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_resgid, 147 Opt_resuid, 148 Opt_mode, 149 Opt_fault_injection, 150 Opt_fault_type, 151 Opt_lazytime, 152 Opt_quota, 153 Opt_usrquota, 154 Opt_grpquota, 155 Opt_prjquota, 156 Opt_usrjquota, 157 Opt_grpjquota, 158 Opt_prjjquota, 159 Opt_alloc, 160 Opt_fsync, 161 Opt_test_dummy_encryption, 162 Opt_inlinecrypt, 163 Opt_checkpoint_disable, 164 Opt_checkpoint_disable_cap, 165 Opt_checkpoint_disable_cap_perc, 166 Opt_checkpoint_enable, 167 Opt_checkpoint_merge, 168 Opt_compress_algorithm, 169 Opt_compress_log_size, 170 Opt_nocompress_extension, 171 Opt_compress_extension, 172 Opt_compress_chksum, 173 Opt_compress_mode, 174 Opt_compress_cache, 175 Opt_atgc, 176 Opt_gc_merge, 177 Opt_discard_unit, 178 Opt_memory_mode, 179 Opt_age_extent_cache, 180 Opt_errors, 181 Opt_nat_bits, 182 Opt_jqfmt, 183 Opt_checkpoint, 184 Opt_err, 185 }; 186 187 static const struct constant_table f2fs_param_background_gc[] = { 188 {"on", BGGC_MODE_ON}, 189 {"off", BGGC_MODE_OFF}, 190 {"sync", BGGC_MODE_SYNC}, 191 {} 192 }; 193 194 static const struct constant_table f2fs_param_mode[] = { 195 {"adaptive", FS_MODE_ADAPTIVE}, 196 {"lfs", FS_MODE_LFS}, 197 {"fragment:segment", FS_MODE_FRAGMENT_SEG}, 198 {"fragment:block", FS_MODE_FRAGMENT_BLK}, 199 {} 200 }; 201 202 static const struct constant_table f2fs_param_jqfmt[] = { 203 {"vfsold", QFMT_VFS_OLD}, 204 {"vfsv0", QFMT_VFS_V0}, 205 {"vfsv1", QFMT_VFS_V1}, 206 {} 207 }; 208 209 static const struct constant_table f2fs_param_alloc_mode[] = { 210 {"default", ALLOC_MODE_DEFAULT}, 211 {"reuse", ALLOC_MODE_REUSE}, 212 {} 213 }; 214 static const struct constant_table f2fs_param_fsync_mode[] = { 215 {"posix", FSYNC_MODE_POSIX}, 216 {"strict", FSYNC_MODE_STRICT}, 217 {"nobarrier", FSYNC_MODE_NOBARRIER}, 218 {} 219 }; 220 221 static const struct constant_table f2fs_param_compress_mode[] = { 222 {"fs", COMPR_MODE_FS}, 223 {"user", COMPR_MODE_USER}, 224 {} 225 }; 226 227 static const struct constant_table f2fs_param_discard_unit[] = { 228 {"block", DISCARD_UNIT_BLOCK}, 229 {"segment", DISCARD_UNIT_SEGMENT}, 230 {"section", DISCARD_UNIT_SECTION}, 231 {} 232 }; 233 234 static const struct constant_table f2fs_param_memory_mode[] = { 235 {"normal", MEMORY_MODE_NORMAL}, 236 {"low", MEMORY_MODE_LOW}, 237 {} 238 }; 239 240 static const struct constant_table f2fs_param_errors[] = { 241 {"remount-ro", MOUNT_ERRORS_READONLY}, 242 {"continue", MOUNT_ERRORS_CONTINUE}, 243 {"panic", MOUNT_ERRORS_PANIC}, 244 {} 245 }; 246 247 static const struct fs_parameter_spec f2fs_param_specs[] = { 248 fsparam_enum("background_gc", Opt_gc_background, f2fs_param_background_gc), 249 fsparam_flag("disable_roll_forward", Opt_disable_roll_forward), 250 fsparam_flag("norecovery", Opt_norecovery), 251 fsparam_flag_no("discard", Opt_discard), 252 fsparam_flag("no_heap", Opt_noheap), 253 fsparam_flag("heap", Opt_heap), 254 fsparam_flag_no("user_xattr", Opt_user_xattr), 255 fsparam_flag_no("acl", Opt_acl), 256 fsparam_s32("active_logs", Opt_active_logs), 257 fsparam_flag("disable_ext_identify", Opt_disable_ext_identify), 258 fsparam_flag_no("inline_xattr", Opt_inline_xattr), 259 fsparam_s32("inline_xattr_size", Opt_inline_xattr_size), 260 fsparam_flag_no("inline_data", Opt_inline_data), 261 fsparam_flag_no("inline_dentry", Opt_inline_dentry), 262 fsparam_flag_no("flush_merge", Opt_flush_merge), 263 fsparam_flag_no("barrier", Opt_barrier), 264 fsparam_flag("fastboot", Opt_fastboot), 265 fsparam_flag_no("extent_cache", Opt_extent_cache), 266 fsparam_flag("data_flush", Opt_data_flush), 267 fsparam_u32("reserve_root", Opt_reserve_root), 268 fsparam_gid("resgid", Opt_resgid), 269 fsparam_uid("resuid", Opt_resuid), 270 fsparam_enum("mode", Opt_mode, f2fs_param_mode), 271 fsparam_s32("fault_injection", Opt_fault_injection), 272 fsparam_u32("fault_type", Opt_fault_type), 273 fsparam_flag_no("lazytime", Opt_lazytime), 274 fsparam_flag_no("quota", Opt_quota), 275 fsparam_flag("usrquota", Opt_usrquota), 276 fsparam_flag("grpquota", Opt_grpquota), 277 fsparam_flag("prjquota", Opt_prjquota), 278 fsparam_string_empty("usrjquota", Opt_usrjquota), 279 fsparam_string_empty("grpjquota", Opt_grpjquota), 280 fsparam_string_empty("prjjquota", Opt_prjjquota), 281 fsparam_flag("nat_bits", Opt_nat_bits), 282 fsparam_enum("jqfmt", Opt_jqfmt, f2fs_param_jqfmt), 283 fsparam_enum("alloc_mode", Opt_alloc, f2fs_param_alloc_mode), 284 fsparam_enum("fsync_mode", Opt_fsync, f2fs_param_fsync_mode), 285 fsparam_string("test_dummy_encryption", Opt_test_dummy_encryption), 286 fsparam_flag("test_dummy_encryption", Opt_test_dummy_encryption), 287 fsparam_flag("inlinecrypt", Opt_inlinecrypt), 288 fsparam_string("checkpoint", Opt_checkpoint), 289 fsparam_flag_no("checkpoint_merge", Opt_checkpoint_merge), 290 fsparam_string("compress_algorithm", Opt_compress_algorithm), 291 fsparam_u32("compress_log_size", Opt_compress_log_size), 292 fsparam_string("compress_extension", Opt_compress_extension), 293 fsparam_string("nocompress_extension", Opt_nocompress_extension), 294 fsparam_flag("compress_chksum", Opt_compress_chksum), 295 fsparam_enum("compress_mode", Opt_compress_mode, f2fs_param_compress_mode), 296 fsparam_flag("compress_cache", Opt_compress_cache), 297 fsparam_flag("atgc", Opt_atgc), 298 fsparam_flag_no("gc_merge", Opt_gc_merge), 299 fsparam_enum("discard_unit", Opt_discard_unit, f2fs_param_discard_unit), 300 fsparam_enum("memory", Opt_memory_mode, f2fs_param_memory_mode), 301 fsparam_flag("age_extent_cache", Opt_age_extent_cache), 302 fsparam_enum("errors", Opt_errors, f2fs_param_errors), 303 {} 304 }; 305 306 /* Resort to a match_table for this interestingly formatted option */ 307 static match_table_t f2fs_checkpoint_tokens = { 308 {Opt_checkpoint_disable, "disable"}, 309 {Opt_checkpoint_disable_cap, "disable:%u"}, 310 {Opt_checkpoint_disable_cap_perc, "disable:%u%%"}, 311 {Opt_checkpoint_enable, "enable"}, 312 {Opt_err, NULL}, 313 }; 314 315 #define F2FS_SPEC_background_gc (1 << 0) 316 #define F2FS_SPEC_inline_xattr_size (1 << 1) 317 #define F2FS_SPEC_active_logs (1 << 2) 318 #define F2FS_SPEC_reserve_root (1 << 3) 319 #define F2FS_SPEC_resgid (1 << 4) 320 #define F2FS_SPEC_resuid (1 << 5) 321 #define F2FS_SPEC_mode (1 << 6) 322 #define F2FS_SPEC_fault_injection (1 << 7) 323 #define F2FS_SPEC_fault_type (1 << 8) 324 #define F2FS_SPEC_jqfmt (1 << 9) 325 #define F2FS_SPEC_alloc_mode (1 << 10) 326 #define F2FS_SPEC_fsync_mode (1 << 11) 327 #define F2FS_SPEC_checkpoint_disable_cap (1 << 12) 328 #define F2FS_SPEC_checkpoint_disable_cap_perc (1 << 13) 329 #define F2FS_SPEC_compress_level (1 << 14) 330 #define F2FS_SPEC_compress_algorithm (1 << 15) 331 #define F2FS_SPEC_compress_log_size (1 << 16) 332 #define F2FS_SPEC_compress_extension (1 << 17) 333 #define F2FS_SPEC_nocompress_extension (1 << 18) 334 #define F2FS_SPEC_compress_chksum (1 << 19) 335 #define F2FS_SPEC_compress_mode (1 << 20) 336 #define F2FS_SPEC_discard_unit (1 << 21) 337 #define F2FS_SPEC_memory_mode (1 << 22) 338 #define F2FS_SPEC_errors (1 << 23) 339 340 struct f2fs_fs_context { 341 struct f2fs_mount_info info; 342 unsigned int opt_mask; /* Bits changed */ 343 unsigned int spec_mask; 344 unsigned short qname_mask; 345 }; 346 347 #define F2FS_CTX_INFO(ctx) ((ctx)->info) 348 349 static inline void ctx_set_opt(struct f2fs_fs_context *ctx, 350 unsigned int flag) 351 { 352 ctx->info.opt |= flag; 353 ctx->opt_mask |= flag; 354 } 355 356 static inline void ctx_clear_opt(struct f2fs_fs_context *ctx, 357 unsigned int flag) 358 { 359 ctx->info.opt &= ~flag; 360 ctx->opt_mask |= flag; 361 } 362 363 static inline bool ctx_test_opt(struct f2fs_fs_context *ctx, 364 unsigned int flag) 365 { 366 return ctx->info.opt & flag; 367 } 368 369 void f2fs_printk(struct f2fs_sb_info *sbi, bool limit_rate, 370 const char *fmt, ...) 371 { 372 struct va_format vaf; 373 va_list args; 374 int level; 375 376 va_start(args, fmt); 377 378 level = printk_get_level(fmt); 379 vaf.fmt = printk_skip_level(fmt); 380 vaf.va = &args; 381 if (limit_rate) 382 if (sbi) 383 printk_ratelimited("%c%cF2FS-fs (%s): %pV\n", 384 KERN_SOH_ASCII, level, sbi->sb->s_id, &vaf); 385 else 386 printk_ratelimited("%c%cF2FS-fs: %pV\n", 387 KERN_SOH_ASCII, level, &vaf); 388 else 389 if (sbi) 390 printk("%c%cF2FS-fs (%s): %pV\n", 391 KERN_SOH_ASCII, level, sbi->sb->s_id, &vaf); 392 else 393 printk("%c%cF2FS-fs: %pV\n", 394 KERN_SOH_ASCII, level, &vaf); 395 396 va_end(args); 397 } 398 399 #if IS_ENABLED(CONFIG_UNICODE) 400 static const struct f2fs_sb_encodings { 401 __u16 magic; 402 char *name; 403 unsigned int version; 404 } f2fs_sb_encoding_map[] = { 405 {F2FS_ENC_UTF8_12_1, "utf8", UNICODE_AGE(12, 1, 0)}, 406 }; 407 408 static const struct f2fs_sb_encodings * 409 f2fs_sb_read_encoding(const struct f2fs_super_block *sb) 410 { 411 __u16 magic = le16_to_cpu(sb->s_encoding); 412 int i; 413 414 for (i = 0; i < ARRAY_SIZE(f2fs_sb_encoding_map); i++) 415 if (magic == f2fs_sb_encoding_map[i].magic) 416 return &f2fs_sb_encoding_map[i]; 417 418 return NULL; 419 } 420 421 struct kmem_cache *f2fs_cf_name_slab; 422 static int __init f2fs_create_casefold_cache(void) 423 { 424 f2fs_cf_name_slab = f2fs_kmem_cache_create("f2fs_casefolded_name", 425 F2FS_NAME_LEN); 426 return f2fs_cf_name_slab ? 0 : -ENOMEM; 427 } 428 429 static void f2fs_destroy_casefold_cache(void) 430 { 431 kmem_cache_destroy(f2fs_cf_name_slab); 432 } 433 #else 434 static int __init f2fs_create_casefold_cache(void) { return 0; } 435 static void f2fs_destroy_casefold_cache(void) { } 436 #endif 437 438 static inline void limit_reserve_root(struct f2fs_sb_info *sbi) 439 { 440 block_t limit = min((sbi->user_block_count >> 3), 441 sbi->user_block_count - sbi->reserved_blocks); 442 443 /* limit is 12.5% */ 444 if (test_opt(sbi, RESERVE_ROOT) && 445 F2FS_OPTION(sbi).root_reserved_blocks > limit) { 446 F2FS_OPTION(sbi).root_reserved_blocks = limit; 447 f2fs_info(sbi, "Reduce reserved blocks for root = %u", 448 F2FS_OPTION(sbi).root_reserved_blocks); 449 } 450 if (!test_opt(sbi, RESERVE_ROOT) && 451 (!uid_eq(F2FS_OPTION(sbi).s_resuid, 452 make_kuid(&init_user_ns, F2FS_DEF_RESUID)) || 453 !gid_eq(F2FS_OPTION(sbi).s_resgid, 454 make_kgid(&init_user_ns, F2FS_DEF_RESGID)))) 455 f2fs_info(sbi, "Ignore s_resuid=%u, s_resgid=%u w/o reserve_root", 456 from_kuid_munged(&init_user_ns, 457 F2FS_OPTION(sbi).s_resuid), 458 from_kgid_munged(&init_user_ns, 459 F2FS_OPTION(sbi).s_resgid)); 460 } 461 462 static inline void adjust_unusable_cap_perc(struct f2fs_sb_info *sbi) 463 { 464 if (!F2FS_OPTION(sbi).unusable_cap_perc) 465 return; 466 467 if (F2FS_OPTION(sbi).unusable_cap_perc == 100) 468 F2FS_OPTION(sbi).unusable_cap = sbi->user_block_count; 469 else 470 F2FS_OPTION(sbi).unusable_cap = (sbi->user_block_count / 100) * 471 F2FS_OPTION(sbi).unusable_cap_perc; 472 473 f2fs_info(sbi, "Adjust unusable cap for checkpoint=disable = %u / %u%%", 474 F2FS_OPTION(sbi).unusable_cap, 475 F2FS_OPTION(sbi).unusable_cap_perc); 476 } 477 478 static void init_once(void *foo) 479 { 480 struct f2fs_inode_info *fi = (struct f2fs_inode_info *) foo; 481 482 inode_init_once(&fi->vfs_inode); 483 } 484 485 #ifdef CONFIG_QUOTA 486 static const char * const quotatypes[] = INITQFNAMES; 487 #define QTYPE2NAME(t) (quotatypes[t]) 488 /* 489 * Note the name of the specified quota file. 490 */ 491 static int f2fs_note_qf_name(struct fs_context *fc, int qtype, 492 struct fs_parameter *param) 493 { 494 struct f2fs_fs_context *ctx = fc->fs_private; 495 char *qname; 496 497 if (param->size < 1) { 498 f2fs_err(NULL, "Missing quota name"); 499 return -EINVAL; 500 } 501 if (strchr(param->string, '/')) { 502 f2fs_err(NULL, "quotafile must be on filesystem root"); 503 return -EINVAL; 504 } 505 if (ctx->info.s_qf_names[qtype]) { 506 if (strcmp(ctx->info.s_qf_names[qtype], param->string) != 0) { 507 f2fs_err(NULL, "Quota file already specified"); 508 return -EINVAL; 509 } 510 return 0; 511 } 512 513 qname = kmemdup_nul(param->string, param->size, GFP_KERNEL); 514 if (!qname) { 515 f2fs_err(NULL, "Not enough memory for storing quotafile name"); 516 return -ENOMEM; 517 } 518 F2FS_CTX_INFO(ctx).s_qf_names[qtype] = qname; 519 ctx->qname_mask |= 1 << qtype; 520 return 0; 521 } 522 523 /* 524 * Clear the name of the specified quota file. 525 */ 526 static int f2fs_unnote_qf_name(struct fs_context *fc, int qtype) 527 { 528 struct f2fs_fs_context *ctx = fc->fs_private; 529 530 kfree(ctx->info.s_qf_names[qtype]); 531 ctx->info.s_qf_names[qtype] = NULL; 532 ctx->qname_mask |= 1 << qtype; 533 return 0; 534 } 535 536 static void f2fs_unnote_qf_name_all(struct fs_context *fc) 537 { 538 int i; 539 540 for (i = 0; i < MAXQUOTAS; i++) 541 f2fs_unnote_qf_name(fc, i); 542 } 543 #endif 544 545 static int f2fs_parse_test_dummy_encryption(const struct fs_parameter *param, 546 struct f2fs_fs_context *ctx) 547 { 548 int err; 549 550 if (!IS_ENABLED(CONFIG_FS_ENCRYPTION)) { 551 f2fs_warn(NULL, "test_dummy_encryption option not supported"); 552 return -EINVAL; 553 } 554 err = fscrypt_parse_test_dummy_encryption(param, 555 &ctx->info.dummy_enc_policy); 556 if (err) { 557 if (err == -EINVAL) 558 f2fs_warn(NULL, "Value of option \"%s\" is unrecognized", 559 param->key); 560 else if (err == -EEXIST) 561 f2fs_warn(NULL, "Conflicting test_dummy_encryption options"); 562 else 563 f2fs_warn(NULL, "Error processing option \"%s\" [%d]", 564 param->key, err); 565 return -EINVAL; 566 } 567 return 0; 568 } 569 570 #ifdef CONFIG_F2FS_FS_COMPRESSION 571 static bool is_compress_extension_exist(struct f2fs_mount_info *info, 572 const char *new_ext, bool is_ext) 573 { 574 unsigned char (*ext)[F2FS_EXTENSION_LEN]; 575 int ext_cnt; 576 int i; 577 578 if (is_ext) { 579 ext = info->extensions; 580 ext_cnt = info->compress_ext_cnt; 581 } else { 582 ext = info->noextensions; 583 ext_cnt = info->nocompress_ext_cnt; 584 } 585 586 for (i = 0; i < ext_cnt; i++) { 587 if (!strcasecmp(new_ext, ext[i])) 588 return true; 589 } 590 591 return false; 592 } 593 594 /* 595 * 1. The same extension name cannot not appear in both compress and non-compress extension 596 * at the same time. 597 * 2. If the compress extension specifies all files, the types specified by the non-compress 598 * extension will be treated as special cases and will not be compressed. 599 * 3. Don't allow the non-compress extension specifies all files. 600 */ 601 static int f2fs_test_compress_extension(unsigned char (*noext)[F2FS_EXTENSION_LEN], 602 int noext_cnt, 603 unsigned char (*ext)[F2FS_EXTENSION_LEN], 604 int ext_cnt) 605 { 606 int index = 0, no_index = 0; 607 608 if (!noext_cnt) 609 return 0; 610 611 for (no_index = 0; no_index < noext_cnt; no_index++) { 612 if (strlen(noext[no_index]) == 0) 613 continue; 614 if (!strcasecmp("*", noext[no_index])) { 615 f2fs_info(NULL, "Don't allow the nocompress extension specifies all files"); 616 return -EINVAL; 617 } 618 for (index = 0; index < ext_cnt; index++) { 619 if (strlen(ext[index]) == 0) 620 continue; 621 if (!strcasecmp(ext[index], noext[no_index])) { 622 f2fs_info(NULL, "Don't allow the same extension %s appear in both compress and nocompress extension", 623 ext[index]); 624 return -EINVAL; 625 } 626 } 627 } 628 return 0; 629 } 630 631 #ifdef CONFIG_F2FS_FS_LZ4 632 static int f2fs_set_lz4hc_level(struct f2fs_fs_context *ctx, const char *str) 633 { 634 #ifdef CONFIG_F2FS_FS_LZ4HC 635 unsigned int level; 636 637 if (strlen(str) == 3) { 638 F2FS_CTX_INFO(ctx).compress_level = 0; 639 ctx->spec_mask |= F2FS_SPEC_compress_level; 640 return 0; 641 } 642 643 str += 3; 644 645 if (str[0] != ':') { 646 f2fs_info(NULL, "wrong format, e.g. <alg_name>:<compr_level>"); 647 return -EINVAL; 648 } 649 if (kstrtouint(str + 1, 10, &level)) 650 return -EINVAL; 651 652 if (!f2fs_is_compress_level_valid(COMPRESS_LZ4, level)) { 653 f2fs_info(NULL, "invalid lz4hc compress level: %d", level); 654 return -EINVAL; 655 } 656 657 F2FS_CTX_INFO(ctx).compress_level = level; 658 ctx->spec_mask |= F2FS_SPEC_compress_level; 659 return 0; 660 #else 661 if (strlen(str) == 3) { 662 F2FS_CTX_INFO(ctx).compress_level = 0; 663 ctx->spec_mask |= F2FS_SPEC_compress_level; 664 return 0; 665 } 666 f2fs_info(NULL, "kernel doesn't support lz4hc compression"); 667 return -EINVAL; 668 #endif 669 } 670 #endif 671 672 #ifdef CONFIG_F2FS_FS_ZSTD 673 static int f2fs_set_zstd_level(struct f2fs_fs_context *ctx, const char *str) 674 { 675 int level; 676 int len = 4; 677 678 if (strlen(str) == len) { 679 F2FS_CTX_INFO(ctx).compress_level = F2FS_ZSTD_DEFAULT_CLEVEL; 680 ctx->spec_mask |= F2FS_SPEC_compress_level; 681 return 0; 682 } 683 684 str += len; 685 686 if (str[0] != ':') { 687 f2fs_info(NULL, "wrong format, e.g. <alg_name>:<compr_level>"); 688 return -EINVAL; 689 } 690 if (kstrtoint(str + 1, 10, &level)) 691 return -EINVAL; 692 693 /* f2fs does not support negative compress level now */ 694 if (level < 0) { 695 f2fs_info(NULL, "do not support negative compress level: %d", level); 696 return -ERANGE; 697 } 698 699 if (!f2fs_is_compress_level_valid(COMPRESS_ZSTD, level)) { 700 f2fs_info(NULL, "invalid zstd compress level: %d", level); 701 return -EINVAL; 702 } 703 704 F2FS_CTX_INFO(ctx).compress_level = level; 705 ctx->spec_mask |= F2FS_SPEC_compress_level; 706 return 0; 707 } 708 #endif 709 #endif 710 711 static int f2fs_parse_param(struct fs_context *fc, struct fs_parameter *param) 712 { 713 struct f2fs_fs_context *ctx = fc->fs_private; 714 #ifdef CONFIG_F2FS_FS_COMPRESSION 715 unsigned char (*ext)[F2FS_EXTENSION_LEN]; 716 unsigned char (*noext)[F2FS_EXTENSION_LEN]; 717 int ext_cnt, noext_cnt; 718 char *name; 719 #endif 720 substring_t args[MAX_OPT_ARGS]; 721 struct fs_parse_result result; 722 int token, ret, arg; 723 724 token = fs_parse(fc, f2fs_param_specs, param, &result); 725 if (token < 0) 726 return token; 727 728 switch (token) { 729 case Opt_gc_background: 730 F2FS_CTX_INFO(ctx).bggc_mode = result.uint_32; 731 ctx->spec_mask |= F2FS_SPEC_background_gc; 732 break; 733 case Opt_disable_roll_forward: 734 ctx_set_opt(ctx, F2FS_MOUNT_DISABLE_ROLL_FORWARD); 735 break; 736 case Opt_norecovery: 737 /* requires ro mount, checked in f2fs_validate_options */ 738 ctx_set_opt(ctx, F2FS_MOUNT_NORECOVERY); 739 break; 740 case Opt_discard: 741 if (result.negated) 742 ctx_clear_opt(ctx, F2FS_MOUNT_DISCARD); 743 else 744 ctx_set_opt(ctx, F2FS_MOUNT_DISCARD); 745 break; 746 case Opt_noheap: 747 case Opt_heap: 748 f2fs_warn(NULL, "heap/no_heap options were deprecated"); 749 break; 750 #ifdef CONFIG_F2FS_FS_XATTR 751 case Opt_user_xattr: 752 if (result.negated) 753 ctx_clear_opt(ctx, F2FS_MOUNT_XATTR_USER); 754 else 755 ctx_set_opt(ctx, F2FS_MOUNT_XATTR_USER); 756 break; 757 case Opt_inline_xattr: 758 if (result.negated) 759 ctx_clear_opt(ctx, F2FS_MOUNT_INLINE_XATTR); 760 else 761 ctx_set_opt(ctx, F2FS_MOUNT_INLINE_XATTR); 762 break; 763 case Opt_inline_xattr_size: 764 if (result.int_32 < MIN_INLINE_XATTR_SIZE || 765 result.int_32 > MAX_INLINE_XATTR_SIZE) { 766 f2fs_err(NULL, "inline xattr size is out of range: %u ~ %u", 767 (u32)MIN_INLINE_XATTR_SIZE, (u32)MAX_INLINE_XATTR_SIZE); 