved_blocks = 0; 5071 sbi->current_reserved_blocks = 0; 5072 limit_reserve_root(sbi); 5073 adjust_unusable_cap_perc(sbi); 5074 5075 f2fs_init_extent_cache_info(sbi); 5076 5077 f2fs_init_ino_entry_info(sbi); 5078 5079 f2fs_init_fsync_node_info(sbi); 5080 5081 /* setup checkpoint request control and start checkpoint issue thread */ 5082 f2fs_init_ckpt_req_control(sbi); 5083 if (!f2fs_readonly(sb) && !test_opt(sbi, DISABLE_CHECKPOINT) && 5084 test_opt(sbi, MERGE_CHECKPOINT)) { 5085 err = f2fs_start_ckpt_thread(sbi); 5086 if (err) { 5087 f2fs_err(sbi, 5088 "Failed to start F2FS issue_checkpoint_thread (%d)", 5089 err); 5090 goto stop_ckpt_thread; 5091 } 5092 } 5093 5094 /* setup f2fs internal modules */ 5095 err = f2fs_build_segment_manager(sbi); 5096 if (err) { 5097 f2fs_err(sbi, "Failed to initialize F2FS segment manager (%d)", 5098 err); 5099 goto free_sm; 5100 } 5101 err = f2fs_build_node_manager(sbi); 5102 if (err) { 5103 f2fs_err(sbi, "Failed to initialize F2FS node manager (%d)", 5104 err); 5105 goto free_nm; 5106 } 5107 5108 /* For write statistics */ 5109 sbi->sectors_written_start = f2fs_get_sectors_written(sbi); 5110 5111 /* get segno of first zoned block device */ 5112 sbi->first_seq_zone_segno = get_first_seq_zone_segno(sbi); 5113 5114 sbi->reserved_pin_section = f2fs_sb_has_blkzoned(sbi) ? 5115 ZONED_PIN_SEC_REQUIRED_COUNT : 5116 GET_SEC_FROM_SEG(sbi, overprovision_segments(sbi)); 5117 5118 /* Read accumulated write IO statistics if exists */ 5119 seg_i = CURSEG_I(sbi, CURSEG_HOT_NODE); 5120 if (__exist_node_summaries(sbi)) 5121 sbi->kbytes_written = 5122 le64_to_cpu(seg_i->journal->info.kbytes_written); 5123 5124 f2fs_build_gc_manager(sbi); 5125 5126 err = f2fs_build_stats(sbi); 5127 if (err) 5128 goto free_nm; 5129 5130 /* get an inode for node space */ 5131 sbi->node_inode = f2fs_iget(sb, F2FS_NODE_INO(sbi)); 5132 if (IS_ERR(sbi->node_inode)) { 5133 f2fs_err(sbi, "Failed to read node inode"); 5134 err = PTR_ERR(sbi->node_inode); 5135 goto free_stats; 5136 } 5137 5138 /* read root inode and dentry */ 5139 root = f2fs_iget(sb, F2FS_ROOT_INO(sbi)); 5140 if (IS_ERR(root)) { 5141 f2fs_err(sbi, "Failed to read root inode"); 5142 err = PTR_ERR(root); 5143 goto free_node_inode; 5144 } 5145 if (!S_ISDIR(root->i_mode) || !root->i_blocks || 5146 !root->i_size || !root->i_nlink) { 5147 iput(root); 5148 err = -EINVAL; 5149 goto free_node_inode; 5150 } 5151 5152 generic_set_sb_d_ops(sb); 5153 sb->s_root = d_make_root(root); /* allocate root dentry */ 5154 if (!sb->s_root) { 5155 err = -ENOMEM; 5156 goto free_node_inode; 5157 } 5158 5159 err = f2fs_init_compress_inode(sbi); 5160 if (err) 5161 goto free_root_inode; 5162 5163 err = f2fs_register_sysfs(sbi); 5164 if (err) 5165 goto free_compress_inode; 5166 5167 sbi->umount_lock_holder = current; 5168 #ifdef CONFIG_QUOTA 5169 /* Enable quota usage during mount */ 5170 if (f2fs_sb_has_quota_ino(sbi) && !f2fs_readonly(sb)) { 5171 err = f2fs_enable_quotas(sb); 5172 if (err) 5173 f2fs_err(sbi, "Cannot turn on quotas: error %d", err); 5174 } 5175 5176 quota_enabled = f2fs_recover_quota_begin(sbi); 5177 #endif 5178 /* if there are any orphan inodes, free them */ 5179 err = f2fs_recover_orphan_inodes(sbi); 5180 if (err) 5181 goto free_meta; 5182 5183 if (unlikely(is_set_ckpt_flags(sbi, CP_DISABLED_FLAG))) { 5184 skip_recovery = true; 5185 goto reset_checkpoint; 5186 } 5187 5188 /* recover fsynced data */ 5189 if (!test_opt(sbi, DISABLE_ROLL_FORWARD) && 5190 !test_opt(sbi, NORECOVERY)) { 5191 /* 5192 * mount should be failed, when device has readonly mode, and 5193 * previous checkpoint was not done by clean system shutdown. 