768 return -EINVAL; 769 } 770 ctx_set_opt(ctx, F2FS_MOUNT_INLINE_XATTR_SIZE); 771 F2FS_CTX_INFO(ctx).inline_xattr_size = result.int_32; 772 ctx->spec_mask |= F2FS_SPEC_inline_xattr_size; 773 break; 774 #else 775 case Opt_user_xattr: 776 case Opt_inline_xattr: 777 case Opt_inline_xattr_size: 778 f2fs_info(NULL, "%s options not supported", param->key); 779 break; 780 #endif 781 #ifdef CONFIG_F2FS_FS_POSIX_ACL 782 case Opt_acl: 783 if (result.negated) 784 ctx_clear_opt(ctx, F2FS_MOUNT_POSIX_ACL); 785 else 786 ctx_set_opt(ctx, F2FS_MOUNT_POSIX_ACL); 787 break; 788 #else 789 case Opt_acl: 790 f2fs_info(NULL, "%s options not supported", param->key); 791 break; 792 #endif 793 case Opt_active_logs: 794 if (result.int_32 != 2 && result.int_32 != 4 && 795 result.int_32 != NR_CURSEG_PERSIST_TYPE) 796 return -EINVAL; 797 ctx->spec_mask |= F2FS_SPEC_active_logs; 798 F2FS_CTX_INFO(ctx).active_logs = result.int_32; 799 break; 800 case Opt_disable_ext_identify: 801 ctx_set_opt(ctx, F2FS_MOUNT_DISABLE_EXT_IDENTIFY); 802 break; 803 case Opt_inline_data: 804 if (result.negated) 805 ctx_clear_opt(ctx, F2FS_MOUNT_INLINE_DATA); 806 else 807 ctx_set_opt(ctx, F2FS_MOUNT_INLINE_DATA); 808 break; 809 case Opt_inline_dentry: 810 if (result.negated) 811 ctx_clear_opt(ctx, F2FS_MOUNT_INLINE_DENTRY); 812 else 813 ctx_set_opt(ctx, F2FS_MOUNT_INLINE_DENTRY); 814 break; 815 case Opt_flush_merge: 816 if (result.negated) 817 ctx_clear_opt(ctx, F2FS_MOUNT_FLUSH_MERGE); 818 else 819 ctx_set_opt(ctx, F2FS_MOUNT_FLUSH_MERGE); 820 break; 821 case Opt_barrier: 822 if (result.negated) 823 ctx_set_opt(ctx, F2FS_MOUNT_NOBARRIER); 824 else 825 ctx_clear_opt(ctx, F2FS_MOUNT_NOBARRIER); 826 break; 827 case Opt_fastboot: 828 ctx_set_opt(ctx, F2FS_MOUNT_FASTBOOT); 829 break; 830 case Opt_extent_cache: 831 if (result.negated) 832 ctx_clear_opt(ctx, F2FS_MOUNT_READ_EXTENT_CACHE); 833 else 834 ctx_set_opt(ctx, F2FS_MOUNT_READ_EXTENT_CACHE); 835 break; 836 case Opt_data_flush: 837 ctx_set_opt(ctx, F2FS_MOUNT_DATA_FLUSH); 838 break; 839 case Opt_reserve_root: 840 ctx_set_opt(ctx, F2FS_MOUNT_RESERVE_ROOT); 841 F2FS_CTX_INFO(ctx).root_reserved_blocks = result.uint_32; 842 ctx->spec_mask |= F2FS_SPEC_reserve_root; 843 break; 844 case Opt_resuid: 845 F2FS_CTX_INFO(ctx).s_resuid = result.uid; 846 ctx->spec_mask |= F2FS_SPEC_resuid; 847 break; 848 case Opt_resgid: 849 F2FS_CTX_INFO(ctx).s_resgid = result.gid; 850 ctx->spec_mask |= F2FS_SPEC_resgid; 851 break; 852 case Opt_mode: 853 F2FS_CTX_INFO(ctx).fs_mode = result.uint_32; 854 ctx->spec_mask |= F2FS_SPEC_mode; 855 break; 856 #ifdef CONFIG_F2FS_FAULT_INJECTION 857 case Opt_fault_injection: 858 F2FS_CTX_INFO(ctx).fault_info.inject_rate = result.int_32; 859 ctx->spec_mask |= F2FS_SPEC_fault_injection; 860 ctx_set_opt(ctx, F2FS_MOUNT_FAULT_INJECTION); 861 break; 862 863 case Opt_fault_type: 864 if (result.uint_32 > BIT(FAULT_MAX)) 865 return -EINVAL; 866 F2FS_CTX_INFO(ctx).fault_info.inject_type = result.uint_32; 867 ctx->spec_mask |= F2FS_SPEC_fault_type; 868 ctx_set_opt(ctx, F2FS_MOUNT_FAULT_INJECTION); 869 break; 870 #else 871 case Opt_fault_injection: 872 case Opt_fault_type: 873 f2fs_info(NULL, "%s options not supported", param->key); 874 break; 875 #endif 876 case Opt_lazytime: 877 if (result.negated) 878 ctx_clear_opt(ctx, F2FS_MOUNT_LAZYTIME); 879 else 880 ctx_set_opt(ctx, F2FS_MOUNT_LAZYTIME); 881 break; 882 #ifdef CONFIG_QUOTA 883 case Opt_quota: 884 if (result.negated) { 885 ctx_clear_opt(ctx, F2FS_MOUNT_QUOTA); 886 ctx_clear_opt(ctx, F2FS_MOUNT_USRQUOTA); 887 ctx_clear_opt(ctx, F2FS_MOUNT_GRPQUOTA); 888 ctx_clear_opt(ctx, F2FS_MOUNT_PRJQUOTA); 889 } else 890 ctx_set_opt(ctx, F2FS_MOUNT_USRQUOTA); 891 break; 892 case Opt_usrquota: 893 ctx_set_opt(ctx, F2FS_MOUNT_USRQUOTA); 894 break; 895 case Opt_grpquota: 896 ctx_set_opt(ctx, F2FS_MOUNT_GRPQUOTA); 897 break; 898 case Opt_prjquota: 899 ctx_set_opt(ctx, F2FS_MOUNT_PRJQUOTA); 900 break; 901 case Opt_usrjquota: 902 if (!*param->string) 903 ret = f2fs_unnote_qf_name(fc, USRQUOTA); 904 else 905 ret = f2fs_note_qf_name(fc, USRQUOTA, param); 906 if (ret) 907 return ret; 908 break; 909 case Opt_grpjquota: 910 if (!*param->string) 911 ret = f2fs_unnote_qf_name(fc, GRPQUOTA); 912 else 913 ret = f2fs_note_qf_name(fc, GRPQUOTA, param); 914 if (ret) 915 return ret; 916 break; 917 case Opt_prjjquota: 918 if (!*param->string) 919 ret = f2fs_unnote_qf_name(fc, PRJQUOTA); 920 else 921 ret = f2fs_note_qf_name(fc, PRJQUOTA, param); 922 if (ret) 923 return ret; 924 break; 925 case Opt_jqfmt: 926 F2FS_CTX_INFO(ctx).s_jquota_fmt = result.int_32; 927 ctx->spec_mask |= F2FS_SPEC_jqfmt; 928 break; 929 #else 930 case Opt_quota: 931 case Opt_usrquota: 932 case Opt_grpquota: 933 case Opt_prjquota: 934 case Opt_usrjquota: 935 case Opt_grpjquota: 936 case Opt_prjjquota: 937 f2fs_info(NULL, "quota operations not supported"); 938 break; 939 #endif 940 case Opt_alloc: 941 F2FS_CTX_INFO(ctx).alloc_mode = result.uint_32; 942 ctx->spec_mask |= F2FS_SPEC_alloc_mode; 943 break; 944 case Opt_fsync: 945 F2FS_CTX_INFO(ctx).fsync_mode = result.uint_32; 946 ctx->spec_mask |= F2FS_SPEC_fsync_mode; 947 break; 948 case Opt_test_dummy_encryption: 949 ret = f2fs_parse_test_dummy_encryption(param, ctx); 950 if (ret) 951 return ret; 952 break; 953 case Opt_inlinecrypt: 954 #ifdef CONFIG_FS_ENCRYPTION_INLINE_CRYPT 955 ctx_set_opt(ctx, F2FS_MOUNT_INLINECRYPT); 956 #else 957 f2fs_info(NULL, "inline encryption not supported"); 958 #endif 959 break; 960 case Opt_checkpoint: 961 /* 962 * Initialize args struct so we know whether arg was 963 * found; some options take optional arguments. 964 */ 965 args[0].from = args[0].to = NULL; 966 arg = 0; 967 968 /* revert to match_table for checkpoint= options */ 969 token = match_token(param->string, f2fs_checkpoint_tokens, args); 970 switch (token) { 971 case Opt_checkpoint_disable_cap_perc: 972 if (args->from && match_int(args, &arg)) 973 return -EINVAL; 974 if (arg < 0 || arg > 100) 975 return -EINVAL; 976 F2FS_CTX_INFO(ctx).unusable_cap_perc = arg; 977 ctx->spec_mask |= F2FS_SPEC_checkpoint_disable_cap_perc; 978 ctx_set_opt(ctx, F2FS_MOUNT_DISABLE_CHECKPOINT); 979 break; 980 case Opt_checkpoint_disable_cap: 981 if (args->from && match_int(args, &arg)) 982 return -EINVAL; 983 F2FS_CTX_INFO(ctx).unusable_cap = arg; 984 ctx->spec_mask |= F2FS_SPEC_checkpoint_disable_cap; 985 ctx_set_opt(ctx, F2FS_MOUNT_DISABLE_CHECKPOINT); 986 break; 987 case Opt_checkpoint_disable: 988 ctx_set_opt(ctx, F2FS_MOUNT_DISABLE_CHECKPOINT); 989 break; 990 case Opt_checkpoint_enable: 991 ctx_clear_opt(ctx, F2FS_MOUNT_DISABLE_CHECKPOINT); 992 break; 993 default: 994 return -EINVAL; 995 } 996 break; 997 case Opt_checkpoint_merge: 998 if (result.negated) 999 ctx_clear_opt(ctx, F2FS_MOUNT_MERGE_CHECKPOINT); 1000 else 1001 ctx_set_opt(ctx, F2FS_MOUNT_MERGE_CHECKPOINT); 1002 break; 1003 #ifdef CONFIG_F2FS_FS_COMPRESSION 1004 case Opt_compress_algorithm: 1005 name = param->string; 1006 if (!strcmp(name, "lzo")) { 1007 #ifdef CONFIG_F2FS_FS_LZO 1008 F2FS_CTX_INFO(ctx).compress_level = 0; 1009 F2FS_CTX_INFO(ctx).compress_algorithm = COMPRESS_LZO; 1010 ctx->spec_mask |= F2FS_SPEC_compress_level; 1011 ctx->spec_mask |= F2FS_SPEC_compress_algorithm; 1012 #else 1013 f2fs_info(NULL, "kernel doesn't support lzo compression"); 1014 #endif 1015 } else if (!strncmp(name, "lz4", 3)) { 1016 #ifdef CONFIG_F2FS_FS_LZ4 1017 ret = f2fs_set_lz4hc_level(ctx, name); 1018 if (ret) 1019 return -EINVAL; 1020 F2FS_CTX_INFO(ctx).compress_algorithm = COMPRESS_LZ4; 1021 ctx->spec_mask |= F2FS_SPEC_compress_algorithm; 1022 #else 1023 f2fs_info(NULL, "kernel doesn't support lz4 compression"); 1024 #endif 1025 } else if (!strncmp(name, "zstd", 4)) { 1026 #ifdef CONFIG_F2FS_FS_ZSTD 1027 ret = f2fs_set_zstd_level(ctx, name); 1028 if (ret) 1029 return -EINVAL; 1030 F2FS_CTX_INFO(ctx).compress_algorithm = COMPRESS_ZSTD; 1031 ctx->spec_mask |= F2FS_SPEC_compress_algorithm; 1032 #else 1033 f2fs_info(NULL, "kernel doesn't support zstd compression"); 1034 #endif 1035 } else if (!strcmp(name, "lzo-rle")) { 1036 #ifdef CONFIG_F2FS_FS_LZORLE 1037 F2FS_CTX_INFO(ctx).compress_level = 0; 1038 F2FS_CTX_INFO(ctx).compress_algorithm = COMPRESS_LZORLE; 1039 ctx->spec_mask |= F2FS_SPEC_compress_level; 1040 ctx->spec_mask |= F2FS_SPEC_compress_algorithm; 1041 #else 1042 f2fs_info(NULL, "kernel doesn't support lzorle compression"); 1043 #endif 1044 } else 1045 return -EINVAL; 1046 break; 1047 case Opt_compress_log_size: 1048 if (result.uint_32 < MIN_COMPRESS_LOG_SIZE || 1049 result.uint_32 > MAX_COMPRESS_LOG_SIZE) { 1050 f2fs_err(NULL, 1051 "Compress cluster log size is out of range"); 1052 return -EINVAL; 1053 } 1054 F2FS_CTX_INFO(ctx).compress_log_size = result.uint_32; 1055 ctx->spec_mask |= F2FS_SPEC_compress_log_size; 1056 break; 1057 case Opt_compress_extension: 1058 name = param->string; 1059 ext = F2FS_CTX_INFO(ctx).extensions; 1060 ext_cnt = F2FS_CTX_INFO(ctx).compress_ext_cnt; 1061 1062 if (strlen(name) >= F2FS_EXTENSION_LEN || 1063 ext_cnt >= COMPRESS_EXT_NUM) { 1064 f2fs_err(NULL, "invalid extension length/number"); 1065 return -EINVAL; 1066 } 1067 1068 if (is_compress_extension_exist(&ctx->info, name, true)) 1069 break; 1070 1071 ret = strscpy(ext[ext_cnt], name, F2FS_EXTENSION_LEN); 1072 if (ret < 0) 1073 return ret; 1074 F2FS_CTX_INFO(ctx).compress_ext_cnt++; 1075 ctx->spec_mask |= F2FS_SPEC_compress_extension; 1076 break; 1077 case Opt_nocompress_extension: 1078 name = param->string; 1079 noext = F2FS_CTX_INFO(ctx).noextensions; 1080 noext_cnt = F2FS_CTX_INFO(ctx).nocompress_ext_cnt; 1081 1082 if (strlen(name) >= F2FS_EXTENSION_LEN || 1083 noext_cnt >= COMPRESS_EXT_NUM) { 1084 f2fs_err(NULL, "invalid extension length/number"); 1085 return -EINVAL; 1086 } 1087 1088 if (is_compress_extension_exist(&ctx->info, name, false)) 1089 break; 1090 1091 ret = strscpy(noext[noext_cnt], name, F2FS_EXTENSION_LEN); 1092 if (ret < 0) 1093 return ret; 1094 F2FS_CTX_INFO(ctx).nocompress_ext_cnt++; 1095 ctx->spec_mask |= F2FS_SPEC_nocompress_extension; 1096 break; 1097 case Opt_compress_chksum: 1098 F2FS_CTX_INFO(ctx).compress_chksum = true; 1099 ctx->spec_mask |= F2FS_SPEC_compress_chksum; 1100 break; 1101 case Opt_compress_mode: 1102 F2FS_CTX_INFO(ctx).compress_mode = result.uint_32; 1103 ctx->spec_mask |= F2FS_SPEC_compress_mode; 1104 break; 1105 case Opt_compress_cache: 1106 ctx_set_opt(ctx, F2FS_MOUNT_COMPRESS_CACHE); 1107 break; 1108 #else 1109 case Opt_compress_algorithm: 1110 case Opt_compress_log_size: 1111 case Opt_compress_extension: 1112 case Opt_nocompress_extension: 1113 case Opt_compress_chksum: 1114 case Opt_compress_mode: 1115 case Opt_compress_cache: 1116 f2fs_info(NULL, "compression options not supported"); 1117 break; 1118 #endif 1119 case Opt_atgc: 1120 ctx_set_opt(ctx, F2FS_MOUNT_ATGC); 1121 break; 1122 case Opt_gc_merge: 1123 if (result.negated) 1124 ctx_clear_opt(ctx, F2FS_MOUNT_GC_MERGE); 1125 else 1126 ctx_set_opt(ctx, F2FS_MOUNT_GC_MERGE); 1127 break; 1128 case Opt_discard_unit: 1129 F2FS_CTX_INFO(ctx).discard_unit = result.uint_32; 1130 ctx->spec_mask |= F2FS_SPEC_discard_unit; 1131 break; 1132 case Opt_memory_mode: 1133 F2FS_CTX_INFO(ctx).memory_mode = result.uint_32; 1134 ctx->spec_mask |= F2FS_SPEC_memory_mode; 1135 break; 1136 case Opt_age_extent_cache: 1137 ctx_set_opt(ctx, F2FS_MOUNT_AGE_EXTENT_CACHE); 1138 break; 1139 case Opt_errors: 1140 F2FS_CTX_INFO(ctx).errors = result.uint_32; 1141 ctx->spec_mask |= F2FS_SPEC_errors; 1142 break; 1143 case Opt_nat_bits: 1144 ctx_set_opt(ctx, F2FS_MOUNT_NAT_BITS); 1145 break; 1146 } 1147 return 0; 1148 } 1149 1150 /* 1151 * Check quota settings consistency. 1152 */ 1153 static int f2fs_check_quota_consistency(struct fs_context *fc, 1154 struct super_block *sb) 1155 { 1156 struct f2fs_sb_info *sbi = F2FS_SB(sb); 1157 #ifdef CONFIG_QUOTA 1158 struct f2fs_fs_context *ctx = fc->fs_private; 1159 bool quota_feature = f2fs_sb_has_quota_ino(sbi); 1160 bool quota_turnon = sb_any_quota_loaded(sb); 1161 char *old_qname, *new_qname; 1162 bool usr_qf_name, grp_qf_name, prj_qf_name, usrquota, grpquota, prjquota; 1163 int i; 1164 1165 /* 1166 * We do the test below only for project quotas. 'usrquota' and 1167 * 'grpquota' mount options are allowed even without quota feature 1168 * to support legacy quotas in quota files. 1169 */ 1170 if (ctx_test_opt(ctx, F2FS_MOUNT_PRJQUOTA) && 1171 !f2fs_sb_has_project_quota(sbi)) { 1172 f2fs_err(sbi, "Project quota feature not enabled. Cannot enable project quota enforcement."); 1173 return -EINVAL; 1174 } 1175 1176 if (ctx->qname_mask) { 1177 for (i = 0; i < MAXQUOTAS; i++) { 1178 if (!(ctx->qname_mask & (1 << i))) 1179 continue; 1180 1181 old_qname = F2FS_OPTION(sbi).s_qf_names[i]; 1182 new_qname = F2FS_CTX_INFO(ctx).s_qf_names[i]; 1183 if (quota_turnon && 1184 !!old_qname != !!new_qname) 1185 goto err_jquota_change; 1186 1187 if (old_qname) { 1188 if (strcmp(old_qname, new_qname) == 0) { 1189 ctx->qname_mask &= ~(1 << i); 1190 continue; 1191 } 1192 goto err_jquota_specified; 1193 } 1194 1195 if (quota_feature) { 1196 f2fs_info(sbi, "QUOTA feature is enabled, so ignore qf_name"); 1197 ctx->qname_mask &= ~(1 << i); 1198 kfree(F2FS_CTX_INFO(ctx).s_qf_names[i]); 1199 F2FS_CTX_INFO(ctx).s_qf_names[i] = NULL; 1200 } 1201 } 1202 } 1203 1204 /* Make sure we don't mix old and new quota format */ 1205 usr_qf_name = F2FS_OPTION(sbi).s_qf_names[USRQUOTA] || 1206 F2FS_CTX_INFO(ctx).s_qf_names[USRQUOTA]; 1207 grp_qf_name = F2FS_OPTION(sbi).s_qf_names[GRPQUOTA] || 1208 F2FS_CTX_INFO(ctx).s_qf_names[GRPQUOTA]; 1209 prj_qf_name = F2FS_OPTION(sbi).s_qf_names[PRJQUOTA] || 1210 F2FS_CTX_INFO(ctx).s_qf_names[PRJQUOTA]; 1211 usrquota = test_opt(sbi, USRQUOTA) || 1212 ctx_test_opt(ctx, F2FS_MOUNT_USRQUOTA); 1213 grpquota = test_opt(sbi, GRPQUOTA) || 1214 ctx_test_opt(ctx, F2FS_MOUNT_GRPQUOTA); 1215 prjquota = test_opt(sbi, PRJQUOTA) || 1216 ctx_test_opt(ctx, F2FS_MOUNT_PRJQUOTA); 1217 1218 if (usr_qf_name) { 1219 ctx_clear_opt(ctx, F2FS_MOUNT_USRQUOTA); 1220 usrquota = false; 1221 } 1222 if (grp_qf_name) { 1223 ctx_clear_opt(ctx, F2FS_MOUNT_GRPQUOTA); 1224 grpquota = false; 1225 } 1226 if (prj_qf_name) { 1227 ctx_clear_opt(ctx, F2FS_MOUNT_PRJQUOTA); 1228 prjquota = false; 1229 } 1230 if (usr_qf_name || grp_qf_name || prj_qf_name) { 1231 if (grpquota || usrquota || prjquota) { 1232 f2fs_err(sbi, "old and new quota format mixing"); 1233 return -EINVAL; 1234 } 1235 if (!(ctx->spec_mask & F2FS_SPEC_jqfmt || 1236 F2FS_OPTION(sbi).s_jquota_fmt)) { 1237 f2fs_err(sbi, "journaled quota format not specified"); 1238 return -EINVAL; 1239 } 1240 } 1241 return 0; 1242 1243 err_jquota_change: 1244 f2fs_err(sbi, "Cannot change journaled quota options when quota turned on"); 1245 return -EINVAL; 1246 err_jquota_specified: 1247 f2fs_err(sbi, "%s quota file already specified", 1248 QTYPE2NAME(i)); 1249 return -EINVAL; 1250 1251 #else 1252 if (f2fs_readonly(sbi->sb)) 1253 return 0; 1254 if (f2fs_sb_has_quota_ino(sbi)) { 1255 f2fs_info(sbi, "Filesystem with quota feature cannot be mounted RDWR without CONFIG_QUOTA"); 1256 return -EINVAL; 1257 } 1258 if (f2fs_sb_has_project_quota(sbi)) { 1259 f2fs_err(sbi, "Filesystem with project quota feature cannot be mounted RDWR without CONFIG_QUOTA"); 1260 return -EINVAL; 1261 } 1262 1263 return 0; 1264 #endif 1265 } 1266 1267 static int f2fs_check_test_dummy_encryption(struct fs_context *fc, 1268 struct super_block *sb) 1269 { 1270 struct f2fs_fs_context *ctx = fc->fs_private; 1271 struct f2fs_sb_info *sbi = F2FS_SB(sb); 1272 1273 if (!fscrypt_is_dummy_policy_set(&F2FS_CTX_INFO(ctx).dummy_enc_policy)) 1274 return 0; 1275 1276 if (!f2fs_sb_has_encrypt(sbi)) { 1277 f2fs_err(sbi, "Encrypt feature is off"); 1278 return -EINVAL; 1279 } 1280 1281 /* 1282 * This mount option is just for testing, and it's not worthwhile to 1283 * implement the extra complexity (e.g. RCU protection) that would be 1284 * needed to allow it to be set or changed during remount. We do allow 1285 * it to be specified during remount, but only if there is no change. 1286 */ 1287 if (fc->purpose == FS_CONTEXT_FOR_RECONFIGURE) { 1288 if (fscrypt_dummy_policies_equal(&F2FS_OPTION(sbi).dummy_enc_policy, 1289 &F2FS_CTX_INFO(ctx).dummy_enc_policy)) 1290 return 0; 1291 f2fs_warn(sbi, "Can't set or change test_dummy_encryption on remount"); 1292 return -EINVAL; 1293 } 1294 return 0; 1295 } 1296 1297 static inline bool test_compression_spec(unsigned int mask) 1298 { 1299 return mask & (F2FS_SPEC_compress_algorithm 1300 | F2FS_SPEC_compress_log_size 1301 | F2FS_SPEC_compress_extension 1302 | F2FS_SPEC_nocompress_extension 1303 | F2FS_SPEC_compress_chksum 1304 | F2FS_SPEC_compress_mode); 1305 } 1306 1307 static inline void clear_compression_spec(struct f2fs_fs_context *ctx) 1308 { 1309 ctx->spec_mask &= ~(F2FS_SPEC_compress_algorithm 1310 | F2FS_SPEC_compress_log_size 1311 | F2FS_SPEC_compress_extension 1312 | F2FS_SPEC_nocompress_extension 1313 | F2FS_SPEC_compress_chksum 1314 | F2FS_SPEC_compress_mode); 1315 } 1316 1317 static int f2fs_check_compression(struct fs_context *fc, 1318 struct super_block *sb) 1319 { 1320 #ifdef CONFIG_F2FS_FS_COMPRESSION 1321 struct f2fs_fs_context *ctx = fc->fs_private; 1322 struct f2fs_sb_info *sbi = F2FS_SB(sb); 1323 int i, cnt; 1324 1325 if (!f2fs_sb_has_compression(sbi)) { 1326 if (test_compression_spec(ctx->spec_mask) || 1327 ctx_test_opt(ctx, F2FS_MOUNT_COMPRESS_CACHE)) 1328 f2fs_info(sbi, "Image doesn't support compression"); 1329 clear_compression_spec(ctx); 1330 ctx->opt_mask &= ~F2FS_MOUNT_COMPRESS_CACHE; 1331 return 0; 1332 } 1333 if (ctx->spec_mask & F2FS_SPEC_compress_extension) { 1334 cnt = F2FS_CTX_INFO(ctx).compress_ext_cnt; 1335 for (i = 0; i < F2FS_CTX_INFO(ctx).compress_ext_cnt; i++) { 1336 if (is_compress_extension_exist(&F2FS_OPTION(sbi), 1337 F2FS_CTX_INFO(ctx).extensions[i], true)) { 1338 F2FS_CTX_INFO(ctx).extensions[i][0] = '\0'; 1339 cnt--; 1340 } 1341 } 1342 if (F2FS_OPTION(sbi).compress_ext_cnt + cnt > COMPRESS_EXT_NUM) { 1343 f2fs_err(sbi, "invalid extension length/number"); 1344 return -EINVAL; 1345 } 1346 } 1347 if (ctx->spec_mask & F2FS_SPEC_nocompress_extension) { 1348 cnt = F2FS_CTX_INFO(ctx).nocompress_ext_cnt; 1349 for (i = 0; i < F2FS_CTX_INFO(ctx).nocompress_ext_cnt; i++) { 1350 if (is_compress_extension_exist(&F2FS_OPTION(sbi), 1351 F2FS_CTX_INFO(ctx).noextensions[i], false)) { 1352 F2FS_CTX_INFO(ctx).noextensions[i][0] = '\0'; 1353 cnt--; 1354 } 1355 } 1356 if (F2FS_OPTION(sbi).nocompress_ext_cnt + cnt > COMPRESS_EXT_NUM) { 1357 f2fs_err(sbi, "invalid noextension length/number"); 1358 return -EINVAL; 1359 } 1360 } 1361 1362 if (f2fs_test_compress_extension(F2FS_CTX_INFO(ctx).noextensions, 1363 F2FS_CTX_INFO(ctx).nocompress_ext_cnt, 1364 F2FS_CTX_INFO(ctx).extensions, 1365 F2FS_CTX_INFO(ctx).compress_ext_cnt)) { 1366 f2fs_err(sbi, "new noextensions conflicts with new extensions"); 1367 return -EINVAL; 1368 } 1369 if (f2fs_test_compress_extension(F2FS_CTX_INFO(ctx).noextensions, 1370 F2FS_CTX_INFO(ctx).nocompress_ext_cnt, 1371 F2FS_OPTION(sbi).extensions, 1372 F2FS_OPTION(sbi).compress_ext_cnt)) { 1373 f2fs_err(sbi, "new noextensions conflicts with old extensions"); 1374 return -EINVAL; 1375 } 1376 if (f2fs_test_compress_extension(F2FS_OPTION(sbi).noextensions, 1377 F2FS_OPTION(sbi).nocompress_ext_cnt, 1378 F2FS_CTX_INFO(ctx).extensions, 1379 F2FS_CTX_INFO(ctx).compress_ext_cnt)) { 1380 f2fs_err(sbi, "new extensions conflicts with old noextensions"); 1381 return -EINVAL; 1382 } 1383 #endif 1384 return 0; 1385 } 1386 1387 static int f2fs_check_opt_consistency(struct fs_context *fc, 1388 struct super_block *sb) 1389 { 1390 struct f2fs_fs_context *ctx = fc->fs_private; 1391 struct f2fs_sb_info *sbi = F2FS_SB(sb); 1392 int err; 1393 1394 if (ctx_test_opt(ctx, F2FS_MOUNT_NORECOVERY) && !f2fs_readonly(sb)) 1395 return -EINVAL; 1396 1397 if (f2fs_hw_should_discard(sbi) && 1398 (ctx->opt_mask & F2FS_MOUNT_DISCARD) && 1399 !ctx_test_opt(ctx, F2FS_MOUNT_DISCARD)) { 1400 f2fs_warn(sbi, "discard is required for zoned block devices"); 1401 return -EINVAL; 1402 } 1403 1404 if (!f2fs_hw_support_discard(sbi) && 1405 (ctx->opt_mask & F2FS_MOUNT_DISCARD) && 1406 ctx_test_opt(ctx, F2FS_MOUNT_DISCARD)) { 1407 f2fs_warn(sbi, "device does not support discard"); 1408 ctx_clear_opt(ctx, F2FS_MOUNT_DISCARD); 1409 ctx->opt_mask &= ~F2FS_MOUNT_DISCARD; 1410 } 1411 1412 if (f2fs_sb_has_device_alias(sbi) && 1413 (ctx->opt_mask & F2FS_MOUNT_READ_EXTENT_CACHE) && 1414 !ctx_test_opt(ctx, F2FS_MOUNT_READ_EXTENT_CACHE)) { 1415 f2fs_err(sbi, "device aliasing requires extent cache"); 1416 return -EINVAL; 1417 } 1418 1419 if (test_opt(sbi, RESERVE_ROOT) && 1420 (ctx->opt_mask & F2FS_MOUNT_RESERVE_ROOT) && 1421 ctx_test_opt(ctx, F2FS_MOUNT_RESERVE_ROOT)) { 1422 f2fs_info(sbi, "Preserve previous reserve_root=%u", 1423 F2FS_OPTION(sbi).root_reserved_blocks); 1424 ctx_clear_opt(ctx, F2FS_MOUNT_RESERVE_ROOT); 1425 ctx->opt_mask &= ~F2FS_MOUNT_RESERVE_ROOT; 1426 } 1427 1428 err = f2fs_check_test_dummy_encryption(fc, sb); 1429 if (err) 1430 return err; 1431 1432 err = f2fs_check_compression(fc, sb); 1433 if (err) 1434 return err; 1435 1436 err = f2fs_check_quota_consistency(fc, sb); 1437 if (err) 1438 return err; 1439 1440 if (!IS_ENABLED(CONFIG_UNICODE) && f2fs_sb_has_casefold(sbi)) { 1441 f2fs_err(sbi, 1442 "Filesystem with casefold feature cannot be mounted without CONFIG_UNICODE"); 1443 return -EINVAL; 1444 } 1445 1446 /* 1447 * The BLKZONED feature indicates that the drive was formatted with 1448 * zone alignment optimization. This is optional for host-aware 1449 * devices, but mandatory for host-managed zoned block devices. 