5194 */ 5195 if (f2fs_hw_is_readonly(sbi)) { 5196 if (!is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG)) { 5197 err = f2fs_recover_fsync_data(sbi, true); 5198 if (err > 0) { 5199 err = -EROFS; 5200 f2fs_err(sbi, "Need to recover fsync data, but " 5201 "write access unavailable, please try " 5202 "mount w/ disable_roll_forward or norecovery"); 5203 } 5204 if (err < 0) 5205 goto free_meta; 5206 } 5207 f2fs_info(sbi, "write access unavailable, skipping recovery"); 5208 goto reset_checkpoint; 5209 } 5210 5211 if (need_fsck) 5212 set_sbi_flag(sbi, SBI_NEED_FSCK); 5213 5214 if (skip_recovery) 5215 goto reset_checkpoint; 5216 5217 err = f2fs_recover_fsync_data(sbi, false); 5218 if (err < 0) { 5219 if (err != -ENOMEM) 5220 skip_recovery = true; 5221 need_fsck = true; 5222 f2fs_err(sbi, "Cannot recover all fsync data errno=%d", 5223 err); 5224 goto free_meta; 5225 } 5226 } else { 5227 err = f2fs_recover_fsync_data(sbi, true); 5228 if (err > 0) { 5229 if (!f2fs_readonly(sb)) { 5230 f2fs_err(sbi, "Need to recover fsync data"); 5231 err = -EINVAL; 5232 goto free_meta; 5233 } else { 5234 f2fs_info(sbi, "drop all fsynced data"); 5235 err = 0; 5236 } 5237 } 5238 } 5239 5240 reset_checkpoint: 5241 #ifdef CONFIG_QUOTA 5242 f2fs_recover_quota_end(sbi, quota_enabled); 5243 #endif 5244 /* 5245 * If the f2fs is not readonly and fsync data recovery succeeds, 5246 * write pointer consistency of cursegs and other zones are already 5247 * checked and fixed during recovery. However, if recovery fails, 5248 * write pointers are left untouched, and retry-mount should check 5249 * them here. 5250 */ 5251 if (skip_recovery) 5252 err = f2fs_check_and_fix_write_pointer(sbi); 5253 if (err) 5254 goto free_meta; 5255 5256 /* f2fs_recover_fsync_data() cleared this already */ 5257 clear_sbi_flag(sbi, SBI_POR_DOING); 5258 5259 err = f2fs_init_inmem_curseg(sbi); 5260 if (err) 5261 goto sync_free_meta; 5262 5263 if (test_opt(sbi, DISABLE_CHECKPOINT)) 5264 err = f2fs_disable_checkpoint(sbi); 5265 else if (is_set_ckpt_flags(sbi, CP_DISABLED_FLAG)) 5266 err = f2fs_enable_checkpoint(sbi); 5267 if (err) 5268 goto sync_free_meta; 5269 5270 /* 5271 * If filesystem is not mounted as read-only then 5272 * do start the gc_thread. 5273 */ 5274 if ((F2FS_OPTION(sbi).bggc_mode != BGGC_MODE_OFF || 5275 test_opt(sbi, GC_MERGE)) && !f2fs_readonly(sb)) { 5276 /* After POR, we can run background GC thread.*/ 5277 err = f2fs_start_gc_thread(sbi); 5278 if (err) 5279 goto sync_free_meta; 5280 } 5281 5282 /* recover broken superblock */ 5283 if (recovery) { 5284 err = f2fs_commit_super(sbi, true); 5285 f2fs_info(sbi, "Try to recover %dth superblock, ret: %d", 5286 sbi->valid_super_block ? 