1450 */ 1451 if (f2fs_sb_has_blkzoned(sbi)) { 1452 if (F2FS_CTX_INFO(ctx).bggc_mode == BGGC_MODE_OFF) { 1453 f2fs_warn(sbi, "zoned devices need bggc"); 1454 return -EINVAL; 1455 } 1456 #ifdef CONFIG_BLK_DEV_ZONED 1457 if ((ctx->spec_mask & F2FS_SPEC_discard_unit) && 1458 F2FS_CTX_INFO(ctx).discard_unit != DISCARD_UNIT_SECTION) { 1459 f2fs_info(sbi, "Zoned block device doesn't need small discard, set discard_unit=section by default"); 1460 F2FS_CTX_INFO(ctx).discard_unit = DISCARD_UNIT_SECTION; 1461 } 1462 1463 if ((ctx->spec_mask & F2FS_SPEC_mode) && 1464 F2FS_CTX_INFO(ctx).fs_mode != FS_MODE_LFS) { 1465 f2fs_info(sbi, "Only lfs mode is allowed with zoned block device feature"); 1466 return -EINVAL; 1467 } 1468 #else 1469 f2fs_err(sbi, "Zoned block device support is not enabled"); 1470 return -EINVAL; 1471 #endif 1472 } 1473 1474 if (ctx_test_opt(ctx, F2FS_MOUNT_INLINE_XATTR_SIZE)) { 1475 if (!f2fs_sb_has_extra_attr(sbi) || 1476 !f2fs_sb_has_flexible_inline_xattr(sbi)) { 1477 f2fs_err(sbi, "extra_attr or flexible_inline_xattr feature is off"); 1478 return -EINVAL; 1479 } 1480 if (!ctx_test_opt(ctx, F2FS_MOUNT_INLINE_XATTR) && !test_opt(sbi, INLINE_XATTR)) { 1481 f2fs_err(sbi, "inline_xattr_size option should be set with inline_xattr option"); 1482 return -EINVAL; 1483 } 1484 } 1485 1486 if (ctx_test_opt(ctx, F2FS_MOUNT_ATGC) && 1487 F2FS_CTX_INFO(ctx).fs_mode == FS_MODE_LFS) { 1488 f2fs_err(sbi, "LFS is not compatible with ATGC"); 1489 return -EINVAL; 1490 } 1491 1492 if (f2fs_is_readonly(sbi) && ctx_test_opt(ctx, F2FS_MOUNT_FLUSH_MERGE)) { 1493 f2fs_err(sbi, "FLUSH_MERGE not compatible with readonly mode"); 1494 return -EINVAL; 1495 } 1496 1497 if (f2fs_sb_has_readonly(sbi) && !f2fs_readonly(sbi->sb)) { 1498 f2fs_err(sbi, "Allow to mount readonly mode only"); 1499 return -EROFS; 1500 } 1501 return 0; 1502 } 1503 1504 static void f2fs_apply_quota_options(struct fs_context *fc, 1505 struct super_block *sb) 1506 { 1507 #ifdef CONFIG_QUOTA 1508 struct f2fs_fs_context *ctx = fc->fs_private; 1509 struct f2fs_sb_info *sbi = F2FS_SB(sb); 1510 bool quota_feature = f2fs_sb_has_quota_ino(sbi); 1511 char *qname; 1512 int i; 1513 1514 if (quota_feature) 1515 return; 1516 1517 for (i = 0; i < MAXQUOTAS; i++) { 1518 if (!(ctx->qname_mask & (1 << i))) 1519 continue; 1520 1521 qname = F2FS_CTX_INFO(ctx).s_qf_names[i]; 1522 if (qname) { 1523 qname = kstrdup(F2FS_CTX_INFO(ctx).s_qf_names[i], 1524 GFP_KERNEL | __GFP_NOFAIL); 1525 set_opt(sbi, QUOTA); 1526 } 1527 F2FS_OPTION(sbi).s_qf_names[i] = qname; 1528 } 1529 1530 if (ctx->spec_mask & F2FS_SPEC_jqfmt) 1531 F2FS_OPTION(sbi).s_jquota_fmt = F2FS_CTX_INFO(ctx).s_jquota_fmt; 1532 1533 if (quota_feature && F2FS_OPTION(sbi).s_jquota_fmt) { 1534 f2fs_info(sbi, "QUOTA feature is enabled, so ignore jquota_fmt"); 1535 F2FS_OPTION(sbi).s_jquota_fmt = 0; 1536 } 1537 #endif 1538 } 1539 1540 static void f2fs_apply_test_dummy_encryption(struct fs_context *fc, 1541 struct super_block *sb) 1542 { 1543 struct f2fs_fs_context *ctx = fc->fs_private; 1544 struct f2fs_sb_info *sbi = F2FS_SB(sb); 1545 1546 if (!fscrypt_is_dummy_policy_set(&F2FS_CTX_INFO(ctx).dummy_enc_policy) || 1547 /* if already set, it was already verified to be the same */ 1548 fscrypt_is_dummy_policy_set(&F2FS_OPTION(sbi).dummy_enc_policy)) 1549 return; 1550 swap(F2FS_OPTION(sbi).dummy_enc_policy, F2FS_CTX_INFO(ctx).dummy_enc_policy); 1551 f2fs_warn(sbi, "Test dummy encryption mode enabled"); 1552 } 1553 1554 static void f2fs_apply_compression(struct fs_context *fc, 1555 struct super_block *sb) 1556 { 1557 #ifdef CONFIG_F2FS_FS_COMPRESSION 1558 struct f2fs_fs_context *ctx = fc->fs_private; 1559 struct f2fs_sb_info *sbi = F2FS_SB(sb); 1560 unsigned char (*ctx_ext)[F2FS_EXTENSION_LEN]; 1561 unsigned char (*sbi_ext)[F2FS_EXTENSION_LEN]; 1562 int ctx_cnt, sbi_cnt, i; 1563 1564 if (ctx->spec_mask & F2FS_SPEC_compress_level) 1565 F2FS_OPTION(sbi).compress_level = 1566 F2FS_CTX_INFO(ctx).compress_level; 1567 if (ctx->spec_mask & F2FS_SPEC_compress_algorithm) 1568 F2FS_OPTION(sbi).compress_algorithm = 1569 F2FS_CTX_INFO(ctx).compress_algorithm; 1570 if (ctx->spec_mask & F2FS_SPEC_compress_log_size) 1571 F2FS_OPTION(sbi).compress_log_size = 1572 F2FS_CTX_INFO(ctx).compress_log_size; 1573 if (ctx->spec_mask & F2FS_SPEC_compress_chksum) 1574 F2FS_OPTION(sbi).compress_chksum = 1575 F2FS_CTX_INFO(ctx).compress_chksum; 1576 if (ctx->spec_mask & F2FS_SPEC_compress_mode) 1577 F2FS_OPTION(sbi).compress_mode = 1578 F2FS_CTX_INFO(ctx).compress_mode; 1579 if (ctx->spec_mask & F2FS_SPEC_compress_extension) { 1580 ctx_ext = F2FS_CTX_INFO(ctx).extensions; 1581 ctx_cnt = F2FS_CTX_INFO(ctx).compress_ext_cnt; 1582 sbi_ext = F2FS_OPTION(sbi).extensions; 1583 sbi_cnt = F2FS_OPTION(sbi).compress_ext_cnt; 1584 for (i = 0; i < ctx_cnt; i++) { 1585 if (strlen(ctx_ext[i]) == 0) 1586 continue; 1587 strscpy(sbi_ext[sbi_cnt], ctx_ext[i]); 1588 sbi_cnt++; 1589 } 1590 F2FS_OPTION(sbi).compress_ext_cnt = sbi_cnt; 1591 } 1592 if (ctx->spec_mask & F2FS_SPEC_nocompress_extension) { 1593 ctx_ext = F2FS_CTX_INFO(ctx).noextensions; 1594 ctx_cnt = F2FS_CTX_INFO(ctx).nocompress_ext_cnt; 1595 sbi_ext = F2FS_OPTION(sbi).noextensions; 1596 sbi_cnt = F2FS_OPTION(sbi).nocompress_ext_cnt; 1597 for (i = 0; i < ctx_cnt; i++) { 1598 if (strlen(ctx_ext[i]) == 0) 1599 continue; 1600 strscpy(sbi_ext[sbi_cnt], ctx_ext[i]); 1601 sbi_cnt++; 1602 } 1603 F2FS_OPTION(sbi).nocompress_ext_cnt = sbi_cnt; 1604 } 1605 #endif 1606 } 1607 1608 static void f2fs_apply_options(struct fs_context *fc, struct super_block *sb) 1609 { 1610 struct f2fs_fs_context *ctx = fc->fs_private; 1611 struct f2fs_sb_info *sbi = F2FS_SB(sb); 1612 1613 F2FS_OPTION(sbi).opt &= ~ctx->opt_mask; 1614 F2FS_OPTION(sbi).opt |= F2FS_CTX_INFO(ctx).opt; 1615 1616 if (ctx->spec_mask & F2FS_SPEC_background_gc) 1617 F2FS_OPTION(sbi).bggc_mode = F2FS_CTX_INFO(ctx).bggc_mode; 1618 if (ctx->spec_mask & F2FS_SPEC_inline_xattr_size) 1619 F2FS_OPTION(sbi).inline_xattr_size = 1620 F2FS_CTX_INFO(ctx).inline_xattr_size; 1621 if (ctx->spec_mask & F2FS_SPEC_active_logs) 1622 F2FS_OPTION(sbi).active_logs = F2FS_CTX_INFO(ctx).active_logs; 1623 if (ctx->spec_mask & F2FS_SPEC_reserve_root) 1624 F2FS_OPTION(sbi).root_reserved_blocks = 1625 F2FS_CTX_INFO(ctx).root_reserved_blocks; 1626 if (ctx->spec_mask & F2FS_SPEC_resgid) 1627 F2FS_OPTION(sbi).s_resgid = F2FS_CTX_INFO(ctx).s_resgid; 1628 if (ctx->spec_mask & F2FS_SPEC_resuid) 1629 F2FS_OPTION(sbi).s_resuid = F2FS_CTX_INFO(ctx).s_resuid; 1630 if (ctx->spec_mask & F2FS_SPEC_mode) 1631 F2FS_OPTION(sbi).fs_mode = F2FS_CTX_INFO(ctx).fs_mode; 1632 #ifdef CONFIG_F2FS_FAULT_INJECTION 1633 if (ctx->spec_mask & F2FS_SPEC_fault_injection) 1634 (void)f2fs_build_fault_attr(sbi, 1635 F2FS_CTX_INFO(ctx).fault_info.inject_rate, 0, FAULT_RATE); 1636 if (ctx->spec_mask & F2FS_SPEC_fault_type) 1637 (void)f2fs_build_fault_attr(sbi, 0, 1638 F2FS_CTX_INFO(ctx).fault_info.inject_type, FAULT_TYPE); 1639 #endif 1640 if (ctx->spec_mask & F2FS_SPEC_alloc_mode) 1641 F2FS_OPTION(sbi).alloc_mode = F2FS_CTX_INFO(ctx).alloc_mode; 1642 if (ctx->spec_mask & F2FS_SPEC_fsync_mode) 1643 F2FS_OPTION(sbi).fsync_mode = F2FS_CTX_INFO(ctx).fsync_mode; 1644 if (ctx->spec_mask & F2FS_SPEC_checkpoint_disable_cap) 1645 F2FS_OPTION(sbi).unusable_cap = F2FS_CTX_INFO(ctx).unusable_cap; 1646 if (ctx->spec_mask & F2FS_SPEC_checkpoint_disable_cap_perc) 1647 F2FS_OPTION(sbi).unusable_cap_perc = 1648 F2FS_CTX_INFO(ctx).unusable_cap_perc; 1649 if (ctx->spec_mask & F2FS_SPEC_discard_unit) 1650 F2FS_OPTION(sbi).discard_unit = F2FS_CTX_INFO(ctx).discard_unit; 1651 if (ctx->spec_mask & F2FS_SPEC_memory_mode) 1652 F2FS_OPTION(sbi).memory_mode = F2FS_CTX_INFO(ctx).memory_mode; 1653 if (ctx->spec_mask & F2FS_SPEC_errors) 1654 F2FS_OPTION(sbi).errors = F2FS_CTX_INFO(ctx).errors; 1655 1656 f2fs_apply_compression(fc, sb); 1657 f2fs_apply_test_dummy_encryption(fc, sb); 1658 f2fs_apply_quota_options(fc, sb); 1659 } 1660 1661 static int f2fs_sanity_check_options(struct f2fs_sb_info *sbi, bool remount) 1662 { 1663 if (f2fs_sb_has_device_alias(sbi) && 1664 !test_opt(sbi, READ_EXTENT_CACHE)) { 1665 f2fs_err(sbi, "device aliasing requires extent cache"); 1666 return -EINVAL; 1667 } 1668 1669 if (!remount) 1670 return 0; 1671 1672 #ifdef CONFIG_BLK_DEV_ZONED 1673 if (f2fs_sb_has_blkzoned(sbi) && 1674 sbi->max_open_zones < F2FS_OPTION(sbi).active_logs) { 1675 f2fs_err(sbi, 1676 "zoned: max open zones %u is too small, need at least %u open zones", 1677 sbi->max_open_zones, F2FS_OPTION(sbi).active_logs); 1678 return -EINVAL; 1679 } 1680 #endif 1681 if (f2fs_lfs_mode(sbi) && !IS_F2FS_IPU_DISABLE(sbi)) { 1682 f2fs_warn(sbi, "LFS is not compatible with IPU"); 1683 return -EINVAL; 1684 } 1685 return 0; 1686 } 1687 1688 static struct inode *f2fs_alloc_inode(struct super_block *sb) 1689 { 1690 struct f2fs_inode_info *fi; 1691 1692 if (time_to_inject(F2FS_SB(sb), FAULT_SLAB_ALLOC)) 1693 return NULL; 1694 1695 fi = alloc_inode_sb(sb, f2fs_inode_cachep, GFP_F2FS_ZERO); 1696 if (!fi) 1697 return NULL; 1698 1699 init_once((void *) fi); 1700 1701 /* Initialize f2fs-specific inode info */ 1702 atomic_set(&fi->dirty_pages, 0); 1703 atomic_set(&fi->i_compr_blocks, 0); 1704 atomic_set(&fi->open_count, 0); 1705 init_f2fs_rwsem(&fi->i_sem); 1706 spin_lock_init(&fi->i_size_lock); 1707 INIT_LIST_HEAD(&fi->dirty_list); 1708 INIT_LIST_HEAD(&fi->gdirty_list); 1709 INIT_LIST_HEAD(&fi->gdonate_list); 1710 init_f2fs_rwsem(&fi->i_gc_rwsem[READ]); 1711 init_f2fs_rwsem(&fi->i_gc_rwsem[WRITE]); 1712 init_f2fs_rwsem(&fi->i_xattr_sem); 1713 1714 /* Will be used by directory only */ 1715 fi->i_dir_level = F2FS_SB(sb)->dir_level; 1716 1717 return &fi->vfs_inode; 1718 } 1719 1720 static int f2fs_drop_inode(struct inode *inode) 1721 { 1722 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1723 int ret; 1724 1725 /* 1726 * during filesystem shutdown, if checkpoint is disabled, 1727 * drop useless meta/node dirty pages. 1728 */ 1729 if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED))) { 1730 if (inode->i_ino == F2FS_NODE_INO(sbi) || 1731 inode->i_ino == F2FS_META_INO(sbi)) { 1732 trace_f2fs_drop_inode(inode, 1); 1733 return 1; 1734 } 1735 } 1736 1737 /* 1738 * This is to avoid a deadlock condition like below. 1739 * writeback_single_inode(inode) 1740 * - f2fs_write_data_page 1741 * - f2fs_gc -> iput -> evict 1742 * - inode_wait_for_writeback(inode) 1743 */ 1744 if ((!inode_unhashed(inode) && inode->i_state & I_SYNC)) { 1745 if (!inode->i_nlink && !is_bad_inode(inode)) { 1746 /* to avoid evict_inode call simultaneously */ 1747 atomic_inc(&inode->i_count); 1748 spin_unlock(&inode->i_lock); 1749 1750 /* should remain fi->extent_tree for writepage */ 1751 f2fs_destroy_extent_node(inode); 1752 1753 sb_start_intwrite(inode->i_sb); 1754 f2fs_i_size_write(inode, 0); 1755 1756 f2fs_submit_merged_write_cond(F2FS_I_SB(inode), 1757 inode, NULL, 0, DATA); 1758 truncate_inode_pages_final(inode->i_mapping); 1759 1760 if (F2FS_HAS_BLOCKS(inode)) 1761 f2fs_truncate(inode); 1762 1763 sb_end_intwrite(inode->i_sb); 1764 1765 spin_lock(&inode->i_lock); 1766 atomic_dec(&inode->i_count); 1767 } 1768 trace_f2fs_drop_inode(inode, 0); 1769 return 0; 1770 } 1771 ret = generic_drop_inode(inode); 1772 if (!ret) 1773 ret = fscrypt_drop_inode(inode); 1774 trace_f2fs_drop_inode(inode, ret); 1775 return ret; 1776 } 1777 1778 int f2fs_inode_dirtied(struct inode *inode, bool sync) 1779 { 1780 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1781 int ret = 0; 1782 1783 spin_lock(&sbi->inode_lock[DIRTY_META]); 1784 if (is_inode_flag_set(inode, FI_DIRTY_INODE)) { 1785 ret = 1; 1786 } else { 1787 set_inode_flag(inode, FI_DIRTY_INODE); 1788 stat_inc_dirty_inode(sbi, DIRTY_META); 1789 } 1790 if (sync && list_empty(&F2FS_I(inode)->gdirty_list)) { 1791 list_add_tail(&F2FS_I(inode)->gdirty_list, 1792 &sbi->inode_list[DIRTY_META]); 1793 inc_page_count(sbi, F2FS_DIRTY_IMETA); 1794 } 1795 spin_unlock(&sbi->inode_lock[DIRTY_META]); 1796 1797 /* if atomic write is not committed, set inode w/ atomic dirty */ 1798 if (!ret && f2fs_is_atomic_file(inode) && 1799 !is_inode_flag_set(inode, FI_ATOMIC_COMMITTED)) 1800 set_inode_flag(inode, FI_ATOMIC_DIRTIED); 1801 1802 return ret; 1803 } 1804 1805 void f2fs_inode_synced(struct inode *inode) 1806 { 1807 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1808 1809 spin_lock(&sbi->inode_lock[DIRTY_META]); 1810 if (!is_inode_flag_set(inode, FI_DIRTY_INODE)) { 1811 spin_unlock(&sbi->inode_lock[DIRTY_META]); 1812 return; 1813 } 1814 if (!list_empty(&F2FS_I(inode)->gdirty_list)) { 1815 list_del_init(&F2FS_I(inode)->gdirty_list); 1816 dec_page_count(sbi, F2FS_DIRTY_IMETA); 1817 } 1818 clear_inode_flag(inode, FI_DIRTY_INODE); 1819 clear_inode_flag(inode, FI_AUTO_RECOVER); 1820 stat_dec_dirty_inode(F2FS_I_SB(inode), DIRTY_META); 1821 spin_unlock(&sbi->inode_lock[DIRTY_META]); 1822 } 1823 1824 /* 1825 * f2fs_dirty_inode() is called from __mark_inode_dirty() 1826 * 1827 * We should call set_dirty_inode to write the dirty inode through write_inode. 1828 */ 1829 static void f2fs_dirty_inode(struct inode *inode, int flags) 1830 { 1831 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1832 1833 if (inode->i_ino == F2FS_NODE_INO(sbi) || 1834 inode->i_ino == F2FS_META_INO(sbi)) 1835 return; 1836 1837 if (is_inode_flag_set(inode, FI_AUTO_RECOVER)) 1838 clear_inode_flag(inode, FI_AUTO_RECOVER); 1839 1840 f2fs_inode_dirtied(inode, false); 1841 } 1842 1843 static void f2fs_free_inode(struct inode *inode) 1844 { 1845 fscrypt_free_inode(inode); 1846 kmem_cache_free(f2fs_inode_cachep, F2FS_I(inode)); 1847 } 1848 1849 static void destroy_percpu_info(struct f2fs_sb_info *sbi) 1850 { 1851 percpu_counter_destroy(&sbi->total_valid_inode_count); 1852 percpu_counter_destroy(&sbi->rf_node_block_count); 1853 percpu_counter_destroy(&sbi->alloc_valid_block_count); 1854 } 1855 1856 static void destroy_device_list(struct f2fs_sb_info *sbi) 1857 { 1858 int i; 1859 1860 for (i = 0; i < sbi->s_ndevs; i++) { 1861 if (i > 0) 1862 bdev_fput(FDEV(i).bdev_file); 1863 #ifdef CONFIG_BLK_DEV_ZONED 1864 kvfree(FDEV(i).blkz_seq); 1865 #endif 1866 } 1867 kvfree(sbi->devs); 1868 } 1869 1870 static void f2fs_put_super(struct super_block *sb) 1871 { 1872 struct f2fs_sb_info *sbi = F2FS_SB(sb); 1873 int i; 1874 int err = 0; 1875 bool done; 1876 1877 /* unregister procfs/sysfs entries in advance to avoid race case */ 1878 f2fs_unregister_sysfs(sbi); 1879 1880 f2fs_quota_off_umount(sb); 1881 1882 /* prevent remaining shrinker jobs */ 1883 mutex_lock(&sbi->umount_mutex); 1884 1885 /* 1886 * flush all issued checkpoints and stop checkpoint issue thread. 1887 * after then, all checkpoints should be done by each process context. 1888 */ 1889 f2fs_stop_ckpt_thread(sbi); 1890 1891 /* 1892 * We don't need to do checkpoint when superblock is clean. 1893 * But, the previous checkpoint was not done by umount, it needs to do 1894 * clean checkpoint again. 1895 */ 1896 if ((is_sbi_flag_set(sbi, SBI_IS_DIRTY) || 1897 !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG))) { 1898 struct cp_control cpc = { 1899 .reason = CP_UMOUNT, 1900 }; 1901 stat_inc_cp_call_count(sbi, TOTAL_CALL); 1902 err = f2fs_write_checkpoint(sbi, &cpc); 1903 } 1904 1905 /* be sure to wait for any on-going discard commands */ 1906 done = f2fs_issue_discard_timeout(sbi); 1907 if (f2fs_realtime_discard_enable(sbi) && !sbi->discard_blks && done) { 1908 struct cp_control cpc = { 1909 .reason = CP_UMOUNT | CP_TRIMMED, 1910 }; 1911 stat_inc_cp_call_count(sbi, TOTAL_CALL); 1912 err = f2fs_write_checkpoint(sbi, &cpc); 1913 } 1914 1915 /* 1916 * normally superblock is clean, so we need to release this. 1917 * In addition, EIO will skip do checkpoint, we need this as well. 1918 */ 1919 f2fs_release_ino_entry(sbi, true); 1920 1921 f2fs_leave_shrinker(sbi); 1922 mutex_unlock(&sbi->umount_mutex); 1923 1924 /* our cp_error case, we can wait for any writeback page */ 1925 f2fs_flush_merged_writes(sbi); 1926 1927 f2fs_wait_on_all_pages(sbi, F2FS_WB_CP_DATA); 1928 1929 if (err || f2fs_cp_error(sbi)) { 1930 truncate_inode_pages_final(NODE_MAPPING(sbi)); 1931 truncate_inode_pages_final(META_MAPPING(sbi)); 1932 } 1933 1934 for (i = 0; i < NR_COUNT_TYPE; i++) { 1935 if (!get_pages(sbi, i)) 1936 continue; 1937 f2fs_err(sbi, "detect filesystem reference count leak during " 1938 "umount, type: %d, count: %lld", i, get_pages(sbi, i)); 1939 f2fs_bug_on(sbi, 1); 1940 } 1941 1942 f2fs_bug_on(sbi, sbi->fsync_node_num); 1943 1944 f2fs_destroy_compress_inode(sbi); 1945 1946 iput(sbi->node_inode); 1947 sbi->node_inode = NULL; 1948 1949 iput(sbi->meta_inode); 1950 sbi->meta_inode = NULL; 1951 1952 /* 1953 * iput() can update stat information, if f2fs_write_checkpoint() 1954 * above failed with error. 1955 */ 1956 f2fs_destroy_stats(sbi); 1957 1958 /* destroy f2fs internal modules */ 1959 f2fs_destroy_node_manager(sbi); 1960 f2fs_destroy_segment_manager(sbi); 1961 1962 /* flush s_error_work before sbi destroy */ 1963 flush_work(&sbi->s_error_work); 1964 1965 f2fs_destroy_post_read_wq(sbi); 1966 1967 kvfree(sbi->ckpt); 1968 1969 kfree(sbi->raw_super); 1970 1971 f2fs_destroy_page_array_cache(sbi); 1972 f2fs_destroy_xattr_caches(sbi); 1973 #ifdef CONFIG_QUOTA 1974 for (i = 0; i < MAXQUOTAS; i++) 1975 kfree(F2FS_OPTION(sbi).s_qf_names[i]); 1976 #endif 1977 fscrypt_free_dummy_policy(&F2FS_OPTION(sbi).dummy_enc_policy); 1978 destroy_percpu_info(sbi); 1979 f2fs_destroy_iostat(sbi); 1980 for (i = 0; i < NR_PAGE_TYPE; i++) 1981 kfree(sbi->write_io[i]); 1982 #if IS_ENABLED(CONFIG_UNICODE) 1983 utf8_unload(sb->s_encoding); 1984 #endif 1985 } 1986 1987 int f2fs_sync_fs(struct super_block *sb, int sync) 1988 { 1989 struct f2fs_sb_info *sbi = F2FS_SB(sb); 1990 int err = 0; 1991 1992 if (unlikely(f2fs_cp_error(sbi))) 1993 return 0; 1994 if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED))) 1995 return 0; 1996 1997 trace_f2fs_sync_fs(sb, sync); 1998 1999 if (unlikely(is_sbi_flag_set(sbi, SBI_POR_DOING))) 2000 return -EAGAIN; 2001 2002 if (sync) { 2003 stat_inc_cp_call_count(sbi, TOTAL_CALL); 2004 err = f2fs_issue_checkpoint(sbi); 2005 } 2006 2007 return err; 2008 } 2009 2010 static int f2fs_freeze(struct super_block *sb) 2011 { 2012 struct f2fs_sb_info *sbi = F2FS_SB(sb); 2013 2014 if (f2fs_readonly(sb)) 2015 return 0; 2016 2017 /* IO error happened before */ 2018 if (unlikely(f2fs_cp_error(sbi))) 2019 return -EIO; 2020 2021 /* must be clean, since sync_filesystem() was already called */ 2022 if (is_sbi_flag_set(sbi, SBI_IS_DIRTY)) 2023 return -EINVAL; 2024 2025 sbi->umount_lock_holder = current; 2026 2027 /* Let's flush checkpoints and stop the thread. */ 2028 f2fs_flush_ckpt_thread(sbi); 2029 2030 sbi->umount_lock_holder = NULL; 2031 2032 /* to avoid deadlock on f2fs_evict_inode->SB_FREEZE_FS */ 2033 set_sbi_flag(sbi, SBI_IS_FREEZING); 2034 return 0; 2035 } 2036 2037 static int f2fs_unfreeze(struct super_block *sb) 2038 { 2039 struct f2fs_sb_info *sbi = F2FS_SB(sb); 2040 2041 /* 2042 * It will update discard_max_bytes of mounted lvm device to zero 2043 * after creating snapshot on this lvm device, let's drop all 2044 * remained discards. 2045 * We don't need to disable real-time discard because discard_max_bytes 2046 * will recover after removal of snapshot. 2047 */ 2048 if (test_opt(sbi, DISCARD) && !f2fs_hw_support_discard(sbi)) 2049 f2fs_issue_discard_timeout(sbi); 2050 2051 clear_sbi_flag(F2FS_SB(sb), SBI_IS_FREEZING); 2052 return 0; 2053 } 2054 2055 #ifdef CONFIG_QUOTA 2056 static int f2fs_statfs_project(struct super_block *sb, 2057 kprojid_t projid, struct kstatfs *buf) 2058 { 2059 struct kqid qid; 2060 struct dquot *dquot; 2061 u64 limit; 2062 u64 curblock; 2063 2064 qid = make_kqid_projid(projid); 2065 dquot = dqget(sb, qid); 2066 if (IS_ERR(dquot)) 2067 return PTR_ERR(dquot); 2068 spin_lock(&dquot->dq_dqb_lock); 2069 2070 limit = min_not_zero(dquot->dq_dqb.dqb_bsoftlimit, 2071 dquot->dq_dqb.dqb_bhardlimit); 2072 limit >>= sb->s_blocksize_bits; 2073 2074 if (limit) { 2075 uint64_t remaining = 0; 2076 2077 curblock = (dquot->dq_dqb.dqb_curspace + 2078 dquot->dq_dqb.dqb_rsvspace) >> sb->s_blocksize_bits; 2079 if (limit > curblock) 2080 remaining = limit - curblock; 2081 2082 buf->f_blocks = min(buf->f_blocks, limit); 2083 buf->f_bfree = min(buf->f_bfree, remaining); 2084 buf->f_bavail = min(buf->f_bavail, remaining); 2085 } 2086 2087 limit = min_not_zero(dquot->dq_dqb.dqb_isoftlimit, 2088 dquot->dq_dqb.dqb_ihardlimit); 2089 2090 if (limit) { 2091 uint64_t remaining = 0; 2092 2093 if (limit > dquot->dq_dqb.dqb_curinodes) 2094 remaining = limit - dquot->dq_dqb.dqb_curinodes; 2095 2096 buf->f_files = min(buf->f_files, limit); 2097 buf->f_ffree = min(buf->f_ffree, remaining); 2098 } 2099 2100 spin_unlock(&dquot->dq_dqb_lock); 2101 dqput(dquot); 2102 return 0; 2103 } 2104 #endif 2105 2106 static int f2fs_statfs(struct dentry *dentry, struct kstatfs *buf) 2107 { 2108 struct super_block *sb = dentry->d_sb; 2109 struct f2fs_sb_info *sbi = F2FS_SB(sb); 2110 u64 id = huge_encode_dev(sb->s_bdev->bd_dev); 2111 block_t total_count, user_block_count, start_count; 2112 u64 avail_node_count; 2113 unsigned int total_valid_node_count; 2114 2115 total_count = le64_to_cpu(sbi->raw_super->block_count); 2116 start_count = le32_to_cpu(sbi->raw_super->segment0_blkaddr); 2117 buf->f_type = F2FS_SUPER_MAGIC; 2118 buf->f_bsize = sbi->blocksize; 2119 2120 buf->f_blocks = total_count - start_count; 2121 2122 spin_lock(&sbi->stat_lock); 2123 if (sbi->carve_out) 2124 buf->f_blocks -= sbi->current_reserved_blocks; 2125 user_block_count = sbi->user_block_count; 2126 total_valid_node_count = valid_node_count(sbi); 2127 avail_node_count = sbi->total_node_count - F2FS_RESERVED_NODE_NUM; 2128 buf->f_bfree = user_block_count - valid_user_blocks(sbi) - 2129 sbi->current_reserved_blocks; 2130 2131 if (unlikely(buf->f_bfree <= sbi->unusable_block_count)) 2132 buf->f_bfree = 0; 2133 else 2134 buf->f_bfree -= sbi->unusable_block_count; 2135 spin_unlock(&sbi->stat_lock); 2136 2137 if (buf->f_bfree > F2FS_OPTION(sbi).root_reserved_blocks) 2138 buf->f_bavail = buf->f_bfree - 2139 F2FS_OPTION(sbi).root_reserved_blocks; 2140 else 2141 buf->f_bavail = 0; 2142 2143 if (avail_node_count > user_block_count) { 2144 buf->f_files = user_block_count; 2145 buf->f_ffree = buf->f_bavail; 2146 } else { 2147 buf->f_files = avail_node_count; 2148 buf->f_ffree = min(avail_node_count - total_valid_node_count, 2149 buf->f_bavail); 2150 } 2151 2152 buf->f_namelen = F2FS_NAME_LEN; 2153 buf->f_fsid = u64_to_fsid(id); 2154 2155 #ifdef CONFIG_QUOTA 2156 if (is_inode_flag_set(d_inode(dentry), FI_PROJ_INHERIT) && 2157 sb_has_quota_limits_enabled(sb, PRJQUOTA)) { 2158 f2fs_statfs_project(sb, F2FS_I(d_inode(dentry))->i_projid, buf); 2159 } 2160 #endif 2161 return 0; 2162 } 2163 2164 static inline void f2fs_show_quota_options(struct seq_file *seq, 2165 struct super_block *sb) 2166 { 2167 #ifdef CONFIG_QUOTA 2168 struct f2fs_sb_info *sbi = F2FS_SB(sb); 2169 2170 if (F2FS_OPTION(sbi).s_jquota_fmt) { 2171 char *fmtname = ""; 2172 2173 switch (F2FS_OPTION(sbi).s_jquota_fmt) { 2174 case QFMT_VFS_OLD: 2175 fmtname = "vfsold"; 2176 break; 2177 case QFMT_VFS_V0: 2178 fmtname = "vfsv0"; 2179 break; 2180 case QFMT_VFS_V1: 2181 fmtname = "vfsv1"; 2182 break; 2183 } 2184 seq_printf(seq, ",jqfmt=%s", fmtname); 2185 } 2186 2187 if (F2FS_OPTION(sbi).s_qf_names[USRQUOTA]) 2188 seq_show_option(seq, "usrjquota", 2189 F2FS_OPTION(sbi).s_qf_names[USRQUOTA]); 2190 2191 if (F2FS_OPTION(sbi).s_qf_names[GRPQUOTA]) 2192 seq_show_option(seq, "grpjquota", 2193 F2FS_OPTION(sbi).s_qf_names[GRPQUOTA]); 2194 2195 if (F2FS_OPTION(sbi).s_qf_names[PRJQUOTA]) 2196 seq_show_option(seq, "prjjquota", 2197 F2FS_OPTION(sbi).s_qf_names[PRJQUOTA]); 2198 #endif 2199 } 2200 2201 #ifdef CONFIG_F2FS_FS_COMPRESSION 2202 static inline void f2fs_show_compress_options(struct seq_file *seq, 2203 struct super_block *sb) 2204 { 2205 struct f2fs_sb_info *sbi = F2FS_SB(sb); 2206 char *algtype = ""; 2207 int i; 2208 2209 if (!f2fs_sb_has_compression(sbi)) 2210 return; 2211 2212 switch (F2FS_OPTION(sbi).compress_algorithm) { 2213 case COMPRESS_LZO: 2214 algtype = "lzo"; 2215 break; 2216 case COMPRESS_LZ4: 2217 algtype = "lz4"; 2218 break; 2219 case COMPRESS_ZSTD: 2220 algtype = "zstd"; 2221 break; 2222 case COMPRESS_LZORLE: 2223 algtype = "lzo-rle"; 2224 break; 2225 } 2226 seq_printf(seq, ",compress_algorithm=%s", algtype); 2227 2228 if (F2FS_OPTION(sbi).compress_level) 2229 seq_printf(seq, ":%d", F2FS_OPTION(sbi).compress_level); 2230 2231 seq_printf(seq, ",compress_log_size=%u", 2232 F2FS_OPTION(sbi).compress_log_size); 2233 2234 for (i = 0; i < F2FS_OPTION(sbi).compress_ext_cnt; i++) { 2235 seq_printf(seq, ",compress_extension=%s", 2236 F2FS_OPTION(sbi).extensions[i]); 2237 } 2238 2239 for (i = 0; i < F2FS_OPTION(sbi).nocompress_ext_cnt; i++) { 2240 seq_printf(seq, ",nocompress_extension=%s", 2241 F2FS_OPTION(sbi).noextensions[i]); 2242 } 2243 2244 if (F2FS_OPTION(sbi).compress_chksum) 2245 seq_puts(seq, ",compress_chksum"); 2246 2247 if (F2FS_OPTION(sbi).compress_mode == COMPR_MODE_FS) 2248 seq_printf(seq, ",compress_mode=%s", "fs"); 2249 else if (F2FS_OPTION(sbi).compress_mode == COMPR_MODE_USER) 2250 seq_printf(seq, ",compress_mode=%s", "user"); 2251 2252 if (test_opt(sbi, COMPRESS_CACHE)) 2253 seq_puts(seq, ",compress_cache"); 2254 } 2255 #endif 2256 2257 static int f2fs_show_options(struct seq_file *seq, struct dentry *root) 2258 { 2259 struct f2fs_sb_info *sbi = F2FS_SB(root->d_sb); 2260 2261 if (F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_SYNC) 2262 seq_printf(seq, ",background_gc=%s", "sync"); 2263 else if (F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_ON) 2264 seq_printf(seq, ",background_gc=%s", "on"); 2265 else if (F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_OFF) 2266 seq_printf(seq, ",background_gc=%s", "off"); 2267 2268 if (test_opt(sbi, GC_MERGE)) 2269 seq_puts(seq, ",gc_merge"); 2270 else 2271 seq_puts(seq, ",nogc_merge"); 2272 2273 if (test_opt(sbi, DISABLE_ROLL_FORWARD)) 2274 seq_puts(seq, ",disable_roll_forward"); 2275 if (test_opt(sbi, NORECOVERY)) 2276 seq_puts(seq, ",norecovery"); 2277 if (test_opt(sbi, DISCARD)) { 2278 seq_puts(seq, ",discard"); 2279 if (F2FS_OPTION(sbi).discard_unit == DISCARD_UNIT_BLOCK) 2280 seq_printf(seq, ",discard_unit=%s", "block"); 2281 else if (F2FS_OPTION(sbi).discard_unit == DISCARD_UNIT_SEGMENT) 2282 seq_printf(seq, ",discard_unit=%s", "segment"); 2283 else if (F2FS_OPTION(sbi).discard_unit == DISCARD_UNIT_SECTION) 2284 seq_printf(seq, ",discard_unit=%s", "section"); 2285 } else { 2286 seq_puts(seq, ",nodiscard"); 2287 } 2288 #ifdef CONFIG_F2FS_FS_XATTR 2289 if (test_opt(sbi, XATTR_USER)) 2290 seq_puts(seq, ",user_xattr"); 2291 else 2292 seq_puts(seq, ",nouser_xattr"); 2293 if (test_opt(sbi, INLINE_XATTR)) 2294 seq_puts(seq, ",inline_xattr"); 2295 else 2296 seq_puts(seq, ",noinline_xattr"); 2297 if (test_opt(sbi, INLINE_XATTR_SIZE)) 2298 seq_printf(seq, ",inline_xattr_size=%u", 2299 F2FS_OPTION(sbi).inline_xattr_size); 2300 #endif 2301 #ifdef CONFIG_F2FS_FS_POSIX_ACL 2302 if (test_opt(sbi, POSIX_ACL)) 2303 seq_puts(seq, ",acl"); 2304 else 2305 seq_puts(seq, ",noacl"); 2306 #endif 2307 if (test_opt(sbi, DISABLE_EXT_IDENTIFY)) 2308 seq_puts(seq, ",disable_ext_identify"); 2309 if (test_opt(sbi, INLINE_DATA)) 2310 seq_puts(seq, ",inline_data"); 2311 else 2312 seq_puts(seq, ",noinline_data"); 2313 if (test_opt(sbi, INLINE_DENTRY)) 2314 seq_puts(seq, ",inline_dentry"); 2315 else 2316 seq_puts(seq, ",noinline_dentry"); 2317 if (test_opt(sbi, FLUSH_MERGE)) 2318 seq_puts(seq, ",flush_merge"); 2319 else 2320 seq_puts(seq, ",noflush_merge"); 2321 if (test_opt(sbi, NOBARRIER)) 2322 seq_puts(seq, ",nobarrier"); 2323 else 2324 seq_puts(seq, ",barrier"); 2325 if (test_opt(sbi, FASTBOOT)) 2326 seq_puts(seq, ",fastboot"); 2327 if (test_opt(sbi, READ_EXTENT_CACHE)) 2328 seq_puts(seq, ",extent_cache"); 2329 else 2330 seq_puts(seq, ",noextent_cache"); 2331 if (test_opt(sbi, AGE_EXTENT_CACHE)) 2332 seq_puts(seq, ",age_extent_cache"); 2333 if (test_opt(sbi, DATA_FLUSH)) 2334 seq_puts(seq, ",data_flush"); 2335 2336 seq_puts(seq, ",mode="); 2337 if (F2FS_OPTION(sbi).fs_mode == FS_MODE_ADAPTIVE) 2338 seq_puts(seq, "adaptive"); 2339 else if (F2FS_OPTION(sbi).fs_mode == FS_MODE_LFS) 2340 seq_puts(seq, "lfs"); 2341 else if (F2FS_OPTION(sbi).fs_mode == FS_MODE_FRAGMENT_SEG) 2342 seq_puts(seq, "fragment:segment"); 2343 else if (F2FS_OPTION(sbi).fs_mode == FS_MODE_FRAGMENT_BLK) 2344 seq_puts(seq, "fragment:block"); 2345 seq_printf(seq, ",active_logs=%u", F2FS_OPTION(sbi).active_logs); 2346 if (test_opt(sbi, RESERVE_ROOT)) 2347 seq_printf(seq, ",reserve_root=%u,resuid=%u,resgid=%u", 2348 F2FS_OPTION(sbi).root_reserved_blocks, 2349 from_kuid_munged(&init_user_ns, 2350 F2FS_OPTION(sbi).s_resuid), 2351 from_kgid_munged(&init_user_ns, 2352 F2FS_OPTION(sbi).s_resgid)); 2353 #ifdef CONFIG_F2FS_FAULT_INJECTION 2354 if (test_opt(sbi, FAULT_INJECTION)) { 2355 seq_printf(seq, ",fault_injection=%u", 2356 F2FS_OPTION(sbi).fault_info.inject_rate); 2357 seq_printf(seq, ",fault_type=%u", 2358 F2FS_OPTION(sbi).fault_info.inject_type); 2359 } 2360 #endif 2361 #ifdef CONFIG_QUOTA 2362 if (test_opt(sbi, QUOTA)) 2363 seq_puts(seq, ",quota"); 2364 if (test_opt(sbi, USRQUOTA)) 2365 seq_puts(seq, ",usrquota"); 2366 if (test_opt(sbi, GRPQUOTA)) 2367 seq_puts(seq, ",grpquota"); 2368 if (test_opt(sbi, PRJQUOTA)) 2369 seq_puts(seq, ",prjquota"); 2370 #endif 2371 f2fs_show_quota_options(seq, sbi->sb); 2372 2373 fscrypt_show_test_dummy_encryption(seq, ',', sbi->sb); 2374 2375 if (sbi->sb->s_flags & SB_INLINECRYPT) 2376 seq_puts(seq, ",inlinecrypt"); 2377 2378 if (F2FS_OPTION(sbi).alloc_mode == ALLOC_MODE_DEFAULT) 2379 seq_printf(seq, ",alloc_mode=%s", "default"); 2380 else if (F2FS_OPTION(sbi).alloc_mode == ALLOC_MODE_REUSE) 2381 seq_printf(seq, ",alloc_mode=%s", "reuse"); 2382 2383 if (test_opt(sbi, DISABLE_CHECKPOINT)) 2384 seq_printf(seq, ",checkpoint=disable:%u", 2385 F2FS_OPTION(sbi).unusable_cap); 2386 if (test_opt(sbi, MERGE_CHECKPOINT)) 2387 seq_puts(seq, ",checkpoint_merge"); 2388 else 2389 seq_puts(seq, ",nocheckpoint_merge"); 2390 if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_POSIX) 2391 seq_printf(seq, ",fsync_mode=%s", "posix"); 2392 else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_STRICT) 2393 seq_printf(seq, ",fsync_mode=%s", "strict"); 2394 else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_NOBARRIER) 2395 seq_printf(seq, ",fsync_mode=%s", "nobarrier"); 2396 2397 #ifdef CONFIG_F2FS_FS_COMPRESSION 2398 f2fs_show_compress_options(seq, sbi->sb); 2399 #endif 2400 2401 if (test_opt(sbi, ATGC)) 2402 seq_puts(seq, ",atgc"); 2403 2404 if (F2FS_OPTION(sbi).memory_mode == MEMORY_MODE_NORMAL) 2405 seq_printf(seq, ",memory=%s", "normal"); 2406 else if (F2FS_OPTION(sbi).memory_mode == MEMORY_MODE_LOW) 2407 seq_printf(seq, ",memory=%s", "low"); 2408 2409 if (F2FS_OPTION(sbi).errors == MOUNT_ERRORS_READONLY) 2410 seq_printf(seq, ",errors=%s", "remount-ro"); 2411 else if (F2FS_OPTION(sbi).errors == MOUNT_ERRORS_CONTINUE) 2412 seq_printf(seq, ",errors=%s", "continue"); 2413 else if (F2FS_OPTION(sbi).errors == MOUNT_ERRORS_PANIC) 2414 seq_printf(seq, ",errors=%s", "panic"); 2415 2416 if (test_opt(sbi, NAT_BITS)) 2417 seq_puts(seq, ",nat_bits"); 2418 2419 return 0; 2420 } 2421 2422 static void default_options(struct f2fs_sb_info *sbi, bool remount) 2423 { 2424 /* init some FS parameters */ 2425 if (!remount) { 2426 set_opt(sbi, READ_EXTENT_CACHE); 2427 clear_opt(sbi, DISABLE_CHECKPOINT); 2428 2429 if (f2fs_hw_support_discard(sbi) || f2fs_hw_should_discard(sbi)) 2430 set_opt(sbi, DISCARD); 2431 2432 if (f2fs_sb_has_blkzoned(sbi)) 2433 F2FS_OPTION(sbi).discard_unit = DISCARD_UNIT_SECTION; 2434 else 2435 F2FS_OPTION(sbi).discard_unit = DISCARD_UNIT_BLOCK; 2436 } 2437 2438 if (f2fs_sb_has_readonly(sbi)) 2439 F2FS_OPTION(sbi).active_logs = NR_CURSEG_RO_TYPE; 2440 else 2441 F2FS_OPTION(sbi).active_logs = NR_CURSEG_PERSIST_TYPE; 2442 2443 F2FS_OPTION(sbi).inline_xattr_size = DEFAULT_INLINE_XATTR_ADDRS; 2444 if (le32_to_cpu(F2FS_RAW_SUPER(sbi)->segment_count_main) <= 2445 SMALL_VOLUME_SEGMENTS) 2446 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_REUSE; 2447 else 2448 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_DEFAULT; 2449 F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_POSIX; 2450 F2FS_OPTION(sbi).s_resuid = make_kuid(&init_user_ns, F2FS_DEF_RESUID); 2451 F2FS_OPTION(sbi).s_resgid = make_kgid(&init_user_ns, F2FS_DEF_RESGID); 2452 if (f2fs_sb_has_compression(sbi)) { 2453 F2FS_OPTION(sbi).compress_algorithm = COMPRESS_LZ4; 2454 F2FS_OPTION(sbi).compress_log_size = MIN_COMPRESS_LOG_SIZE; 2455 F2FS_OPTION(sbi).compress_ext_cnt = 0; 2456 F2FS_OPTION(sbi).compress_mode = COMPR_MODE_FS; 2457 } 2458 F2FS_OPTION(sbi).bggc_mode = BGGC_MODE_ON; 2459 F2FS_OPTION(sbi).memory_mode = MEMORY_MODE_NORMAL; 2460 F2FS_OPTION(sbi).errors = MOUNT_ERRORS_CONTINUE; 2461 2462 set_opt(sbi, INLINE_XATTR); 2463 set_opt(sbi, INLINE_DATA); 2464 set_opt(sbi, INLINE_DENTRY); 2465 set_opt(sbi, MERGE_CHECKPOINT); 2466 set_opt(sbi, LAZYTIME); 2467 F2FS_OPTION(sbi).unusable_cap = 0; 2468 if (!f2fs_is_readonly(sbi)) 2469 set_opt(sbi, FLUSH_MERGE); 2470 if (f2fs_sb_has_blkzoned(sbi)) 2471 F2FS_OPTION(sbi).fs_mode = FS_MODE_LFS; 2472 else 2473 F2FS_OPTION(sbi).fs_mode = FS_MODE_ADAPTIVE; 2474 2475 #ifdef CONFIG_F2FS_FS_XATTR 2476 set_opt(sbi, XATTR_USER); 2477 #endif 2478 #ifdef CONFIG_F2FS_FS_POSIX_ACL 2479 set_opt(sbi, POSIX_ACL); 2480 #endif 2481 2482 f2fs_build_fault_attr(sbi, 0, 0, FAULT_ALL); 2483 } 2484 2485 #ifdef CONFIG_QUOTA 2486 static int f2fs_enable_quotas(struct super_block *sb); 2487 #endif 2488 2489 static int f2fs_disable_checkpoint(struct f2fs_sb_info *sbi) 2490 { 2491 unsigned int s_flags = sbi->sb->s_flags; 2492 struct cp_control cpc; 2493 unsigned int gc_mode = sbi->gc_mode; 2494 int err = 0; 2495 int ret; 2496 block_t unusable; 2497 2498 if (s_flags & SB_RDONLY) { 2499 f2fs_err(sbi, "checkpoint=disable on readonly fs"); 2500 return -EINVAL; 2501 } 2502 sbi->sb->s_flags |= SB_ACTIVE; 2503 2504 /* check if we need more GC first */ 2505 unusable = f2fs_get_unusable_blocks(sbi); 2506 if (!f2fs_disable_cp_again(sbi, unusable)) 2507 goto skip_gc; 2508 2509 f2fs_update_time(sbi, DISABLE_TIME); 2510 2511 sbi->gc_mode = GC_URGENT_HIGH; 2512 2513 while (!f2fs_time_over(sbi, DISABLE_TIME)) { 2514 struct f2fs_gc_control gc_control = { 2515 .victim_segno = NULL_SEGNO, 2516 .init_gc_type = FG_GC, 2517 .should_migrate_blocks = false, 2518 .err_gc_skipped = true, 2519 .no_bg_gc = true, 2520 .nr_free_secs = 1 }; 2521 2522 f2fs_down_write(&sbi->gc_lock); 2523 stat_inc_gc_call_count(sbi, FOREGROUND); 2524 err = f2fs_gc(sbi, &gc_control); 2525 if (err == -ENODATA) { 2526 err = 0; 2527 break; 2528 } 2529 if (err && err != -EAGAIN) 2530 break; 2531 } 2532 2533 ret = sync_filesystem(sbi->sb); 2534 if (ret || err) { 2535 err = ret ? ret : err; 2536 goto restore_flag; 2537 } 2538 2539 unusable = f2fs_get_unusable_blocks(sbi); 2540 if (f2fs_disable_cp_again(sbi, unusable)) { 2541 err = -EAGAIN; 2542 goto restore_flag; 2543 } 2544 2545 skip_gc: 2546 f2fs_down_write(&sbi->gc_lock); 2547 cpc.reason = CP_PAUSE; 2548 set_sbi_flag(sbi, SBI_CP_DISABLED); 2549 stat_inc_cp_call_count(sbi, TOTAL_CALL); 2550 err = f2fs_write_checkpoint(sbi, &cpc); 2551 if (err) 2552 goto out_unlock; 2553 2554 spin_lock(&sbi->stat_lock); 2555 sbi->unusable_block_count = unusable; 2556 spin_unlock(&sbi->stat_lock); 2557 2558 out_unlock: 2559 f2fs_up_write(&sbi->gc_lock); 2560 restore_flag: 2561 sbi->gc_mode = gc_mode; 2562 sbi->sb->s_flags = s_flags; /* Restore SB_RDONLY status */ 2563 return err; 2564 } 2565 2566 static void f2fs_enable_checkpoint(struct f2fs_sb_info *sbi) 2567 { 2568 int retry = DEFAULT_RETRY_IO_COUNT; 2569 2570 /* we should flush all the data to keep data consistency */ 2571 do { 2572 sync_inodes_sb(sbi->sb); 2573 f2fs_io_schedule_timeout(DEFAULT_IO_TIMEOUT); 2574 } while (get_pages(sbi, F2FS_DIRTY_DATA) && retry--); 2575 2576 if (unlikely(retry < 0)) 2577 f2fs_warn(sbi, "checkpoint=enable has some unwritten data."); 2578 2579 f2fs_down_write(&sbi->gc_lock); 2580 f2fs_dirty_to_prefree(sbi); 2581 2582 clear_sbi_flag(sbi, SBI_CP_DISABLED); 2583 set_sbi_flag(sbi, SBI_IS_DIRTY); 2584 f2fs_up_write(&sbi->gc_lock); 2585 2586 f2fs_sync_fs(sbi->sb, 1); 2587 2588 /* Let's ensure there's no pending checkpoint anymore */ 2589 f2fs_flush_ckpt_thread(sbi); 2590 } 2591 2592 static int __f2fs_remount(struct fs_context *fc, struct super_block *sb) 2593 { 2594 struct f2fs_sb_info *sbi = F2FS_SB(sb); 2595 struct f2fs_mount_info org_mount_opt; 2596 unsigned long old_sb_flags; 2597 unsigned int flags = fc->sb_flags; 2598 int err; 2599 bool need_restart_gc = false, need_stop_gc = false; 2600 bool need_restart_flush = false, need_stop_flush = false; 2601 bool need_restart_discard = false, need_stop_discard = false; 2602 bool need_enable_checkpoint = false, need_disable_checkpoint = false; 2603 bool no_read_extent_cache = !test_opt(sbi, READ_EXTENT_CACHE); 2604 bool no_age_extent_cache = !test_opt(sbi, AGE_EXTENT_CACHE); 2605 bool enable_checkpoint = !test_opt(sbi, DISABLE_CHECKPOINT); 2606 bool no_atgc = !test_opt(sbi, ATGC); 2607 bool no_discard = !test_opt(sbi, DISCARD); 2608 bool no_compress_cache = !test_opt(sbi, COMPRESS_CACHE); 2609 bool block_unit_discard = f2fs_block_unit_discard(sbi); 2610 bool no_nat_bits = !test_opt(sbi, NAT_BITS); 2611 #ifdef CONFIG_QUOTA 2612 int i, j; 2613 #endif 2614 2615 /* 2616 * Save the old mount options in case we 2617 * need to restore them. 2618 */ 2619 org_mount_opt = sbi->mount_opt; 2620 old_sb_flags = sb->s_flags; 2621 2622 sbi->umount_lock_holder = current; 2623 2624 #ifdef CONFIG_QUOTA 2625 org_mount_opt.s_jquota_fmt = F2FS_OPTION(sbi).s_jquota_fmt; 2626 for (i = 0; i < MAXQUOTAS; i++) { 2627 if (F2FS_OPTION(sbi).s_qf_names[i]) { 2628 org_mount_opt.s_qf_names[i] = 2629 kstrdup(F2FS_OPTION(sbi).s_qf_names[i], 2630 GFP_KERNEL); 2631 if (!org_mount_opt.s_qf_names[i]) { 2632 for (j = 0; j < i; j++) 2633 kfree(org_mount_opt.s_qf_names[j]); 2634 return -ENOMEM; 2635 } 2636 } else { 2637 org_mount_opt.s_qf_names[i] = NULL; 2638 } 2639 } 2640 #endif 2641 2642 /* recover superblocks we couldn't write due to previous RO mount */ 2643 if (!