1 : 2, err); 5287 } 5288 5289 f2fs_join_shrinker(sbi); 5290 5291 f2fs_tuning_parameters(sbi); 5292 5293 f2fs_notice(sbi, "Mounted with checkpoint version = %llx", 5294 cur_cp_version(F2FS_CKPT(sbi))); 5295 f2fs_update_time(sbi, CP_TIME); 5296 f2fs_update_time(sbi, REQ_TIME); 5297 clear_sbi_flag(sbi, SBI_CP_DISABLED_QUICK); 5298 5299 sbi->umount_lock_holder = NULL; 5300 return 0; 5301 5302 sync_free_meta: 5303 /* safe to flush all the data */ 5304 sync_filesystem(sbi->sb); 5305 retry_cnt = 0; 5306 5307 free_meta: 5308 #ifdef CONFIG_QUOTA 5309 f2fs_truncate_quota_inode_pages(sb); 5310 if (f2fs_sb_has_quota_ino(sbi) && !f2fs_readonly(sb)) 5311 f2fs_quota_off_umount(sbi->sb); 5312 #endif 5313 /* 5314 * Some dirty meta pages can be produced by f2fs_recover_orphan_inodes() 5315 * failed by EIO. Then, iput(node_inode) can trigger balance_fs_bg() 5316 * followed by f2fs_write_checkpoint() through f2fs_write_node_pages(), which 5317 * falls into an infinite loop in f2fs_sync_meta_pages(). 5318 */ 5319 truncate_inode_pages_final(META_MAPPING(sbi)); 5320 /* evict some inodes being cached by GC */ 5321 evict_inodes(sb); 5322 f2fs_unregister_sysfs(sbi); 5323 free_compress_inode: 5324 f2fs_destroy_compress_inode(sbi); 5325 free_root_inode: 5326 dput(sb->s_root); 5327 sb->s_root = NULL; 5328 free_node_inode: 5329 f2fs_release_ino_entry(sbi, true); 5330 truncate_inode_pages_final(NODE_MAPPING(sbi)); 5331 iput(sbi->node_inode); 5332 sbi->node_inode = NULL; 5333 free_stats: 5334 f2fs_destroy_stats(sbi); 5335 free_nm: 5336 /* stop discard thread before destroying node manager */ 5337 f2fs_stop_discard_thread(sbi); 5338 f2fs_destroy_node_manager(sbi); 5339 free_sm: 5340 f2fs_destroy_segment_manager(sbi); 5341 stop_ckpt_thread: 5342 f2fs_stop_ckpt_thread(sbi); 5343 /* flush s_error_work before sbi destroy */ 5344 flush_work(&sbi->s_error_work); 5345 f2fs_destroy_post_read_wq(sbi); 5346 free_devices: 5347 destroy_device_list(sbi); 5348 kvfree(sbi->ckpt); 5349 free_meta_inode: 5350 make_bad_inode(sbi->meta_inode); 5351 iput(sbi->meta_inode); 5352 sbi->meta_inode = NULL; 5353 free_page_array_cache: 5354 f2fs_destroy_page_array_cache(sbi); 5355 free_percpu: 5356 destroy_percpu_info(sbi); 5357 free_iostat: 5358 f2fs_destroy_iostat(sbi); 5359 free_bio_info: 5360 for (i = 0; i < NR_PAGE_TYPE; i++) 5361 kfree(sbi->write_io[i]); 5362 5363 #if IS_ENABLED(CONFIG_UNICODE) 5364 utf8_unload(sb->s_encoding); 5365 sb->s_encoding = NULL; 5366 #endif 5367 free_options: 5368 #ifdef CONFIG_QUOTA 5369 for (i = 0; i < MAXQUOTAS; i++) 5370 kfree(F2FS_OPTION(sbi).s_qf_names[i]); 5371 #endif 5372 /* no need to free dummy_enc_policy, we just keep it in ctx when failed */ 5373 swap(F2FS_CTX_INFO(ctx).dummy_enc_policy, F2FS_OPTION(sbi).dummy_enc_policy); 5374 free_sb_buf: 5375 kfree(raw_super); 5376 free_sbi: 5377 kfree(sbi); 5378 sb->s_fs_info = NULL; 5379 5380 /* give only one another chance */ 5381 if (retry_cnt > 0 && skip_recovery) { 5382 retry_cnt--; 5383 shrink_dcache_sb(sb); 5384 goto try_onemore; 5385 } 5386 return err; 5387 } 5388 5389 static int f2fs_get_tree(struct fs_context *fc) 5390 { 5391 return get_tree_bdev(fc, f2fs_fill_super); 5392 } 5393 5394 static int f2fs_reconfigure(struct fs_context *fc) 5395 { 5396 struct super_block *sb = fc->root->d_sb; 5397 5398 return __f2fs_remount(fc, sb); 5399 } 5400 