(flags & SB_RDONLY) && is_sbi_flag_set(sbi, SBI_NEED_SB_WRITE)) { 2644 err = f2fs_commit_super(sbi, false); 2645 f2fs_info(sbi, "Try to recover all the superblocks, ret: %d", 2646 err); 2647 if (!err) 2648 clear_sbi_flag(sbi, SBI_NEED_SB_WRITE); 2649 } 2650 2651 default_options(sbi, true); 2652 2653 err = f2fs_check_opt_consistency(fc, sb); 2654 if (err) 2655 goto restore_opts; 2656 2657 f2fs_apply_options(fc, sb); 2658 2659 err = f2fs_sanity_check_options(sbi, true); 2660 if (err) 2661 goto restore_opts; 2662 2663 /* flush outstanding errors before changing fs state */ 2664 flush_work(&sbi->s_error_work); 2665 2666 /* 2667 * Previous and new state of filesystem is RO, 2668 * so skip checking GC and FLUSH_MERGE conditions. 2669 */ 2670 if (f2fs_readonly(sb) && (flags & SB_RDONLY)) 2671 goto skip; 2672 2673 if (f2fs_dev_is_readonly(sbi) && !(flags & SB_RDONLY)) { 2674 err = -EROFS; 2675 goto restore_opts; 2676 } 2677 2678 #ifdef CONFIG_QUOTA 2679 if (!f2fs_readonly(sb) && (flags & SB_RDONLY)) { 2680 err = dquot_suspend(sb, -1); 2681 if (err < 0) 2682 goto restore_opts; 2683 } else if (f2fs_readonly(sb) && !(flags & SB_RDONLY)) { 2684 /* dquot_resume needs RW */ 2685 sb->s_flags &= ~SB_RDONLY; 2686 if (sb_any_quota_suspended(sb)) { 2687 dquot_resume(sb, -1); 2688 } else if (f2fs_sb_has_quota_ino(sbi)) { 2689 err = f2fs_enable_quotas(sb); 2690 if (err) 2691 goto restore_opts; 2692 } 2693 } 2694 #endif 2695 /* disallow enable atgc dynamically */ 2696 if (no_atgc == !!test_opt(sbi, ATGC)) { 2697 err = -EINVAL; 2698 f2fs_warn(sbi, "switch atgc option is not allowed"); 2699 goto restore_opts; 2700 } 2701 2702 /* disallow enable/disable extent_cache dynamically */ 2703 if (no_read_extent_cache == !!test_opt(sbi, READ_EXTENT_CACHE)) { 2704 err = -EINVAL; 2705 f2fs_warn(sbi, "switch extent_cache option is not allowed"); 2706 goto restore_opts; 2707 } 2708 /* disallow enable/disable age extent_cache dynamically */ 2709 if (no_age_extent_cache == !!test_opt(sbi, AGE_EXTENT_CACHE)) { 2710 err = -EINVAL; 2711 f2fs_warn(sbi, "switch age_extent_cache option is not allowed"); 2712 goto restore_opts; 2713 } 2714 2715 if (no_compress_cache == !!test_opt(sbi, COMPRESS_CACHE)) { 2716 err = -EINVAL; 2717 f2fs_warn(sbi, "switch compress_cache option is not allowed"); 2718 goto restore_opts; 2719 } 2720 2721 if (block_unit_discard != f2fs_block_unit_discard(sbi)) { 2722 err = -EINVAL; 2723 f2fs_warn(sbi, "switch discard_unit option is not allowed"); 2724 goto restore_opts; 2725 } 2726 2727 if (no_nat_bits == !!test_opt(sbi, NAT_BITS)) { 2728 err = -EINVAL; 2729 f2fs_warn(sbi, "switch nat_bits option is not allowed"); 2730 goto restore_opts; 2731 } 2732 2733 if ((flags & SB_RDONLY) && test_opt(sbi, DISABLE_CHECKPOINT)) { 2734 err = -EINVAL; 2735 f2fs_warn(sbi, "disabling checkpoint not compatible with read-only"); 2736 goto restore_opts; 2737 } 2738 2739 /* 2740 * We stop the GC thread if FS is mounted as RO 2741 * or if background_gc = off is passed in mount 2742 * option. Also sync the filesystem. 2743 */ 2744 if ((flags & SB_RDONLY) || 2745 (F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_OFF && 2746 !test_opt(sbi, GC_MERGE))) { 2747 if (sbi->gc_thread) { 2748 f2fs_stop_gc_thread(sbi); 2749 need_restart_gc = true; 2750 } 2751 } else if (!sbi->gc_thread) { 2752 err = f2fs_start_gc_thread(sbi); 2753 if (err) 2754 goto restore_opts; 2755 need_stop_gc = true; 2756 } 2757 2758 if (flags & SB_RDONLY) { 2759 sync_inodes_sb(sb); 2760 2761 set_sbi_flag(sbi, SBI_IS_DIRTY); 2762 set_sbi_flag(sbi, SBI_IS_CLOSE); 2763 f2fs_sync_fs(sb, 1); 2764 clear_sbi_flag(sbi, SBI_IS_CLOSE); 2765 } 2766 2767 /* 2768 * We stop issue flush thread if FS is mounted as RO 2769 * or if flush_merge is not passed in mount option. 2770 */ 2771 if ((flags & SB_RDONLY) || !test_opt(sbi, FLUSH_MERGE)) { 2772 clear_opt(sbi, FLUSH_MERGE); 2773 f2fs_destroy_flush_cmd_control(sbi, false); 2774 need_restart_flush = true; 2775 } else { 2776 err = f2fs_create_flush_cmd_control(sbi); 2777 if (err) 2778 goto restore_gc; 2779 need_stop_flush = true; 2780 } 2781 2782 if (no_discard == !!test_opt(sbi, DISCARD)) { 2783 if (test_opt(sbi, DISCARD)) { 2784 err = f2fs_start_discard_thread(sbi); 2785 if (err) 2786 goto restore_flush; 2787 need_stop_discard = true; 2788 } else { 2789 f2fs_stop_discard_thread(sbi); 2790 f2fs_issue_discard_timeout(sbi); 2791 need_restart_discard = true; 2792 } 2793 } 2794 2795 adjust_unusable_cap_perc(sbi); 2796 if (enable_checkpoint == !!test_opt(sbi, DISABLE_CHECKPOINT)) { 2797 if (test_opt(sbi, DISABLE_CHECKPOINT)) { 2798 err = f2fs_disable_checkpoint(sbi); 2799 if (err) 2800 goto restore_discard; 2801 need_enable_checkpoint = true; 2802 } else { 2803 f2fs_enable_checkpoint(sbi); 2804 need_disable_checkpoint = true; 2805 } 2806 } 2807 2808 /* 2809 * Place this routine at the end, since a new checkpoint would be 2810 * triggered while remount and we need to take care of it before 2811 * returning from remount. 2812 */ 2813 if ((flags & SB_RDONLY) || test_opt(sbi, DISABLE_CHECKPOINT) || 2814 !test_opt(sbi, MERGE_CHECKPOINT)) { 2815 f2fs_stop_ckpt_thread(sbi); 2816 } else { 2817 /* Flush if the previous checkpoint, if exists. */ 2818 f2fs_flush_ckpt_thread(sbi); 2819 2820 err = f2fs_start_ckpt_thread(sbi); 2821 if (err) { 2822 f2fs_err(sbi, 2823 "Failed to start F2FS issue_checkpoint_thread (%d)", 2824 err); 2825 goto restore_checkpoint; 2826 } 2827 } 2828 2829 skip: 2830 #ifdef CONFIG_QUOTA 2831 /* Release old quota file names */ 2832 for (i = 0; i < MAXQUOTAS; i++) 2833 kfree(org_mount_opt.s_qf_names[i]); 2834 #endif 2835 /* Update the POSIXACL Flag */ 2836 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) | 2837 (test_opt(sbi, POSIX_ACL) ? SB_POSIXACL : 0); 2838 2839 limit_reserve_root(sbi); 2840 fc->sb_flags = (flags & ~SB_LAZYTIME) | (sb->s_flags & SB_LAZYTIME); 2841 2842 sbi->umount_lock_holder = NULL; 2843 return 0; 2844 restore_checkpoint: 2845 if (need_enable_checkpoint) { 2846 f2fs_enable_checkpoint(sbi); 2847 } else if (need_disable_checkpoint) { 2848 if (f2fs_disable_checkpoint(sbi)) 2849 f2fs_warn(sbi, "checkpoint has not been disabled"); 2850 } 2851 restore_discard: 2852 if (need_restart_discard) { 2853 if (f2fs_start_discard_thread(sbi)) 2854 f2fs_warn(sbi, "discard has been stopped"); 2855 } else if (need_stop_discard) { 2856 f2fs_stop_discard_thread(sbi); 2857 } 2858 restore_flush: 2859 if (need_restart_flush) { 2860 if (f2fs_create_flush_cmd_control(sbi)) 2861 f2fs_warn(sbi, "background flush thread has stopped"); 2862 } else if (need_stop_flush) { 2863 clear_opt(sbi, FLUSH_MERGE); 2864 f2fs_destroy_flush_cmd_control(sbi, false); 2865 } 2866 restore_gc: 2867 if (need_restart_gc) { 2868 if (f2fs_start_gc_thread(sbi)) 2869 f2fs_warn(sbi, "background gc thread has stopped"); 2870 } else if (need_stop_gc) { 2871 f2fs_stop_gc_thread(sbi); 2872 } 2873 restore_opts: 2874 #ifdef CONFIG_QUOTA 2875 F2FS_OPTION(sbi).s_jquota_fmt = org_mount_opt.s_jquota_fmt; 2876 for (i = 0; i < MAXQUOTAS; i++) { 2877 kfree(F2FS_OPTION(sbi).s_qf_names[i]); 2878 F2FS_OPTION(sbi).s_qf_names[i] = org_mount_opt.s_qf_names[i]; 2879 } 2880 #endif 2881 sbi->mount_opt = org_mount_opt; 2882 sb->s_flags = old_sb_flags; 2883 2884 sbi->umount_lock_holder = NULL; 2885 return err; 2886 } 2887 2888 static void f2fs_shutdown(struct super_block *sb) 2889 { 2890 f2fs_do_shutdown(F2FS_SB(sb), F2FS_GOING_DOWN_NOSYNC, false, false); 2891 } 2892 2893 #ifdef CONFIG_QUOTA 2894 static bool f2fs_need_recovery(struct f2fs_sb_info *sbi) 2895 { 2896 /* need to recovery orphan */ 2897 if (is_set_ckpt_flags(sbi, CP_ORPHAN_PRESENT_FLAG)) 2898 return true; 2899 /* need to recovery data */ 2900 if (test_opt(sbi, DISABLE_ROLL_FORWARD)) 2901 return false; 2902 if (test_opt(sbi, NORECOVERY)) 2903 return false; 2904 return !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG); 2905 } 2906 2907 static bool f2fs_recover_quota_begin(struct f2fs_sb_info *sbi) 2908 { 2909 bool readonly = f2fs_readonly(sbi->sb); 2910 2911 if (!f2fs_need_recovery(sbi)) 2912 return false; 2913 2914 /* it doesn't need to check f2fs_sb_has_readonly() */ 2915 if (f2fs_hw_is_readonly(sbi)) 2916 return false; 2917 2918 if (readonly) { 2919 sbi->sb->s_flags &= ~SB_RDONLY; 2920 set_sbi_flag(sbi, SBI_IS_WRITABLE); 2921 } 2922 2923 /* 2924 * Turn on quotas which were not enabled for read-only mounts if 2925 * filesystem has quota feature, so that they are updated correctly. 2926 */ 2927 return f2fs_enable_quota_files(sbi, readonly); 2928 } 2929 2930 static void f2fs_recover_quota_end(struct f2fs_sb_info *sbi, 2931 bool quota_enabled) 2932 { 2933 if (quota_enabled) 2934 f2fs_quota_off_umount(sbi->sb); 2935 2936 if (is_sbi_flag_set(sbi, SBI_IS_WRITABLE)) { 2937 clear_sbi_flag(sbi, SBI_IS_WRITABLE); 2938 sbi->sb->s_flags |= SB_RDONLY; 2939 } 2940 } 2941 2942 /* Read data from quotafile */ 2943 static ssize_t f2fs_quota_read(struct super_block *sb, int type, char *data, 2944 size_t len, loff_t off) 2945 { 2946 struct inode *inode = sb_dqopt(sb)->files[type]; 2947 struct address_space *mapping = inode->i_mapping; 2948 int tocopy; 2949 size_t toread; 2950 loff_t i_size = i_size_read(inode); 2951 2952 if (off > i_size) 2953 return 0; 2954 2955 if (off + len > i_size) 2956 len = i_size - off; 2957 toread = len; 2958 while (toread > 0) { 2959 struct folio *folio; 2960 size_t offset; 2961 2962 repeat: 2963 folio = mapping_read_folio_gfp(mapping, off >> PAGE_SHIFT, 2964 GFP_NOFS); 2965 if (IS_ERR(folio)) { 2966 if (PTR_ERR(folio) == -ENOMEM) { 2967 memalloc_retry_wait(GFP_NOFS); 2968 goto repeat; 2969 } 2970 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR); 2971 return PTR_ERR(folio); 2972 } 2973 offset = offset_in_folio(folio, off); 2974 tocopy = min(folio_size(folio) - offset, toread); 2975 2976 folio_lock(folio); 2977 2978 if (unlikely(folio->mapping != mapping)) { 2979 f2fs_folio_put(folio, true); 2980 goto repeat; 2981 } 2982 2983 /* 2984 * should never happen, just leave f2fs_bug_on() here to catch 2985 * any potential bug. 2986 */ 2987 f2fs_bug_on(F2FS_SB(sb), !folio_test_uptodate(folio)); 2988 2989 memcpy_from_folio(data, folio, offset, tocopy); 2990 f2fs_folio_put(folio, true); 2991 2992 toread -= tocopy; 2993 data += tocopy; 2994 off += tocopy; 2995 } 2996 return len; 2997 } 2998 2999 /* Write to quotafile */ 3000 static ssize_t f2fs_quota_write(struct super_block *sb, int type, 3001 const char *data, size_t len, loff_t off) 3002 { 3003 struct inode *inode = sb_dqopt(sb)->files[type]; 3004 struct address_space *mapping = inode->i_mapping; 3005 const struct address_space_operations *a_ops = mapping->a_ops; 3006 int offset = off & (sb->s_blocksize - 1); 3007 size_t towrite = len; 3008 struct folio *folio; 3009 void *fsdata = NULL; 3010 int err = 0; 3011 int tocopy; 3012 3013 while (towrite > 0) { 3014 tocopy = min_t(unsigned long, sb->s_blocksize - offset, 3015 towrite); 3016 retry: 3017 err = a_ops->write_begin(NULL, mapping, off, tocopy, 3018 &folio, &fsdata); 3019 if (unlikely(err)) { 3020 if (err == -ENOMEM) { 3021 f2fs_io_schedule_timeout(DEFAULT_IO_TIMEOUT); 3022 goto retry; 3023 } 3024 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR); 3025 break; 3026 } 3027 3028 memcpy_to_folio(folio, offset_in_folio(folio, off), data, tocopy); 3029 3030 a_ops->write_end(NULL, mapping, off, tocopy, tocopy, 3031 folio, fsdata); 3032 offset = 0; 3033 towrite -= tocopy; 3034 off += tocopy; 3035 data += tocopy; 3036 cond_resched(); 3037 } 3038 3039 if (len == towrite) 3040 return err; 3041 inode_set_mtime_to_ts(inode, inode_set_ctime_current(inode)); 3042 f2fs_mark_inode_dirty_sync(inode, false); 3043 return len - towrite; 3044 } 3045 3046 int f2fs_dquot_initialize(struct inode *inode) 3047 { 3048 if (time_to_inject(F2FS_I_SB(inode), FAULT_DQUOT_INIT)) 3049 return -ESRCH; 3050 3051 return dquot_initialize(inode); 3052 } 3053 3054 static struct dquot __rcu **f2fs_get_dquots(struct inode *inode) 3055 { 3056 return F2FS_I(inode)->i_dquot; 3057 } 3058 3059 static qsize_t *f2fs_get_reserved_space(struct inode *inode) 3060 { 3061 return &F2FS_I(inode)->i_reserved_quota; 3062 } 3063 3064 static int f2fs_quota_on_mount(struct f2fs_sb_info *sbi, int type) 3065 { 3066 if (is_set_ckpt_flags(sbi, CP_QUOTA_NEED_FSCK_FLAG)) { 3067 f2fs_err(sbi, "quota sysfile may be corrupted, skip loading it"); 3068 return 0; 3069 } 3070 3071 return dquot_quota_on_mount(sbi->sb, F2FS_OPTION(sbi).s_qf_names[type], 3072 F2FS_OPTION(sbi).s_jquota_fmt, type); 3073 } 3074 3075 int f2fs_enable_quota_files(struct f2fs_sb_info *sbi, bool rdonly) 3076 { 3077 int enabled = 0; 3078 int i, err; 3079 3080 if (f2fs_sb_has_quota_ino(sbi) && rdonly) { 3081 err = f2fs_enable_quotas(sbi->sb); 3082 if (err) { 3083 f2fs_err(sbi, "Cannot turn on quota_ino: %d", err); 3084 return 0; 3085 } 3086 return 1; 3087 } 3088 3089 for (i = 0; i < MAXQUOTAS; i++) { 3090 if (F2FS_OPTION(sbi).s_qf_names[i]) { 3091 err = f2fs_quota_on_mount(sbi, i); 3092 if (!err) { 3093 enabled = 1; 3094 continue; 3095 } 3096 f2fs_err(sbi, "Cannot turn on quotas: %d on %d", 3097 err, i); 3098 } 3099 } 3100 return enabled; 3101 } 3102 3103 static int f2fs_quota_enable(struct super_block *sb, int type, int format_id, 3104 unsigned int flags) 3105 { 3106 struct inode *qf_inode; 3107 unsigned long qf_inum; 3108 unsigned long qf_flag = F2FS_QUOTA_DEFAULT_FL; 3109 int err; 3110 3111 BUG_ON(!f2fs_sb_has_quota_ino(F2FS_SB(sb))); 3112 3113 qf_inum = f2fs_qf_ino(sb, type); 3114 if (!qf_inum) 3115 return -EPERM; 3116 3117 qf_inode = f2fs_iget(sb, qf_inum); 3118 if (IS_ERR(qf_inode)) { 3119 f2fs_err(F2FS_SB(sb), "Bad quota inode %u:%lu", type, qf_inum); 3120 return PTR_ERR(qf_inode); 3121 } 3122 3123 /* Don't account quota for quota files to avoid recursion */ 3124 inode_lock(qf_inode); 3125 qf_inode->i_flags |= S_NOQUOTA; 3126 3127 if ((F2FS_I(qf_inode)->i_flags & qf_flag) != qf_flag) { 3128 F2FS_I(qf_inode)->i_flags |= qf_flag; 3129 f2fs_set_inode_flags(qf_inode); 3130 } 3131 inode_unlock(qf_inode); 3132 3133 err = dquot_load_quota_inode(qf_inode, type, format_id, flags); 3134 iput(qf_inode); 3135 return err; 3136 } 3137 3138 static int f2fs_enable_quotas(struct super_block *sb) 3139 { 3140 struct f2fs_sb_info *sbi = F2FS_SB(sb); 3141 int type, err = 0; 3142 unsigned long qf_inum; 3143 bool quota_mopt[MAXQUOTAS] = { 3144 test_opt(sbi, USRQUOTA), 3145 test_opt(sbi, GRPQUOTA), 3146 test_opt(sbi, PRJQUOTA), 3147 }; 3148 3149 if (is_set_ckpt_flags(F2FS_SB(sb), CP_QUOTA_NEED_FSCK_FLAG)) { 3150 f2fs_err(sbi, "quota file may be corrupted, skip loading it"); 3151 return 0; 3152 } 3153 3154 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE; 3155 3156 for (type = 0; type < MAXQUOTAS; type++) { 3157 qf_inum = f2fs_qf_ino(sb, type); 3158 if (qf_inum) { 3159 err = f2fs_quota_enable(sb, type, QFMT_VFS_V1, 3160 DQUOT_USAGE_ENABLED | 3161 (quota_mopt[type] ? DQUOT_LIMITS_ENABLED : 0)); 3162 if (err) { 3163 f2fs_err(sbi, "Failed to enable quota tracking (type=%d, err=%d). Please run fsck to fix.", 3164 type, err); 3165 for (type--; type >= 0; type--) 3166 dquot_quota_off(sb, type); 3167 set_sbi_flag(F2FS_SB(sb), 3168 SBI_QUOTA_NEED_REPAIR); 3169 return err; 3170 } 3171 } 3172 } 3173 return 0; 3174 } 3175 3176 static int f2fs_quota_sync_file(struct f2fs_sb_info *sbi, int type) 3177 { 3178 struct quota_info *dqopt = sb_dqopt(sbi->sb); 3179 struct address_space *mapping = dqopt->files[type]->i_mapping; 3180 int ret = 0; 3181 3182 ret = dquot_writeback_dquots(sbi->sb, type); 3183 if (ret) 3184 goto out; 3185 3186 ret = filemap_fdatawrite(mapping); 3187 if (ret) 3188 goto out; 3189 3190 /* if we are using journalled quota */ 3191 if (is_journalled_quota(sbi)) 3192 goto out; 3193 3194 ret = filemap_fdatawait(mapping); 3195 3196 truncate_inode_pages(&dqopt->files[type]->i_data, 0); 3197 out: 3198 if (ret) 3199 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR); 3200 return ret; 3201 } 3202 3203 int f2fs_do_quota_sync(struct super_block *sb, int type) 3204 { 3205 struct f2fs_sb_info *sbi = F2FS_SB(sb); 3206 struct quota_info *dqopt = sb_dqopt(sb); 3207 int cnt; 3208 int ret = 0; 3209 3210 /* 3211 * Now when everything is written we can discard the pagecache so 3212 * that userspace sees the changes. 3213 */ 3214 for (cnt = 0; cnt < MAXQUOTAS; cnt++) { 3215 3216 if (type != -1 && cnt != type) 3217 continue; 3218 3219 if (!sb_has_quota_active(sb, cnt)) 3220 continue; 3221 3222 if (!f2fs_sb_has_quota_ino(sbi)) 3223 inode_lock(dqopt->files[cnt]); 3224 3225 /* 3226 * do_quotactl 3227 * f2fs_quota_sync 3228 * f2fs_down_read(quota_sem) 3229 * dquot_writeback_dquots() 3230 * f2fs_dquot_commit 3231 * block_operation 3232 * f2fs_down_read(quota_sem) 3233 */ 3234 f2fs_lock_op(sbi); 3235 f2fs_down_read(&sbi->quota_sem); 3236 3237 ret = f2fs_quota_sync_file(sbi, cnt); 3238 3239 f2fs_up_read(&sbi->quota_sem); 3240 f2fs_unlock_op(sbi); 3241 3242 if (!f2fs_sb_has_quota_ino(sbi)) 3243 inode_unlock(dqopt->files[cnt]); 3244 3245 if (ret) 3246 break; 3247 } 3248 return ret; 3249 } 3250 3251 static int f2fs_quota_sync(struct super_block *sb, int type) 3252 { 3253 int ret; 3254 3255 F2FS_SB(sb)->umount_lock_holder = current; 3256 ret = f2fs_do_quota_sync(sb, type); 3257 F2FS_SB(sb)->umount_lock_holder = NULL; 3258 return ret; 3259 } 3260 3261 static int f2fs_quota_on(struct super_block *sb, int type, int format_id, 3262 const struct path *path) 3263 { 3264 struct inode *inode; 3265 int err = 0; 3266 3267 /* if quota sysfile exists, deny enabling quota with specific file */ 3268 if (f2fs_sb_has_quota_ino(F2FS_SB(sb))) { 3269 f2fs_err(F2FS_SB(sb), "quota sysfile already exists"); 3270 return -EBUSY; 3271 } 3272 3273 if (path->dentry->d_sb != sb) 3274 return -EXDEV; 3275 3276 F2FS_SB(sb)->umount_lock_holder = current; 3277 3278 err = f2fs_do_quota_sync(sb, type); 3279 if (err) 3280 goto out; 3281 3282 inode = d_inode(path->dentry); 3283 3284 err = filemap_fdatawrite(inode->i_mapping); 3285 if (err) 3286 goto out; 3287 3288 err = filemap_fdatawait(inode->i_mapping); 3289 if (err) 3290 goto out; 3291 3292 err = dquot_quota_on(sb, type, format_id, path); 3293 if (err) 3294 goto out; 3295 3296 inode_lock(inode); 3297 F2FS_I(inode)->i_flags |= F2FS_QUOTA_DEFAULT_FL; 3298 f2fs_set_inode_flags(inode); 3299 inode_unlock(inode); 3300 f2fs_mark_inode_dirty_sync(inode, false); 3301 out: 3302 F2FS_SB(sb)->umount_lock_holder = NULL; 3303 return err; 3304 } 3305 3306 static int __f2fs_quota_off(struct super_block *sb, int type) 3307 { 3308 struct inode *inode = sb_dqopt(sb)->files[type]; 3309 int err; 3310 3311 if (!inode || !igrab(inode)) 3312 return dquot_quota_off(sb, type); 3313 3314 err = f2fs_do_quota_sync(sb, type); 3315 if (err) 3316 goto out_put; 3317 3318 err = dquot_quota_off(sb, type); 3319 if (err || f2fs_sb_has_quota_ino(F2FS_SB(sb))) 3320 goto out_put; 3321 3322 inode_lock(inode); 3323 F2FS_I(inode)->i_flags &= ~F2FS_QUOTA_DEFAULT_FL; 3324 f2fs_set_inode_flags(inode); 3325 inode_unlock(inode); 3326 f2fs_mark_inode_dirty_sync(inode, false); 3327 out_put: 3328 iput(inode); 3329 return err; 3330 } 3331 3332 static int f2fs_quota_off(struct super_block *sb, int type) 3333 { 3334 struct f2fs_sb_info *sbi = F2FS_SB(sb); 3335 int err; 3336 3337 F2FS_SB(sb)->umount_lock_holder = current; 3338 3339 err = __f2fs_quota_off(sb, type); 3340 3341 /* 3342 * quotactl can shutdown journalled quota, result in inconsistence 3343 * between quota record and fs data by following updates, tag the 3344 * flag to let fsck be aware of it. 3345 */ 3346 if (is_journalled_quota(sbi)) 3347 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR); 3348 3349 F2FS_SB(sb)->umount_lock_holder = NULL; 3350 3351 return err; 3352 } 3353 3354 void f2fs_quota_off_umount(struct super_block *sb) 3355 { 3356 int type; 3357 int err; 3358 3359 for (type = 0; type < MAXQUOTAS; type++) { 3360 err = __f2fs_quota_off(sb, type); 3361 if (err) { 3362 int ret = dquot_quota_off(sb, type); 3363 3364 f2fs_err(F2FS_SB(sb), "Fail to turn off disk quota (type: %d, err: %d, ret:%d), Please run fsck to fix it.", 3365 type, err, ret); 3366 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR); 3367 } 3368 } 3369 /* 3370 * In case of checkpoint=disable, we must flush quota blocks. 3371 * This can cause NULL exception for node_inode in end_io, since 3372 * put_super already dropped it. 3373 */ 3374 sync_filesystem(sb); 3375 } 3376 3377 static void f2fs_truncate_quota_inode_pages(struct super_block *sb) 3378 { 3379 struct quota_info *dqopt = sb_dqopt(sb); 3380 int type; 3381 3382 for (type = 0; type < MAXQUOTAS; type++) { 3383 if (!dqopt->files[type]) 3384 continue; 3385 f2fs_inode_synced(dqopt->files[type]); 3386 } 3387 } 3388 3389 static int f2fs_dquot_commit(struct dquot *dquot) 3390 { 3391 struct f2fs_sb_info *sbi = F2FS_SB(dquot->dq_sb); 3392 int ret; 3393 3394 f2fs_down_read_nested(&sbi->quota_sem, SINGLE_DEPTH_NESTING); 3395 ret = dquot_commit(dquot); 3396 if (ret < 0) 3397 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR); 3398 f2fs_up_read(&sbi->quota_sem); 3399 return ret; 3400 } 3401 3402 static int f2fs_dquot_acquire(struct dquot *dquot) 3403 { 3404 struct f2fs_sb_info *sbi = F2FS_SB(dquot->dq_sb); 3405 int ret; 3406 3407 f2fs_down_read(&sbi->quota_sem); 3408 ret = dquot_acquire(dquot); 3409 if (ret < 0) 3410 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR); 3411 f2fs_up_read(&sbi->quota_sem); 3412 return ret; 3413 } 3414 3415 static int f2fs_dquot_release(struct dquot *dquot) 3416 { 3417 struct f2fs_sb_info *sbi = F2FS_SB(dquot->dq_sb); 3418 int ret = dquot_release(dquot); 3419 3420 if (ret < 0) 3421 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR); 3422 return ret; 3423 } 3424 3425 static int f2fs_dquot_mark_dquot_dirty(struct dquot *dquot) 3426 { 3427 struct super_block *sb = dquot->dq_sb; 3428 struct f2fs_sb_info *sbi = F2FS_SB(sb); 3429 int ret = dquot_mark_dquot_dirty(dquot); 3430 3431 /* if we are using journalled quota */ 3432 if (is_journalled_quota(sbi)) 3433 set_sbi_flag(sbi, SBI_QUOTA_NEED_FLUSH); 3434 3435 return ret; 3436 } 3437 3438 static int f2fs_dquot_commit_info(struct super_block *sb, int type) 3439 { 3440 struct f2fs_sb_info *sbi = F2FS_SB(sb); 3441 int ret = dquot_commit_info(sb, type); 3442 3443 if (ret < 0) 3444 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR); 3445 return ret; 3446 } 3447 3448 static int f2fs_get_projid(struct inode *inode, kprojid_t *projid) 3449 { 3450 *projid = F2FS_I(inode)->i_projid; 3451 return 0; 3452 } 3453 3454 static const struct dquot_operations f2fs_quota_operations = { 3455 .get_reserved_space = f2fs_get_reserved_space, 3456 .write_dquot = f2fs_dquot_commit, 3457 .acquire_dquot = f2fs_dquot_acquire, 3458 .release_dquot = f2fs_dquot_release, 3459 .mark_dirty = f2fs_dquot_mark_dquot_dirty, 3460 .write_info = f2fs_dquot_commit_info, 3461 .alloc_dquot = dquot_alloc, 3462 .destroy_dquot = dquot_destroy, 3463 .get_projid = f2fs_get_projid, 3464 .get_next_id = dquot_get_next_id, 3465 }; 3466 3467 static const struct quotactl_ops f2fs_quotactl_ops = { 3468 .quota_on = f2fs_quota_on, 3469 .quota_off = f2fs_quota_off, 3470 .quota_sync = f2fs_quota_sync, 3471 .get_state = dquot_get_state, 3472 .set_info = dquot_set_dqinfo, 3473 .get_dqblk = dquot_get_dqblk, 3474 .set_dqblk = dquot_set_dqblk, 3475 .get_nextdqblk = dquot_get_next_dqblk, 3476 }; 3477 #else 3478 int f2fs_dquot_initialize(struct inode *inode) 3479 { 3480 return 0; 3481 } 3482 3483 int f2fs_do_quota_sync(struct super_block *sb, int type) 3484 { 3485 return 0; 3486 } 3487 3488 void f2fs_quota_off_umount(struct super_block *sb) 3489 { 3490 } 3491 #endif 3492 3493 static const struct super_operations f2fs_sops = { 3494 .alloc_inode = f2fs_alloc_inode, 3495 .free_inode = f2fs_free_inode, 3496 .drop_inode = f2fs_drop_inode, 3497 .write_inode = f2fs_write_inode, 3498 .dirty_inode = f2fs_dirty_inode, 3499 .show_options = f2fs_show_options, 3500 #ifdef CONFIG_QUOTA 3501 .quota_read = f2fs_quota_read, 3502 .quota_write = f2fs_quota_write, 3503 .get_dquots = f2fs_get_dquots, 3504 #endif 3505 .evict_inode = f2fs_evict_inode, 3506 .put_super = f2fs_put_super, 3507 .sync_fs = f2fs_sync_fs, 3508 .freeze_fs = f2fs_freeze, 3509 .unfreeze_fs = f2fs_unfreeze, 3510 .statfs = f2fs_statfs, 3511 .shutdown = f2fs_shutdown, 3512 }; 3513 3514 #ifdef CONFIG_FS_ENCRYPTION 3515 static int f2fs_get_context(struct inode *inode, void *ctx, size_t len) 3516 { 3517 return f2fs_getxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION, 3518 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT, 3519 ctx, len, NULL); 3520 } 3521 3522 static int f2fs_set_context(struct inode *inode, const void *ctx, size_t len, 3523 void *fs_data) 3524 { 3525 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 3526 3527 /* 3528 * Encrypting the root directory is not allowed because fsck 3529 * expects lost+found directory to exist and remain unencrypted 3530 * if LOST_FOUND feature is enabled. 3531 * 3532 */ 3533 if (f2fs_sb_has_lost_found(sbi) && 3534 inode->i_ino == F2FS_ROOT_INO(sbi)) 3535 return -EPERM; 3536 3537 return f2fs_setxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION, 3538 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT, 3539 ctx, len, fs_data, XATTR_CREATE); 3540 } 3541 3542 static const union fscrypt_policy *f2fs_get_dummy_policy(struct super_block *sb) 3543 { 3544 return F2FS_OPTION(F2FS_SB(sb)).dummy_enc_policy.policy; 3545 } 3546 3547 static bool f2fs_has_stable_inodes(struct super_block *sb) 3548 { 3549 return true; 3550 } 3551 3552 static struct block_device **f2fs_get_devices(struct super_block *sb, 3553 unsigned int *num_devs) 3554 { 3555 struct f2fs_sb_info *sbi = F2FS_SB(sb); 3556 struct block_device **devs; 3557 int i; 3558 3559 if (!f2fs_is_multi_device(sbi)) 3560 return NULL; 3561 3562 devs = kmalloc_array(sbi->s_ndevs, sizeof(*devs), GFP_KERNEL); 3563 if (!devs) 3564 return ERR_PTR(-ENOMEM); 3565 3566 for (i = 0; i < sbi->s_ndevs; i++) 3567 devs[i] = FDEV(i).bdev; 3568 *num_devs = sbi->s_ndevs; 3569 return devs; 3570 } 3571 3572 static const struct fscrypt_operations f2fs_cryptops = { 3573 .needs_bounce_pages = 1, 3574 .has_32bit_inodes = 1, 3575 .supports_subblock_data_units = 1, 3576 .legacy_key_prefix = "f2fs:", 3577 .get_context = f2fs_get_context, 3578 .set_context = f2fs_set_context, 3579 .get_dummy_policy = f2fs_get_dummy_policy, 3580 .empty_dir = f2fs_empty_dir, 3581 .has_stable_inodes = f2fs_has_stable_inodes, 3582 .get_devices = f2fs_get_devices, 3583 }; 3584 #endif 3585 3586 static struct inode *f2fs_nfs_get_inode(struct super_block *sb, 3587 u64 ino, u32 generation) 3588 { 3589 struct f2fs_sb_info *sbi = F2FS_SB(sb); 3590 struct inode *inode; 3591 3592 if (f2fs_check_nid_range(sbi, ino)) 3593 return ERR_PTR(-ESTALE); 3594 3595 /* 3596 * f2fs_iget isn't quite right if the inode is currently unallocated! 3597 * However f2fs_iget currently does appropriate checks to handle stale 3598 * inodes so everything is OK. 3599 */ 3600 inode = f2fs_iget(sb, ino); 3601 if (IS_ERR(inode)) 3602 return ERR_CAST(inode); 3603 if (unlikely(generation && inode->i_generation != generation)) { 3604 /* we didn't find the right inode.. */ 3605 iput(inode); 3606 return ERR_PTR(-ESTALE); 3607 } 3608 return inode; 3609 } 3610 3611 static struct dentry *f2fs_fh_to_dentry(struct super_block *sb, struct fid *fid, 3612 int fh_len, int fh_type) 3613 { 3614 return generic_fh_to_dentry(sb, fid, fh_len, fh_type, 3615 f2fs_nfs_get_inode); 3616 } 3617 3618 static struct dentry *f2fs_fh_to_parent(struct super_block *sb, struct fid *fid, 3619 int fh_len, int fh_type) 3620 { 3621 return generic_fh_to_parent(sb, fid, fh_len, fh_type, 3622 f2fs_nfs_get_inode); 3623 } 3624 3625 static const struct export_operations f2fs_export_ops = { 3626 .encode_fh = generic_encode_ino32_fh, 3627 .fh_to_dentry = f2fs_fh_to_dentry, 3628 .fh_to_parent = f2fs_fh_to_parent, 3629 .get_parent = f2fs_get_parent, 3630 }; 3631 3632 loff_t max_file_blocks(struct inode *inode) 3633 { 3634 loff_t result = 0; 3635 loff_t leaf_count; 3636 3637 /* 3638 * note: previously, result is equal to (DEF_ADDRS_PER_INODE - 3639 * DEFAULT_INLINE_XATTR_ADDRS), but now f2fs try to reserve more 3640 * space in inode.i_addr, it will be more safe to reassign 3641 * result as zero. 3642 */ 3643 3644 if (inode && f2fs_compressed_file(inode)) 3645 leaf_count = ADDRS_PER_BLOCK(inode); 3646 else 3647 leaf_count = DEF_ADDRS_PER_BLOCK; 3648 3649 /* two direct node blocks */ 3650 result += (leaf_count * 2); 3651 3652 /* two indirect node blocks */ 3653 leaf_count *= NIDS_PER_BLOCK; 3654 result += (leaf_count * 2); 3655 3656 /* one double indirect node block */ 3657 leaf_count *= NIDS_PER_BLOCK; 3658 result += leaf_count; 3659 3660 /* 3661 * For compatibility with FSCRYPT_POLICY_FLAG_IV_INO_LBLK_{64,32} with 3662 * a 4K crypto data unit, we must restrict the max filesize to what can 3663 * fit within U32_MAX + 1 data units. 3664 */ 3665 3666 result = umin(result, F2FS_BYTES_TO_BLK(((loff_t)U32_MAX + 1) * 4096)); 3667 3668 return result; 3669 } 3670 3671 static int __f2fs_commit_super(struct f2fs_sb_info *sbi, struct folio *folio, 3672 pgoff_t index, bool update) 3673 { 3674 struct bio *bio; 3675 /* it's rare case, we can do fua all the time */ 3676 blk_opf_t opf = REQ_OP_WRITE | REQ_SYNC | REQ_PREFLUSH | REQ_FUA; 3677 int ret; 3678 3679 folio_lock(folio); 3680 folio_wait_writeback(folio); 3681 if (update) 3682 memcpy(F2FS_SUPER_BLOCK(folio, index), F2FS_RAW_SUPER(sbi), 3683 sizeof(struct f2fs_super_block)); 3684 folio_mark_dirty(folio); 3685 folio_clear_dirty_for_io(folio); 3686 folio_start_writeback(folio); 3687 folio_unlock(folio); 3688 3689 bio = bio_alloc(sbi->sb->s_bdev, 1, opf, GFP_NOFS); 3690 3691 /* it doesn't need to set crypto context for superblock update */ 3692 bio->bi_iter.bi_sector = SECTOR_FROM_BLOCK(folio->index); 3693 3694 if (!bio_add_folio(bio, folio, folio_size(folio), 0)) 3695 f2fs_bug_on(sbi, 1); 3696 3697 ret = submit_bio_wait(bio); 3698 bio_put(bio); 3699 folio_end_writeback(folio); 3700 3701 return ret; 3702 } 3703 3704 static inline bool sanity_check_area_boundary(struct f2fs_sb_info *sbi, 3705 struct folio *folio, pgoff_t index) 3706 { 3707 struct f2fs_super_block *raw_super = F2FS_SUPER_BLOCK(folio, index); 3708 struct super_block *sb = sbi->sb; 3709 u32 segment0_blkaddr = le32_to_cpu(raw_super->segment0_blkaddr); 3710 u32 cp_blkaddr = le32_to_cpu(raw_super->cp_blkaddr); 3711 u32 sit_blkaddr = le32_to_cpu(raw_super->sit_blkaddr); 3712 u32 nat_blkaddr = le32_to_cpu(raw_super->nat_blkaddr); 3713 u32 ssa_blkaddr = le32_to_cpu(raw_super->ssa_blkaddr); 3714 u32 main_blkaddr = le32_to_cpu(raw_super->main_blkaddr); 3715 u32 segment_count_ckpt = le32_to_cpu(raw_super->segment_count_ckpt); 3716 u32 segment_count_sit = le32_to_cpu(raw_super->segment_count_sit); 3717 u32 segment_count_nat = le32_to_cpu(raw_super->segment_count_nat); 3718 u32 segment_count_ssa = le32_to_cpu(raw_super->segment_count_ssa); 3719 u32 segment_count_main = le32_to_cpu(raw_super->segment_count_main); 3720 u32 segment_count = le32_to_cpu(raw_super->segment_count); 3721 u32 log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg); 3722 u64 main_end_blkaddr = main_blkaddr + 3723 ((u64)segment_count_main << log_blocks_per_seg); 3724 u64 seg_end_blkaddr = segment0_blkaddr + 3725 ((u64)segment_count << log_blocks_per_seg); 3726 3727 if (segment0_blkaddr != cp_blkaddr) { 3728 f2fs_info(sbi, "Mismatch start address, segment0(%u) cp_blkaddr(%u)", 3729 segment0_blkaddr, cp_blkaddr); 3730 return true; 3731 } 3732 3733 if (cp_blkaddr + (segment_count_ckpt << log_blocks_per_seg) != 3734 sit_blkaddr) { 3735 f2fs_info(sbi, "Wrong CP boundary, start(%u) end(%u) blocks(%u)", 3736 cp_blkaddr, sit_blkaddr, 3737 segment_count_ckpt << log_blocks_per_seg); 3738 return true; 3739 } 3740 3741 if (sit_blkaddr + (segment_count_sit << log_blocks_per_seg) != 3742 nat_blkaddr) { 3743 f2fs_info(sbi, "Wrong SIT boundary, start(%u) end(%u) blocks(%u)", 3744 sit_blkaddr, nat_blkaddr, 3745 segment_count_sit << log_blocks_per_seg); 3746 return true; 3747 } 3748 3749 if (nat_blkaddr + (segment_count_nat << log_blocks_per_seg) != 3750 ssa_blkaddr) { 3751 f2fs_info(sbi, "Wrong NAT boundary, start(%u) end(%u) blocks(%u)", 3752 nat_blkaddr, ssa_blkaddr, 3753 segment_count_nat << log_blocks_per_seg); 3754 return true; 3755 } 3756 3757 if (ssa_blkaddr + (segment_count_ssa << log_blocks_per_seg) != 3758 main_blkaddr) { 3759 f2fs_info(sbi, "Wrong SSA boundary, start(%u) end(%u) blocks(%u)", 3760 ssa_blkaddr, main_blkaddr, 3761 segment_count_ssa << log_blocks_per_seg); 3762 return true; 3763 } 3764 3765 if (main_end_blkaddr > seg_end_blkaddr) { 3766 f2fs_info(sbi, "Wrong MAIN_AREA boundary, start(%u) end(%llu) block(%u)", 3767 main_blkaddr, seg_end_blkaddr, 3768 segment_count_main << log_blocks_per_seg); 3769 return true; 3770 } else if (main_end_blkaddr < seg_end_blkaddr) { 3771 int err = 0; 3772 char *res; 3773 3774 /* fix in-memory information all the time */ 3775 raw_super->segment_count = cpu_to_le32((main_end_blkaddr - 3776 segment0_blkaddr) >> log_blocks_per_seg); 3777 3778 if (f2fs_readonly(sb) || f2fs_hw_is_readonly(sbi)) { 3779 set_sbi_flag(sbi, SBI_NEED_SB_WRITE); 3780 res = "internally"; 3781 } else { 3782 err = __f2fs_commit_super(sbi, folio, index, false); 3783 res = err ? "failed" : "done"; 3784 } 3785 f2fs_info(sbi, "Fix alignment : %s, start(%u) end(%llu) block(%u)", 3786 res, main_blkaddr, seg_end_blkaddr, 3787 segment_count_main << log_blocks_per_seg); 3788 if (err) 3789 return true; 3790 } 3791 return false; 3792 } 3793 3794 static int sanity_check_raw_super(struct f2fs_sb_info *sbi, 3795 struct folio *folio, pgoff_t index) 3796 { 3797 block_t segment_count, segs_per_sec, secs_per_zone, segment_count_main; 3798 block_t total_sections, blocks_per_seg; 3799 struct f2fs_super_block *raw_super = F2FS_SUPER_BLOCK(folio, index); 3800 size_t crc_offset = 0; 3801 __u32 crc = 0; 3802 3803 if (le32_to_cpu(raw_super->magic) != F2FS_SUPER_MAGIC) { 3804 f2fs_info(sbi, "Magic Mismatch, valid(0x%x) - read(0x%x)", 3805 F2FS_SUPER_MAGIC, le32_to_cpu(raw_super->magic)); 3806 return -EINVAL; 3807 } 3808 3809 /* Check checksum_offset and crc in superblock */ 3810 if (__F2FS_HAS_FEATURE(raw_super, F2FS_FEATURE_SB_CHKSUM)) { 3811 crc_offset = le32_to_cpu(raw_super->checksum_offset); 3812 if (crc_offset != 3813 offsetof(struct f2fs_super_block, crc)) { 3814 f2fs_info(sbi, "Invalid SB checksum offset: %zu", 3815 crc_offset); 3816 return -EFSCORRUPTED; 3817 } 3818 crc = le32_to_cpu(raw_super->crc); 3819 if (crc != f2fs_crc32(raw_super, crc_offset)) { 3820 f2fs_info(sbi, "Invalid SB checksum value: %u", crc); 3821 return -EFSCORRUPTED; 3822 } 3823 } 3824 3825 /* only support block_size equals to PAGE_SIZE */ 3826 if (le32_to_cpu(raw_super->log_blocksize) != F2FS_BLKSIZE_BITS) { 3827 f2fs_info(sbi, "Invalid log_blocksize (%u), supports only %u", 3828 le32_to_cpu(raw_super->log_blocksize), 3829 F2FS_BLKSIZE_BITS); 3830 return -EFSCORRUPTED; 3831 } 3832 3833 /* check log blocks per segment */ 3834 if (le32_to_cpu(raw_super->log_blocks_per_seg) != 9) { 3835 f2fs_info(sbi, "Invalid log blocks per segment (%u)", 3836 le32_to_cpu(raw_super->log_blocks_per_seg)); 3837 return -EFSCORRUPTED; 3838 } 3839 3840 /* Currently, support 512/1024/2048/4096/16K bytes sector