5401 static void f2fs_fc_free(struct fs_context *fc) 5402 { 5403 struct f2fs_fs_context *ctx = fc->fs_private; 5404 5405 if (!ctx) 5406 return; 5407 5408 #ifdef CONFIG_QUOTA 5409 f2fs_unnote_qf_name_all(fc); 5410 #endif 5411 fscrypt_free_dummy_policy(&F2FS_CTX_INFO(ctx).dummy_enc_policy); 5412 kfree(ctx); 5413 } 5414 5415 static const struct fs_context_operations f2fs_context_ops = { 5416 .parse_param = f2fs_parse_param, 5417 .get_tree = f2fs_get_tree, 5418 .reconfigure = f2fs_reconfigure, 5419 .free = f2fs_fc_free, 5420 }; 5421 5422 static void kill_f2fs_super(struct super_block *sb) 5423 { 5424 struct f2fs_sb_info *sbi = F2FS_SB(sb); 5425 5426 if (sb->s_root) { 5427 sbi->umount_lock_holder = current; 5428 5429 set_sbi_flag(sbi, SBI_IS_CLOSE); 5430 f2fs_stop_gc_thread(sbi); 5431 f2fs_stop_discard_thread(sbi); 5432 5433 #ifdef CONFIG_F2FS_FS_COMPRESSION 5434 /* 5435 * latter evict_inode() can bypass checking and invalidating 5436 * compress inode cache. 5437 */ 5438 if (test_opt(sbi, COMPRESS_CACHE)) 5439 truncate_inode_pages_final(COMPRESS_MAPPING(sbi)); 5440 #endif 5441 5442 if (is_sbi_flag_set(sbi, SBI_IS_DIRTY) || 5443 !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG)) { 5444 struct cp_control cpc = { 5445 .reason = CP_UMOUNT, 5446 }; 5447 stat_inc_cp_call_count(sbi, TOTAL_CALL); 5448 f2fs_write_checkpoint(sbi, &cpc); 5449 } 5450 5451 if (is_sbi_flag_set(sbi, SBI_IS_RECOVERED) && f2fs_readonly(sb)) 5452 sb->s_flags &= ~SB_RDONLY; 5453 } 5454 kill_block_super(sb); 5455 /* Release block devices last, after fscrypt_destroy_keyring(). */ 5456 if (sbi) { 5457 destroy_device_list(sbi); 5458 kfree(sbi); 5459 sb->s_fs_info = NULL; 5460 } 5461 } 5462 5463 static int f2fs_init_fs_context(struct fs_context *fc) 5464 { 5465 struct f2fs_fs_context *ctx; 5466 5467 ctx = kzalloc(sizeof(struct f2fs_fs_context), GFP_KERNEL); 5468 if (!ctx) 5469 return -ENOMEM; 5470 5471 fc->fs_private = ctx; 5472 fc->ops = &f2fs_context_ops; 5473 5474 return 0; 5475 } 5476 5477 static struct file_system_type f2fs_fs_type = { 5478 .owner = THIS_MODULE, 5479 .name = "f2fs", 5480 .init_fs_context = f2fs_init_fs_context, 5481 .kill_sb = kill_f2fs_super, 5482 .fs_flags = FS_REQUIRES_DEV | FS_ALLOW_IDMAP, 5483 }; 5484 MODULE_ALIAS_FS("f2fs"); 5485 5486 static int __init init_inodecache(void) 5487 { 5488 f2fs_inode_cachep = kmem_cache_create("f2fs_inode_cache", 5489 sizeof(struct f2fs_inode_info), 0, 5490 SLAB_RECLAIM_ACCOUNT|SLAB_ACCOUNT, NULL); 5491 return f2fs_inode_cachep ? 0 : -ENOMEM; 5492 } 5493 5494 static void destroy_inodecache(void) 5495 { 5496 /* 5497 * Make sure all delayed rcu free inodes are flushed before we 5498 * destroy cache. 5499 */ 5500 rcu_barrier(); 5501 kmem_cache_destroy(f2fs_inode_cachep); 5502 } 5503 5504 static int __init init_f2fs_fs(void) 5505 { 5506 int err; 5507 5508 err = init_inodecache(); 5509 if (err) 5510 goto fail; 5511 err = f2fs_create_node_manager_caches(); 5512 if (err) 5513 goto free_inodecache; 5514 err = f2fs_create_segment_manager_caches(); 5515 if (err) 5516 goto free_node_manager_caches; 5517 err = f2fs_create_checkpoint_caches(); 5518 if (err) 5519 goto free_segment_manager_caches; 5520 err = f2fs_create_recovery_cache(); 5521 if (err) 