size */ 3841 if (le32_to_cpu(raw_super->log_sectorsize) > 3842 F2FS_MAX_LOG_SECTOR_SIZE || 3843 le32_to_cpu(raw_super->log_sectorsize) < 3844 F2FS_MIN_LOG_SECTOR_SIZE) { 3845 f2fs_info(sbi, "Invalid log sectorsize (%u)", 3846 le32_to_cpu(raw_super->log_sectorsize)); 3847 return -EFSCORRUPTED; 3848 } 3849 if (le32_to_cpu(raw_super->log_sectors_per_block) + 3850 le32_to_cpu(raw_super->log_sectorsize) != 3851 F2FS_MAX_LOG_SECTOR_SIZE) { 3852 f2fs_info(sbi, "Invalid log sectors per block(%u) log sectorsize(%u)", 3853 le32_to_cpu(raw_super->log_sectors_per_block), 3854 le32_to_cpu(raw_super->log_sectorsize)); 3855 return -EFSCORRUPTED; 3856 } 3857 3858 segment_count = le32_to_cpu(raw_super->segment_count); 3859 segment_count_main = le32_to_cpu(raw_super->segment_count_main); 3860 segs_per_sec = le32_to_cpu(raw_super->segs_per_sec); 3861 secs_per_zone = le32_to_cpu(raw_super->secs_per_zone); 3862 total_sections = le32_to_cpu(raw_super->section_count); 3863 3864 /* blocks_per_seg should be 512, given the above check */ 3865 blocks_per_seg = BIT(le32_to_cpu(raw_super->log_blocks_per_seg)); 3866 3867 if (segment_count > F2FS_MAX_SEGMENT || 3868 segment_count < F2FS_MIN_SEGMENTS) { 3869 f2fs_info(sbi, "Invalid segment count (%u)", segment_count); 3870 return -EFSCORRUPTED; 3871 } 3872 3873 if (total_sections > segment_count_main || total_sections < 1 || 3874 segs_per_sec > segment_count || !segs_per_sec) { 3875 f2fs_info(sbi, "Invalid segment/section count (%u, %u x %u)", 3876 segment_count, total_sections, segs_per_sec); 3877 return -EFSCORRUPTED; 3878 } 3879 3880 if (segment_count_main != total_sections * segs_per_sec) { 3881 f2fs_info(sbi, "Invalid segment/section count (%u != %u * %u)", 3882 segment_count_main, total_sections, segs_per_sec); 3883 return -EFSCORRUPTED; 3884 } 3885 3886 if ((segment_count / segs_per_sec) < total_sections) { 3887 f2fs_info(sbi, "Small segment_count (%u < %u * %u)", 3888 segment_count, segs_per_sec, total_sections); 3889 return -EFSCORRUPTED; 3890 } 3891 3892 if (segment_count > (le64_to_cpu(raw_super->block_count) >> 9)) { 3893 f2fs_info(sbi, "Wrong segment_count / block_count (%u > %llu)", 3894 segment_count, le64_to_cpu(raw_super->block_count)); 3895 return -EFSCORRUPTED; 3896 } 3897 3898 if (RDEV(0).path[0]) { 3899 block_t dev_seg_count = le32_to_cpu(RDEV(0).total_segments); 3900 int i = 1; 3901 3902 while (i < MAX_DEVICES && RDEV(i).path[0]) { 3903 dev_seg_count += le32_to_cpu(RDEV(i).total_segments); 3904 i++; 3905 } 3906 if (segment_count != dev_seg_count) { 3907 f2fs_info(sbi, "Segment count (%u) mismatch with total segments from devices (%u)", 3908 segment_count, dev_seg_count); 3909 return -EFSCORRUPTED; 3910 } 3911 } else { 3912 if (__F2FS_HAS_FEATURE(raw_super, F2FS_FEATURE_BLKZONED) && 3913 !bdev_is_zoned(sbi->sb->s_bdev)) { 3914 f2fs_info(sbi, "Zoned block device path is missing"); 3915 return -EFSCORRUPTED; 3916 } 3917 } 3918 3919 if (secs_per_zone > total_sections || !secs_per_zone) { 3920 f2fs_info(sbi, "Wrong secs_per_zone / total_sections (%u, %u)", 3921 secs_per_zone, total_sections); 3922 return -EFSCORRUPTED; 3923 } 3924 if (le32_to_cpu(raw_super->extension_count) > F2FS_MAX_EXTENSION || 3925 raw_super->hot_ext_count > F2FS_MAX_EXTENSION || 3926 (le32_to_cpu(raw_super->extension_count) + 3927 raw_super->hot_ext_count) > F2FS_MAX_EXTENSION) { 3928 f2fs_info(sbi, "Corrupted extension count (%u + %u > %u)", 3929 le32_to_cpu(raw_super->extension_count), 3930 raw_super->hot_ext_count, 3931 F2FS_MAX_EXTENSION); 3932 return -EFSCORRUPTED; 3933 } 3934 3935 if (le32_to_cpu(raw_super->cp_payload) >= 3936 (blocks_per_seg - F2FS_CP_PACKS - 3937 NR_CURSEG_PERSIST_TYPE)) { 3938 f2fs_info(sbi, "Insane cp_payload (%u >= %u)", 3939 le32_to_cpu(raw_super->cp_payload), 3940 blocks_per_seg - F2FS_CP_PACKS - 3941 NR_CURSEG_PERSIST_TYPE); 3942 return -EFSCORRUPTED; 3943 } 3944 3945 /* check reserved ino info */ 3946 if (le32_to_cpu(raw_super->node_ino) != 1 || 3947 le32_to_cpu(raw_super->meta_ino) != 2 || 3948 le32_to_cpu(raw_super->root_ino) != 3) { 3949 f2fs_info(sbi, "Invalid Fs Meta Ino: node(%u) meta(%u) root(%u)", 3950 le32_to_cpu(raw_super->node_ino), 3951 le32_to_cpu(raw_super->meta_ino), 3952 le32_to_cpu(raw_super->root_ino)); 3953 return -EFSCORRUPTED; 3954 } 3955 3956 /* check CP/SIT/NAT/SSA/MAIN_AREA area boundary */ 3957 if (sanity_check_area_boundary(sbi, folio, index)) 3958 return -EFSCORRUPTED; 3959 3960 return 0; 3961 } 3962 3963 int f2fs_sanity_check_ckpt(struct f2fs_sb_info *sbi) 3964 { 3965 unsigned int total, fsmeta; 3966 struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi); 3967 struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi); 3968 unsigned int ovp_segments, reserved_segments; 3969 unsigned int main_segs, blocks_per_seg; 3970 unsigned int sit_segs, nat_segs; 3971 unsigned int sit_bitmap_size, nat_bitmap_size; 3972 unsigned int log_blocks_per_seg; 3973 unsigned int segment_count_main; 3974 unsigned int cp_pack_start_sum, cp_payload; 3975 block_t user_block_count, valid_user_blocks; 3976 block_t avail_node_count, valid_node_count; 3977 unsigned int nat_blocks, nat_bits_bytes, nat_bits_blocks; 3978 unsigned int sit_blk_cnt; 3979 int i, j; 3980 3981 total = le32_to_cpu(raw_super->segment_count); 3982 fsmeta = le32_to_cpu(raw_super->segment_count_ckpt); 3983 sit_segs = le32_to_cpu(raw_super->segment_count_sit); 3984 fsmeta += sit_segs; 3985 nat_segs = le32_to_cpu(raw_super->segment_count_nat); 3986 fsmeta += nat_segs; 3987 fsmeta += le32_to_cpu(ckpt->rsvd_segment_count); 3988 fsmeta += le32_to_cpu(raw_super->segment_count_ssa); 3989 3990 if (unlikely(fsmeta >= total)) 3991 return 1; 3992 3993 ovp_segments = le32_to_cpu(ckpt->overprov_segment_count); 3994 reserved_segments = le32_to_cpu(ckpt->rsvd_segment_count); 3995 3996 if (!f2fs_sb_has_readonly(sbi) && 3997 unlikely(fsmeta < F2FS_MIN_META_SEGMENTS || 3998 ovp_segments == 0 || reserved_segments == 0)) { 3999 f2fs_err(sbi, "Wrong layout: check mkfs.f2fs version"); 4000 return 1; 4001 } 4002 user_block_count = le64_to_cpu(ckpt->user_block_count); 4003 segment_count_main = le32_to_cpu(raw_super->segment_count_main) + 4004 (f2fs_sb_has_readonly(sbi) ? 1 : 0); 4005 log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg); 4006 if (!user_block_count || user_block_count >= 4007 segment_count_main << log_blocks_per_seg) { 4008 f2fs_err(sbi, "Wrong user_block_count: %u", 4009 user_block_count); 4010 return 1; 4011 } 4012 4013 valid_user_blocks = le64_to_cpu(ckpt->valid_block_count); 4014 if (valid_user_blocks > user_block_count) { 4015 f2fs_err(sbi, "Wrong valid_user_blocks: %u, user_block_count: %u", 4016 valid_user_blocks, user_block_count); 4017 return 1; 4018 } 4019 4020 valid_node_count = le32_to_cpu(ckpt->valid_node_count); 4021 avail_node_count = sbi->total_node_count - F2FS_RESERVED_NODE_NUM; 4022 if (valid_node_count > avail_node_count) { 4023 f2fs_err(sbi, "Wrong valid_node_count: %u, avail_node_count: %u", 4024 valid_node_count, avail_node_count); 4025 return 1; 4026 } 4027 4028 main_segs = le32_to_cpu(raw_super->segment_count_main); 4029 blocks_per_seg = BLKS_PER_SEG(sbi); 4030 4031 for (i = 0; i < NR_CURSEG_NODE_TYPE; i++) { 4032 if (le32_to_cpu(ckpt->cur_node_segno[i]) >= main_segs || 4033 le16_to_cpu(ckpt->cur_node_blkoff[i]) >= blocks_per_seg) 4034 return 1; 4035 4036 if (f2fs_sb_has_readonly(sbi)) 4037 goto check_data; 4038 4039 for (j = i + 1; j < NR_CURSEG_NODE_TYPE; j++) { 4040 if (le32_to_cpu(ckpt->cur_node_segno[i]) == 4041 le32_to_cpu(ckpt->cur_node_segno[j])) { 4042 f2fs_err(sbi, "Node segment (%u, %u) has the same segno: %u", 4043 i, j, 4044 le32_to_cpu(ckpt->cur_node_segno[i])); 4045 return 1; 4046 } 4047 } 4048 } 4049 check_data: 4050 for (i = 0; i < NR_CURSEG_DATA_TYPE; i++) { 4051 if (le32_to_cpu(ckpt->cur_data_segno[i]) >= main_segs || 4052 le16_to_cpu(ckpt->cur_data_blkoff[i]) >= blocks_per_seg) 4053 return 1; 4054 4055 if (f2fs_sb_has_readonly(sbi)) 4056 goto skip_cross; 4057 4058 for (j = i + 1; j < NR_CURSEG_DATA_TYPE; j++) { 4059 if (le32_to_cpu(ckpt->cur_data_segno[i]) == 4060 le32_to_cpu(ckpt->cur_data_segno[j])) { 4061 f2fs_err(sbi, "Data segment (%u, %u) has the same segno: %u", 4062 i, j, 4063 le32_to_cpu(ckpt->cur_data_segno[i])); 4064 return 1; 4065 } 4066 } 4067 } 4068 for (i = 0; i < NR_CURSEG_NODE_TYPE; i++) { 4069 for (j = 0; j < NR_CURSEG_DATA_TYPE; j++) { 4070 if (le32_to_cpu(ckpt->cur_node_segno[i]) == 4071 le32_to_cpu(ckpt->cur_data_segno[j])) { 4072 f2fs_err(sbi, "Node segment (%u) and Data segment (%u) has the same segno: %u", 4073 i, j, 4074 le32_to_cpu(ckpt->cur_node_segno[i])); 4075 return 1; 4076 } 4077 } 4078 } 4079 skip_cross: 4080 sit_bitmap_size = le32_to_cpu(ckpt->sit_ver_bitmap_bytesize); 4081 nat_bitmap_size = le32_to_cpu(ckpt->nat_ver_bitmap_bytesize); 4082 4083 if (sit_bitmap_size != ((sit_segs / 2) << log_blocks_per_seg) / 8 || 4084 nat_bitmap_size != ((nat_segs / 2) << log_blocks_per_seg) / 8) { 4085 f2fs_err(sbi, "Wrong bitmap size: sit: %u, nat:%u", 4086 sit_bitmap_size, nat_bitmap_size); 4087 return 1; 4088 } 4089 4090 sit_blk_cnt = DIV_ROUND_UP(main_segs, SIT_ENTRY_PER_BLOCK); 4091 if (sit_bitmap_size * 8 < sit_blk_cnt) { 4092 f2fs_err(sbi, "Wrong bitmap size: sit: %u, sit_blk_cnt:%u", 4093 sit_bitmap_size, sit_blk_cnt); 4094 return 1; 4095 } 4096 4097 cp_pack_start_sum = __start_sum_addr(sbi); 4098 cp_payload = __cp_payload(sbi); 4099 if (cp_pack_start_sum < cp_payload + 1 || 4100 cp_pack_start_sum > blocks_per_seg - 1 - 4101 NR_CURSEG_PERSIST_TYPE) { 4102 f2fs_err(sbi, "Wrong cp_pack_start_sum: %u", 4103 cp_pack_start_sum); 4104 return 1; 4105 } 4106 4107 if (__is_set_ckpt_flags(ckpt, CP_LARGE_NAT_BITMAP_FLAG) && 4108 le32_to_cpu(ckpt->checksum_offset) != CP_MIN_CHKSUM_OFFSET) { 4109 f2fs_warn(sbi, "using deprecated layout of large_nat_bitmap, " 4110 "please run fsck v1.13.0 or higher to repair, chksum_offset: %u, " 4111 "fixed with patch: \"f2fs-tools: relocate chksum_offset for large_nat_bitmap feature\"", 4112 le32_to_cpu(ckpt->checksum_offset)); 4113 return 1; 4114 } 4115 4116 nat_blocks = nat_segs << log_blocks_per_seg; 4117 nat_bits_bytes = nat_blocks / BITS_PER_BYTE; 4118 nat_bits_blocks = F2FS_BLK_ALIGN((nat_bits_bytes << 1) + 8); 4119 if (__is_set_ckpt_flags(ckpt, CP_NAT_BITS_FLAG) && 4120 (cp_payload + F2FS_CP_PACKS + 4121 NR_CURSEG_PERSIST_TYPE + nat_bits_blocks >= blocks_per_seg)) { 4122 f2fs_warn(sbi, "Insane cp_payload: %u, nat_bits_blocks: %u)", 4123 cp_payload, nat_bits_blocks); 4124 return 1; 4125 } 4126 4127 if (unlikely(f2fs_cp_error(sbi))) { 4128 f2fs_err(sbi, "A bug case: need to run fsck"); 4129 return 1; 4130 } 4131 return 0; 4132 } 4133 4134 static void init_sb_info(struct f2fs_sb_info *sbi) 4135 { 4136 struct f2fs_super_block *raw_super = sbi->raw_super; 4137 int i; 4138 4139 sbi->log_sectors_per_block = 4140 le32_to_cpu(raw_super->log_sectors_per_block); 4141 sbi->log_blocksize = le32_to_cpu(raw_super->log_blocksize); 4142 sbi->blocksize = BIT(sbi->log_blocksize); 4143 sbi->log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg); 4144 sbi->blocks_per_seg = BIT(sbi->log_blocks_per_seg); 4145 sbi->segs_per_sec = le32_to_cpu(raw_super->segs_per_sec); 4146 sbi->secs_per_zone = le32_to_cpu(raw_super->secs_per_zone); 4147 sbi->total_sections = le32_to_cpu(raw_super->section_count); 4148 sbi->total_node_count = SEGS_TO_BLKS(sbi, 4149 ((le32_to_cpu(raw_super->segment_count_nat) / 2) * 4150 NAT_ENTRY_PER_BLOCK)); 4151 F2FS_ROOT_INO(sbi) = le32_to_cpu(raw_super->root_ino); 4152 F2FS_NODE_INO(sbi) = le32_to_cpu(raw_super->node_ino); 4153 F2FS_META_INO(sbi) = le32_to_cpu(raw_super->meta_ino); 4154 sbi->cur_victim_sec = NULL_SECNO; 4155 sbi->gc_mode = GC_NORMAL; 4156 sbi->next_victim_seg[BG_GC] = NULL_SEGNO; 4157 sbi->next_victim_seg[FG_GC] = NULL_SEGNO; 4158 sbi->max_victim_search = DEF_MAX_VICTIM_SEARCH; 4159 sbi->migration_granularity = SEGS_PER_SEC(sbi); 4160 sbi->migration_window_granularity = f2fs_sb_has_blkzoned(sbi) ? 4161 DEF_MIGRATION_WINDOW_GRANULARITY_ZONED : SEGS_PER_SEC(sbi); 4162 sbi->seq_file_ra_mul = MIN_RA_MUL; 4163 sbi->max_fragment_chunk = DEF_FRAGMENT_SIZE; 4164 sbi->max_fragment_hole = DEF_FRAGMENT_SIZE; 4165 spin_lock_init(&sbi->gc_remaining_trials_lock); 4166 atomic64_set(&sbi->current_atomic_write, 0); 4167 4168 sbi->dir_level = DEF_DIR_LEVEL; 4169 sbi->interval_time[CP_TIME] = DEF_CP_INTERVAL; 4170 sbi->interval_time[REQ_TIME] = DEF_IDLE_INTERVAL; 4171 sbi->interval_time[DISCARD_TIME] = DEF_IDLE_INTERVAL; 4172 sbi->interval_time[GC_TIME] = DEF_IDLE_INTERVAL; 4173 sbi->interval_time[DISABLE_TIME] = DEF_DISABLE_INTERVAL; 4174 sbi->interval_time[UMOUNT_DISCARD_TIMEOUT] = 4175 DEF_UMOUNT_DISCARD_TIMEOUT; 4176 clear_sbi_flag(sbi, SBI_NEED_FSCK); 4177 4178 for (i = 0; i < NR_COUNT_TYPE; i++) 4179 atomic_set(&sbi->nr_pages[i], 0); 4180 4181 for (i = 0; i < META; i++) 4182 atomic_set(&sbi->wb_sync_req[i], 0); 4183 4184 INIT_LIST_HEAD(&sbi->s_list); 4185 mutex_init(&sbi->umount_mutex); 4186 init_f2fs_rwsem(&sbi->io_order_lock); 4187 spin_lock_init(&sbi->cp_lock); 4188 4189 sbi->dirty_device = 0; 4190 spin_lock_init(&sbi->dev_lock); 4191 4192 init_f2fs_rwsem(&sbi->sb_lock); 4193 init_f2fs_rwsem(&sbi->pin_sem); 4194 } 4195 4196 static int init_percpu_info(struct f2fs_sb_info *sbi) 4197 { 4198 int err; 4199 4200 err = percpu_counter_init(&sbi->alloc_valid_block_count, 0, GFP_KERNEL); 4201 if (err) 4202 return err; 4203 4204 err = percpu_counter_init(&sbi->rf_node_block_count, 0, GFP_KERNEL); 4205 if (err) 4206 goto err_valid_block; 4207 4208 err = percpu_counter_init(&sbi->total_valid_inode_count, 0, 4209 GFP_KERNEL); 4210 if (err) 4211 goto err_node_block; 4212 return 0; 4213 4214 err_node_block: 4215 percpu_counter_destroy(&sbi->rf_node_block_count); 4216 err_valid_block: 4217 percpu_counter_destroy(&sbi->alloc_valid_block_count); 4218 return err; 4219 } 4220 4221 #ifdef CONFIG_BLK_DEV_ZONED 4222 4223 struct f2fs_report_zones_args { 4224 struct f2fs_sb_info *sbi; 4225 struct f2fs_dev_info *dev; 4226 }; 4227 4228 static int f2fs_report_zone_cb(struct blk_zone *zone, unsigned int idx, 4229 void *data) 4230 { 4231 struct f2fs_report_zones_args *rz_args = data; 4232 block_t unusable_blocks = (zone->len - zone->capacity) >> 4233 F2FS_LOG_SECTORS_PER_BLOCK; 4234 4235 if (zone->type == BLK_ZONE_TYPE_CONVENTIONAL) 4236 return 0; 4237 4238 set_bit(idx, rz_args->dev->blkz_seq); 4239 if (!rz_args->sbi->unusable_blocks_per_sec) { 4240 rz_args->sbi->unusable_blocks_per_sec = unusable_blocks; 4241 return 0; 4242 } 4243 if (rz_args->sbi->unusable_blocks_per_sec != unusable_blocks) { 4244 f2fs_err(rz_args->sbi, "F2FS supports single zone capacity\n"); 4245 return -EINVAL; 4246 } 4247 return 0; 4248 } 4249 4250 static int init_blkz_info(struct f2fs_sb_info *sbi, int devi) 4251 { 4252 struct block_device *bdev = FDEV(devi).bdev; 4253 sector_t nr_sectors = bdev_nr_sectors(bdev); 4254 struct f2fs_report_zones_args rep_zone_arg; 4255 u64 zone_sectors; 4256 unsigned int max_open_zones; 4257 int ret; 4258 4259 if (!f2fs_sb_has_blkzoned(sbi)) 4260 return 0; 4261 4262 if (bdev_is_zoned(FDEV(devi).bdev)) { 4263 max_open_zones = bdev_max_open_zones(bdev); 4264 if (max_open_zones && (max_open_zones < sbi->max_open_zones)) 4265 sbi->max_open_zones = max_open_zones; 4266 if (sbi->max_open_zones < F2FS_OPTION(sbi).active_logs) { 4267 f2fs_err(sbi, 4268 "zoned: max open zones %u is too small, need at least %u open zones", 4269 sbi->max_open_zones, F2FS_OPTION(sbi).active_logs); 4270 return -EINVAL; 4271 } 4272 } 4273 4274 zone_sectors = bdev_zone_sectors(bdev); 4275 if (sbi->blocks_per_blkz && sbi->blocks_per_blkz != 4276 SECTOR_TO_BLOCK(zone_sectors)) 4277 return -EINVAL; 4278 sbi->blocks_per_blkz = SECTOR_TO_BLOCK(zone_sectors); 4279 FDEV(devi).nr_blkz = div_u64(SECTOR_TO_BLOCK(nr_sectors), 4280 sbi->blocks_per_blkz); 4281 if (nr_sectors & (zone_sectors - 1)) 4282 FDEV(devi).nr_blkz++; 4283 4284 FDEV(devi).blkz_seq = f2fs_kvzalloc(sbi, 4285 BITS_TO_LONGS(FDEV(devi).nr_blkz) 4286 * sizeof(unsigned long), 4287 GFP_KERNEL); 4288 if (!FDEV(devi).blkz_seq) 4289 return -ENOMEM; 4290 4291 rep_zone_arg.sbi = sbi; 4292 rep_zone_arg.dev = &FDEV(devi); 4293 4294 ret = blkdev_report_zones(bdev, 0, BLK_ALL_ZONES, f2fs_report_zone_cb, 4295 &rep_zone_arg); 4296 if (ret < 0) 4297 return ret; 4298 return 0; 4299 } 4300 #endif 4301 4302 /* 4303 * Read f2fs raw super block. 4304 * Because we have two copies of super block, so read both of them 4305 * to get the first valid one. If any one of them is broken, we pass 4306 * them recovery flag back to the caller. 4307 */ 4308 static int read_raw_super_block(struct f2fs_sb_info *sbi, 4309 struct f2fs_super_block **raw_super, 4310 int *valid_super_block, int *recovery) 4311 { 4312 struct super_block *sb = sbi->sb; 4313 int block; 4314 struct folio *folio; 4315 struct f2fs_super_block *super; 4316 int err = 0; 4317 4318 super = kzalloc(sizeof(struct f2fs_super_block), GFP_KERNEL); 4319 if (!super) 4320 return -ENOMEM; 4321 4322 for (block = 0; block < 2; block++) { 4323 folio = read_mapping_folio(sb->s_bdev->bd_mapping, block, NULL); 4324 if (IS_ERR(folio)) { 4325 f2fs_err(sbi, "Unable to read %dth superblock", 4326 block + 1); 4327 err = PTR_ERR(folio); 4328 *recovery = 1; 4329 continue; 4330 } 4331 4332 /* sanity checking of raw super */ 4333 err = sanity_check_raw_super(sbi, folio, block); 4334 if (err) { 4335 f2fs_err(sbi, "Can't find valid F2FS filesystem in %dth superblock", 4336 block + 1); 4337 folio_put(folio); 4338 *recovery = 1; 4339 continue; 4340 } 4341 4342 if (!*raw_super) { 4343 memcpy(super, F2FS_SUPER_BLOCK(folio, block), 4344 sizeof(*super)); 4345 *valid_super_block = block; 4346 *raw_super = super; 4347 } 4348 folio_put(folio); 4349 } 4350 4351 /* No valid superblock */ 4352 if (!*raw_super) 4353 kfree(super); 4354 else 4355 err = 0; 4356 4357 return err; 4358 } 4359 4360 int f2fs_commit_super(struct f2fs_sb_info *sbi, bool recover) 4361 { 4362 struct folio *folio; 4363 pgoff_t index; 4364 __u32 crc = 0; 4365 int err; 4366 4367 if ((recover && f2fs_readonly(sbi->sb)) || 4368 f2fs_hw_is_readonly(sbi)) { 4369 set_sbi_flag(sbi, SBI_NEED_SB_WRITE); 4370 return -EROFS; 4371 } 4372 4373 /* we should update superblock crc here */ 4374 if (!recover && f2fs_sb_has_sb_chksum(sbi)) { 4375 crc = f2fs_crc32(F2FS_RAW_SUPER(sbi), 4376 offsetof(struct f2fs_super_block, crc)); 4377 F2FS_RAW_SUPER(sbi)->crc = cpu_to_le32(crc); 4378 } 4379 4380 /* write back-up superblock first */ 4381 index = sbi->valid_super_block ? 