5522 goto free_checkpoint_caches; 5523 err = f2fs_create_extent_cache(); 5524 if (err) 5525 goto free_recovery_cache; 5526 err = f2fs_create_garbage_collection_cache(); 5527 if (err) 5528 goto free_extent_cache; 5529 err = f2fs_init_sysfs(); 5530 if (err) 5531 goto free_garbage_collection_cache; 5532 err = f2fs_init_shrinker(); 5533 if (err) 5534 goto free_sysfs; 5535 f2fs_create_root_stats(); 5536 err = f2fs_init_post_read_processing(); 5537 if (err) 5538 goto free_root_stats; 5539 err = f2fs_init_iostat_processing(); 5540 if (err) 5541 goto free_post_read; 5542 err = f2fs_init_bio_entry_cache(); 5543 if (err) 5544 goto free_iostat; 5545 err = f2fs_init_bioset(); 5546 if (err) 5547 goto free_bio_entry_cache; 5548 err = f2fs_init_compress_mempool(); 5549 if (err) 5550 goto free_bioset; 5551 err = f2fs_init_compress_cache(); 5552 if (err) 5553 goto free_compress_mempool; 5554 err = f2fs_create_casefold_cache(); 5555 if (err) 5556 goto free_compress_cache; 5557 err = f2fs_init_xattr_cache(); 5558 if (err) 5559 goto free_casefold_cache; 5560 err = register_filesystem(&f2fs_fs_type); 5561 if (err) 5562 goto free_xattr_cache; 5563 return 0; 5564 free_xattr_cache: 5565 f2fs_destroy_xattr_cache(); 5566 free_casefold_cache: 5567 f2fs_destroy_casefold_cache(); 5568 free_compress_cache: 5569 f2fs_destroy_compress_cache(); 5570 free_compress_mempool: 5571 f2fs_destroy_compress_mempool(); 5572 free_bioset: 5573 f2fs_destroy_bioset(); 5574 free_bio_entry_cache: 5575 f2fs_destroy_bio_entry_cache(); 5576 free_iostat: 5577 f2fs_destroy_iostat_processing(); 5578 free_post_read: 5579 f2fs_destroy_post_read_processing(); 5580 free_root_stats: 5581 f2fs_destroy_root_stats(); 5582 f2fs_exit_shrinker(); 5583 free_sysfs: 5584 f2fs_exit_sysfs(); 5585 free_garbage_collection_cache: 5586 f2fs_destroy_garbage_collection_cache(); 5587 free_extent_cache: 5588 f2fs_destroy_extent_cache(); 5589 free_recovery_cache: 5590 f2fs_destroy_recovery_cache(); 5591 free_checkpoint_caches: 5592 f2fs_destroy_checkpoint_caches(); 5593 free_segment_manager_caches: 5594 f2fs_destroy_segment_manager_caches(); 5595 free_node_manager_caches: 5596 f2fs_destroy_node_manager_caches(); 5597 free_inodecache: 5598 destroy_inodecache(); 5599 fail: 5600 return err; 5601 } 5602 5603 static void __exit exit_f2fs_fs(void) 5604 { 5605 unregister_filesystem(&f2fs_fs_type); 5606 f2fs_destroy_xattr_cache(); 5607 f2fs_destroy_casefold_cache(); 5608 f2fs_destroy_compress_cache(); 5609 f2fs_destroy_compress_mempool(); 5610 f2fs_destroy_bioset(); 5611 f2fs_destroy_bio_entry_cache(); 5612 f2fs_destroy_iostat_processing(); 5613 f2fs_destroy_post_read_processing(); 5614 f2fs_destroy_root_stats(); 5615 f2fs_exit_shrinker(); 5616 f2fs_exit_sysfs(); 5617 f2fs_destroy_garbage_collection_cache(); 5618 f2fs_destroy_extent_cache(); 5619 f2fs_destroy_recovery_cache(); 5620 f2fs_destroy_checkpoint_caches(); 5621 f2fs_destroy_segment_manager_caches(); 5622 f2fs_destroy_node_manager_caches(); 5623 destroy_inodecache(); 5624 } 5625 5626 module_init(init_f2fs_fs) 5627 module_exit(exit_f2fs_fs) 5628 5629 MODULE_AUTHOR("Samsung Electronics's Praesto Team"); 5630 MODULE_DESCRIPTION("Flash Friendly File System"); 5631 MODULE_LICENSE("GPL"); 5632