0 : 1; 4382 folio = read_mapping_folio(sbi->sb->s_bdev->bd_mapping, index, NULL); 4383 if (IS_ERR(folio)) 4384 return PTR_ERR(folio); 4385 err = __f2fs_commit_super(sbi, folio, index, true); 4386 folio_put(folio); 4387 4388 /* if we are in recovery path, skip writing valid superblock */ 4389 if (recover || err) 4390 return err; 4391 4392 /* write current valid superblock */ 4393 index = sbi->valid_super_block; 4394 folio = read_mapping_folio(sbi->sb->s_bdev->bd_mapping, index, NULL); 4395 if (IS_ERR(folio)) 4396 return PTR_ERR(folio); 4397 err = __f2fs_commit_super(sbi, folio, index, true); 4398 folio_put(folio); 4399 return err; 4400 } 4401 4402 static void save_stop_reason(struct f2fs_sb_info *sbi, unsigned char reason) 4403 { 4404 unsigned long flags; 4405 4406 spin_lock_irqsave(&sbi->error_lock, flags); 4407 if (sbi->stop_reason[reason] < GENMASK(BITS_PER_BYTE - 1, 0)) 4408 sbi->stop_reason[reason]++; 4409 spin_unlock_irqrestore(&sbi->error_lock, flags); 4410 } 4411 4412 static void f2fs_record_stop_reason(struct f2fs_sb_info *sbi) 4413 { 4414 struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi); 4415 unsigned long flags; 4416 int err; 4417 4418 f2fs_down_write(&sbi->sb_lock); 4419 4420 spin_lock_irqsave(&sbi->error_lock, flags); 4421 if (sbi->error_dirty) { 4422 memcpy(F2FS_RAW_SUPER(sbi)->s_errors, sbi->errors, 4423 MAX_F2FS_ERRORS); 4424 sbi->error_dirty = false; 4425 } 4426 memcpy(raw_super->s_stop_reason, sbi->stop_reason, MAX_STOP_REASON); 4427 spin_unlock_irqrestore(&sbi->error_lock, flags); 4428 4429 err = f2fs_commit_super(sbi, false); 4430 4431 f2fs_up_write(&sbi->sb_lock); 4432 if (err) 4433 f2fs_err_ratelimited(sbi, 4434 "f2fs_commit_super fails to record stop_reason, err:%d", 4435 err); 4436 } 4437 4438 void f2fs_save_errors(struct f2fs_sb_info *sbi, unsigned char flag) 4439 { 4440 unsigned long flags; 4441 4442 spin_lock_irqsave(&sbi->error_lock, flags); 4443 if (!test_bit(flag, (unsigned long *)sbi->errors)) { 4444 set_bit(flag, (unsigned long *)sbi->errors); 4445 sbi->error_dirty = true; 4446 } 4447 spin_unlock_irqrestore(&sbi->error_lock, flags); 4448 } 4449 4450 static bool f2fs_update_errors(struct f2fs_sb_info *sbi) 4451 { 4452 unsigned long flags; 4453 bool need_update = false; 4454 4455 spin_lock_irqsave(&sbi->error_lock, flags); 4456 if (sbi->error_dirty) { 4457 memcpy(F2FS_RAW_SUPER(sbi)->s_errors, sbi->errors, 4458 MAX_F2FS_ERRORS); 4459 sbi->error_dirty = false; 4460 need_update = true; 4461 } 4462 spin_unlock_irqrestore(&sbi->error_lock, flags); 4463 4464 return need_update; 4465 } 4466 4467 static void f2fs_record_errors(struct f2fs_sb_info *sbi, unsigned char error) 4468 { 4469 int err; 4470 4471 f2fs_down_write(&sbi->sb_lock); 4472 4473 if (!f2fs_update_errors(sbi)) 4474 goto out_unlock; 4475 4476 err = f2fs_commit_super(sbi, false); 4477 if (err) 4478 f2fs_err_ratelimited(sbi, 4479 "f2fs_commit_super fails to record errors:%u, err:%d", 4480 error, err); 4481 out_unlock: 4482 f2fs_up_write(&sbi->sb_lock); 4483 } 4484 4485 void f2fs_handle_error(struct f2fs_sb_info *sbi, unsigned char error) 4486 { 4487 f2fs_save_errors(sbi, error); 4488 f2fs_record_errors(sbi, error); 4489 } 4490 4491 void f2fs_handle_error_async(struct f2fs_sb_info *sbi, unsigned char error) 4492 { 4493 f2fs_save_errors(sbi, error); 4494 4495 if (!sbi->error_dirty) 4496 return; 4497 if (!test_bit(error, (unsigned long *)sbi->errors)) 4498 return; 4499 schedule_work(&sbi->s_error_work); 4500 } 4501 4502 static bool system_going_down(void) 4503 { 4504 return system_state == SYSTEM_HALT || system_state == SYSTEM_POWER_OFF 4505 || system_state == SYSTEM_RESTART; 4506 } 4507 4508 void f2fs_handle_critical_error(struct f2fs_sb_info *sbi, unsigned char reason) 4509 { 4510 struct super_block *sb = sbi->sb; 4511 bool shutdown = reason == STOP_CP_REASON_SHUTDOWN; 4512 bool continue_fs = !shutdown && 4513 F2FS_OPTION(sbi).errors == MOUNT_ERRORS_CONTINUE; 4514 4515 set_ckpt_flags(sbi, CP_ERROR_FLAG); 4516 4517 if (!f2fs_hw_is_readonly(sbi)) { 4518 save_stop_reason(sbi, reason); 4519 4520 /* 4521 * always create an asynchronous task to record stop_reason 4522 * in order to avoid potential deadlock when running into 4523 * f2fs_record_stop_reason() synchronously. 4524 */ 4525 schedule_work(&sbi->s_error_work); 4526 } 4527 4528 /* 4529 * We force ERRORS_RO behavior when system is rebooting. Otherwise we 4530 * could panic during 'reboot -f' as the underlying device got already 4531 * disabled. 4532 */ 4533 if (F2FS_OPTION(sbi).errors == MOUNT_ERRORS_PANIC && 4534 !shutdown && !system_going_down() && 4535 !is_sbi_flag_set(sbi, SBI_IS_SHUTDOWN)) 4536 panic("F2FS-fs (device %s): panic forced after error\n", 4537 sb->s_id); 4538 4539 if (shutdown) 4540 set_sbi_flag(sbi, SBI_IS_SHUTDOWN); 4541 else 4542 dump_stack(); 4543 4544 /* 4545 * Continue filesystem operators if errors=continue. Should not set 4546 * RO by shutdown, since RO bypasses thaw_super which can hang the 4547 * system. 4548 */ 4549 if (continue_fs || f2fs_readonly(sb) || shutdown) { 4550 f2fs_warn(sbi, "Stopped filesystem due to reason: %d", reason); 4551 return; 4552 } 4553 4554 f2fs_warn(sbi, "Remounting filesystem read-only"); 4555 4556 /* 4557 * We have already set CP_ERROR_FLAG flag to stop all updates 4558 * to filesystem, so it doesn't need to set SB_RDONLY flag here 4559 * because the flag should be set covered w/ sb->s_umount semaphore 4560 * via remount procedure, otherwise, it will confuse code like 4561 * freeze_super() which will lead to deadlocks and other problems. 4562 */ 4563 } 4564 4565 static void f2fs_record_error_work(struct work_struct *work) 4566 { 4567 struct f2fs_sb_info *sbi = container_of(work, 4568 struct f2fs_sb_info, s_error_work); 4569 4570 f2fs_record_stop_reason(sbi); 4571 } 4572 4573 static inline unsigned int get_first_seq_zone_segno(struct f2fs_sb_info *sbi) 4574 { 4575 #ifdef CONFIG_BLK_DEV_ZONED 4576 unsigned int zoneno, total_zones; 4577 int devi; 4578 4579 if (!f2fs_sb_has_blkzoned(sbi)) 4580 return NULL_SEGNO; 4581 4582 for (devi = 0; devi < sbi->s_ndevs; devi++) { 4583 if (!bdev_is_zoned(FDEV(devi).bdev)) 4584 continue; 4585 4586 total_zones = GET_ZONE_FROM_SEG(sbi, FDEV(devi).total_segments); 4587 4588 for (zoneno = 0; zoneno < total_zones; zoneno++) { 4589 unsigned int segs, blks; 4590 4591 if (!f2fs_zone_is_seq(sbi, devi, zoneno)) 4592 continue; 4593 4594 segs = GET_SEG_FROM_SEC(sbi, 4595 zoneno * sbi->secs_per_zone); 4596 blks = SEGS_TO_BLKS(sbi, segs); 4597 return GET_SEGNO(sbi, FDEV(devi).start_blk + blks); 4598 } 4599 } 4600 #endif 4601 return NULL_SEGNO; 4602 } 4603 4604 static int f2fs_scan_devices(struct f2fs_sb_info *sbi) 4605 { 4606 struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi); 4607 unsigned int max_devices = MAX_DEVICES; 4608 unsigned int logical_blksize; 4609 blk_mode_t mode = sb_open_mode(sbi->sb->s_flags); 4610 int i; 4611 4612 /* Initialize single device information */ 4613 if (!RDEV(0).path[0]) { 4614 if (!bdev_is_zoned(sbi->sb->s_bdev)) 4615 return 0; 4616 max_devices = 1; 4617 } 4618 4619 /* 4620 * Initialize multiple devices information, or single 4621 * zoned block device information. 4622 */ 4623 sbi->devs = f2fs_kzalloc(sbi, 4624 array_size(max_devices, 4625 sizeof(struct f2fs_dev_info)), 4626 GFP_KERNEL); 4627 if (!sbi->devs) 4628 return -ENOMEM; 4629 4630 logical_blksize = bdev_logical_block_size(sbi->sb->s_bdev); 4631 sbi->aligned_blksize = true; 4632 #ifdef CONFIG_BLK_DEV_ZONED 4633 sbi->max_open_zones = UINT_MAX; 4634 sbi->blkzone_alloc_policy = BLKZONE_ALLOC_PRIOR_SEQ; 4635 #endif 4636 4637 for (i = 0; i < max_devices; i++) { 4638 if (max_devices == 1) { 4639 FDEV(i).total_segments = 4640 le32_to_cpu(raw_super->segment_count_main); 4641 FDEV(i).start_blk = 0; 4642 FDEV(i).end_blk = FDEV(i).total_segments * 4643 BLKS_PER_SEG(sbi); 4644 } 4645 4646 if (i == 0) 4647 FDEV(0).bdev_file = sbi->sb->s_bdev_file; 4648 else if (!RDEV(i).path[0]) 4649 break; 4650 4651 if (max_devices > 1) { 4652 /* Multi-device mount */ 4653 memcpy(FDEV(i).path, RDEV(i).path, MAX_PATH_LEN); 4654 FDEV(i).total_segments = 4655 le32_to_cpu(RDEV(i).total_segments); 4656 if (i == 0) { 4657 FDEV(i).start_blk = 0; 4658 FDEV(i).end_blk = FDEV(i).start_blk + 4659 SEGS_TO_BLKS(sbi, 4660 FDEV(i).total_segments) - 1 + 4661 le32_to_cpu(raw_super->segment0_blkaddr); 4662 } else { 4663 FDEV(i).start_blk = FDEV(i - 1).end_blk + 1; 4664 FDEV(i).end_blk = FDEV(i).start_blk + 4665 SEGS_TO_BLKS(sbi, 4666 FDEV(i).total_segments) - 1; 4667 FDEV(i).bdev_file = bdev_file_open_by_path( 4668 FDEV(i).path, mode, sbi->sb, NULL); 4669 } 4670 } 4671 if (IS_ERR(FDEV(i).bdev_file)) 4672 return PTR_ERR(FDEV(i).bdev_file); 4673 4674 FDEV(i).bdev = file_bdev(FDEV(i).bdev_file); 4675 /* to release errored devices */ 4676 sbi->s_ndevs = i + 1; 4677 4678 if (logical_blksize != bdev_logical_block_size(FDEV(i).bdev)) 4679 sbi->aligned_blksize = false; 4680 4681 #ifdef CONFIG_BLK_DEV_ZONED 4682 if (bdev_is_zoned(FDEV(i).bdev)) { 4683 if (!f2fs_sb_has_blkzoned(sbi)) { 4684 f2fs_err(sbi, "Zoned block device feature not enabled"); 4685 return -EINVAL; 4686 } 4687 if (init_blkz_info(sbi, i)) { 4688 f2fs_err(sbi, "Failed to initialize F2FS blkzone information"); 4689 return -EINVAL; 4690 } 4691 if (max_devices == 1) 4692 break; 4693 f2fs_info(sbi, "Mount Device [%2d]: %20s, %8u, %8x - %8x (zone: Host-managed)", 4694 i, FDEV(i).path, 4695 FDEV(i).total_segments, 4696 FDEV(i).start_blk, FDEV(i).end_blk); 4697 continue; 4698 } 4699 #endif 4700 f2fs_info(sbi, "Mount Device [%2d]: %20s, %8u, %8x - %8x", 4701 i, FDEV(i).path, 4702 FDEV(i).total_segments, 4703 FDEV(i).start_blk, FDEV(i).end_blk); 4704 } 4705 return 0; 4706 } 4707 4708 static int f2fs_setup_casefold(struct f2fs_sb_info *sbi) 4709 { 4710 #if IS_ENABLED(CONFIG_UNICODE) 4711 if (f2fs_sb_has_casefold(sbi) && !sbi->sb->s_encoding) { 4712 const struct f2fs_sb_encodings *encoding_info; 4713 struct unicode_map *encoding; 4714 __u16 encoding_flags; 4715 4716 encoding_info = f2fs_sb_read_encoding(sbi->raw_super); 4717 if (!encoding_info) { 4718 f2fs_err(sbi, 4719 "Encoding requested by superblock is unknown"); 4720 return -EINVAL; 4721 } 4722 4723 encoding_flags = le16_to_cpu(sbi->raw_super->s_encoding_flags); 4724 encoding = utf8_load(encoding_info->version); 4725 if (IS_ERR(encoding)) { 4726 f2fs_err(sbi, 4727 "can't mount with superblock charset: %s-%u.%u.%u " 4728 "not supported by the kernel. flags: 0x%x.", 4729 encoding_info->name, 4730 unicode_major(encoding_info->version), 4731 unicode_minor(encoding_info->version), 4732 unicode_rev(encoding_info->version), 4733 encoding_flags); 4734 return PTR_ERR(encoding); 4735 } 4736 f2fs_info(sbi, "Using encoding defined by superblock: " 4737 "%s-%u.%u.%u with flags 0x%hx", encoding_info->name, 4738 unicode_major(encoding_info->version), 4739 unicode_minor(encoding_info->version), 4740 unicode_rev(encoding_info->version), 4741 encoding_flags); 4742 4743 sbi->sb->s_encoding = encoding; 4744 sbi->sb->s_encoding_flags = encoding_flags; 4745 } 4746 #else 4747 if (f2fs_sb_has_casefold(sbi)) { 4748 f2fs_err(sbi, "Filesystem with casefold feature cannot be mounted without CONFIG_UNICODE"); 4749 return -EINVAL; 4750 } 4751 #endif 4752 return 0; 4753 } 4754 4755 static void f2fs_tuning_parameters(struct f2fs_sb_info *sbi) 4756 { 4757 /* adjust parameters according to the volume size */ 4758 if (MAIN_SEGS(sbi) <= SMALL_VOLUME_SEGMENTS) { 4759 if (f2fs_block_unit_discard(sbi)) 4760 SM_I(sbi)->dcc_info->discard_granularity = 4761 MIN_DISCARD_GRANULARITY; 4762 if (!f2fs_lfs_mode(sbi)) 4763 SM_I(sbi)->ipu_policy = BIT(F2FS_IPU_FORCE) | 4764 BIT(F2FS_IPU_HONOR_OPU_WRITE); 4765 } 4766 4767 sbi->readdir_ra = true; 4768 } 4769 4770 static int f2fs_fill_super(struct super_block *sb, struct fs_context *fc) 4771 { 4772 struct f2fs_fs_context *ctx = fc->fs_private; 4773 struct f2fs_sb_info *sbi; 4774 struct f2fs_super_block *raw_super; 4775 struct inode *root; 4776 int err; 4777 bool skip_recovery = false, need_fsck = false; 4778 int recovery, i, valid_super_block; 4779 struct curseg_info *seg_i; 4780 int retry_cnt = 1; 4781 #ifdef CONFIG_QUOTA 4782 bool quota_enabled = false; 4783 #endif 4784 4785 try_onemore: 4786 err = -EINVAL; 4787 raw_super = NULL; 4788 valid_super_block = -1; 4789 recovery = 0; 4790 4791 /* allocate memory for f2fs-specific super block info */ 4792 sbi = kzalloc(sizeof(struct f2fs_sb_info), GFP_KERNEL); 4793 if (!sbi) 4794 return -ENOMEM; 4795 4796 sbi->sb = sb; 4797 4798 /* initialize locks within allocated memory */ 4799 init_f2fs_rwsem(&sbi->gc_lock); 4800 mutex_init(&sbi->writepages); 4801 init_f2fs_rwsem(&sbi->cp_global_sem); 4802 init_f2fs_rwsem(&sbi->node_write); 4803 init_f2fs_rwsem(&sbi->node_change); 4804 spin_lock_init(&sbi->stat_lock); 4805 init_f2fs_rwsem(&sbi->cp_rwsem); 4806 init_f2fs_rwsem(&sbi->quota_sem); 4807 init_waitqueue_head(&sbi->cp_wait); 4808 spin_lock_init(&sbi->error_lock); 4809 4810 for (i = 0; i < NR_INODE_TYPE; i++) { 4811 INIT_LIST_HEAD(&sbi->inode_list[i]); 4812 spin_lock_init(&sbi->inode_lock[i]); 4813 } 4814 mutex_init(&sbi->flush_lock); 4815 4816 /* set a block size */ 4817 if (unlikely(!sb_set_blocksize(sb, F2FS_BLKSIZE))) { 4818 f2fs_err(sbi, "unable to set blocksize"); 4819 goto free_sbi; 4820 } 4821 4822 err = read_raw_super_block(sbi, &raw_super, &valid_super_block, 4823 &recovery); 4824 if (err) 4825 goto free_sbi; 4826 4827 sb->s_fs_info = sbi; 4828 sbi->raw_super = raw_super; 4829 4830 INIT_WORK(&sbi->s_error_work, f2fs_record_error_work); 4831 memcpy(sbi->errors, raw_super->s_errors, MAX_F2FS_ERRORS); 4832 memcpy(sbi->stop_reason, raw_super->s_stop_reason, MAX_STOP_REASON); 4833 4834 /* precompute checksum seed for metadata */ 4835 if (f2fs_sb_has_inode_chksum(sbi)) 4836 sbi->s_chksum_seed = f2fs_chksum(~0, raw_super->uuid, 4837 sizeof(raw_super->uuid)); 4838 4839 default_options(sbi, false); 4840 4841 err = f2fs_check_opt_consistency(fc, sb); 4842 if (err) 4843 goto free_sb_buf; 4844 4845 f2fs_apply_options(fc, sb); 4846 4847 err = f2fs_sanity_check_options(sbi, false); 4848 if (err) 4849 goto free_options; 4850 4851 sb->s_maxbytes = max_file_blocks(NULL) << 4852 le32_to_cpu(raw_super->log_blocksize); 4853 sb->s_max_links = F2FS_LINK_MAX; 4854 4855 err = f2fs_setup_casefold(sbi); 4856 if (err) 4857 goto free_options; 4858 4859 #ifdef CONFIG_QUOTA 4860 sb->dq_op = &f2fs_quota_operations; 4861 sb->s_qcop = &f2fs_quotactl_ops; 4862 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ; 4863 4864 if (f2fs_sb_has_quota_ino(sbi)) { 4865 for (i = 0; i < MAXQUOTAS; i++) { 4866 if (f2fs_qf_ino(sbi->sb, i)) 4867 sbi->nquota_files++; 4868 } 4869 } 4870 #endif 4871 4872 sb->s_op = &f2fs_sops; 4873 #ifdef CONFIG_FS_ENCRYPTION 4874 sb->s_cop = &f2fs_cryptops; 4875 #endif 4876 #ifdef CONFIG_FS_VERITY 4877 sb->s_vop = &f2fs_verityops; 4878 #endif 4879 sb->s_xattr = f2fs_xattr_handlers; 4880 sb->s_export_op = &f2fs_export_ops; 4881 sb->s_magic = F2FS_SUPER_MAGIC; 4882 sb->s_time_gran = 1; 4883 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) | 4884 (test_opt(sbi, POSIX_ACL) ? SB_POSIXACL : 0); 4885 if (test_opt(sbi, INLINECRYPT)) 4886 sb->s_flags |= SB_INLINECRYPT; 4887 4888 if (test_opt(sbi, LAZYTIME)) 4889 sb->s_flags |= SB_LAZYTIME; 4890 else 4891 sb->s_flags &= ~SB_LAZYTIME; 4892 4893 super_set_uuid(sb, (void *) raw_super->uuid, sizeof(raw_super->uuid)); 4894 super_set_sysfs_name_bdev(sb); 4895 sb->s_iflags |= SB_I_CGROUPWB; 4896 4897 /* init f2fs-specific super block info */ 4898 sbi->valid_super_block = valid_super_block; 4899 4900 /* disallow all the data/node/meta page writes */ 4901 set_sbi_flag(sbi, SBI_POR_DOING); 4902 4903 err = f2fs_init_write_merge_io(sbi); 4904 if (err) 4905 goto free_bio_info; 4906 4907 init_sb_info(sbi); 4908 4909 err = f2fs_init_iostat(sbi); 4910 if (err) 4911 goto free_bio_info; 4912 4913 err = init_percpu_info(sbi); 4914 if (err) 4915 goto free_iostat; 4916 4917 /* init per sbi slab cache */ 4918 err = f2fs_init_xattr_caches(sbi); 4919 if (err) 4920 goto free_percpu; 4921 err = f2fs_init_page_array_cache(sbi); 4922 if (err) 4923 goto free_xattr_cache; 4924 4925 /* get an inode for meta space */ 4926 sbi->meta_inode = f2fs_iget(sb, F2FS_META_INO(sbi)); 4927 if (IS_ERR(sbi->meta_inode)) { 4928 f2fs_err(sbi, "Failed to read F2FS meta data inode"); 4929 err = PTR_ERR(sbi->meta_inode); 4930 goto free_page_array_cache; 4931 } 4932 4933 err = f2fs_get_valid_checkpoint(sbi); 4934 if (err) { 4935 f2fs_err(sbi, "Failed to get valid F2FS checkpoint"); 4936 goto free_meta_inode; 4937 } 4938 4939 if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_QUOTA_NEED_FSCK_FLAG)) 4940 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR); 4941 if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_DISABLED_QUICK_FLAG)) { 4942 set_sbi_flag(sbi, SBI_CP_DISABLED_QUICK); 4943 sbi->interval_time[DISABLE_TIME] = DEF_DISABLE_QUICK_INTERVAL; 4944 } 4945 4946 if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_FSCK_FLAG)) 4947 set_sbi_flag(sbi, SBI_NEED_FSCK); 4948 4949 /* Initialize device list */ 4950 err = f2fs_scan_devices(sbi); 4951 if (err) { 4952 f2fs_err(sbi, "Failed to find devices"); 4953 goto free_devices; 4954 } 4955 4956 err = f2fs_init_post_read_wq(sbi); 4957 if (err) { 4958 f2fs_err(sbi, "Failed to initialize post read workqueue"); 4959 goto free_devices; 4960 } 4961 4962 sbi->total_valid_node_count = 4963 le32_to_cpu(sbi->ckpt->valid_node_count); 4964 percpu_counter_set(&sbi->total_valid_inode_count, 4965 le32_to_cpu(sbi->ckpt->valid_inode_count)); 4966 sbi->user_block_count = le64_to_cpu(sbi->ckpt->user_block_count); 4967 sbi->total_valid_block_count = 4968 le64_to_cpu(sbi->ckpt->valid_block_count); 4969 sbi->last_valid_block_count = sbi->total_valid_block_count; 4970 sbi->reserved_blocks = 0; 4971 sbi->current_reserved_blocks = 0; 4972 limit_reserve_root(sbi); 4973 adjust_unusable_cap_perc(sbi); 4974 4975 f2fs_init_extent_cache_info(sbi); 4976 4977 f2fs_init_ino_entry_info(sbi); 4978 4979 f2fs_init_fsync_node_info(sbi); 4980 4981 /* setup checkpoint request control and start checkpoint issue thread */ 4982 f2fs_init_ckpt_req_control(sbi); 4983 if (!f2fs_readonly(sb) && !test_opt(sbi, DISABLE_CHECKPOINT) && 4984 test_opt(sbi, MERGE_CHECKPOINT)) { 4985 err = f2fs_start_ckpt_thread(sbi); 4986 if (err) { 4987 f2fs_err(sbi, 4988 "Failed to start F2FS issue_checkpoint_thread (%d)", 4989 err); 4990 goto stop_ckpt_thread; 4991 } 4992 } 4993 4994 /* setup f2fs internal modules */ 4995 err = f2fs_build_segment_manager(sbi); 4996 if (err) { 4997 f2fs_err(sbi, "Failed to initialize F2FS segment manager (%d)", 4998 err); 4999 goto free_sm; 5000 } 5001 err = f2fs_build_node_manager(sbi); 5002 if (err) { 5003 f2fs_err(sbi, "Failed to initialize F2FS node manager (%d)", 5004 err); 5005 goto free_nm; 5006 } 5007 5008 /* For write statistics */ 5009 sbi->sectors_written_start = f2fs_get_sectors_written(sbi); 5010 5011 /* get segno of first zoned block device */ 5012 sbi->first_seq_zone_segno = get_first_seq_zone_segno(sbi); 5013 5014 sbi->reserved_pin_section = f2fs_sb_has_blkzoned(sbi) ? 5015 ZONED_PIN_SEC_REQUIRED_COUNT : 5016 GET_SEC_FROM_SEG(sbi, overprovision_segments(sbi)); 5017 5018 /* Read accumulated write IO statistics if exists */ 5019 seg_i = CURSEG_I(sbi, CURSEG_HOT_NODE); 5020 if (__exist_node_summaries(sbi)) 5021 sbi->kbytes_written = 5022 le64_to_cpu(seg_i->journal->info.kbytes_written); 5023 5024 f2fs_build_gc_manager(sbi); 5025 5026 err = f2fs_build_stats(sbi); 5027 if (err) 5028 goto free_nm; 5029 5030 /* get an inode for node space */ 5031 sbi->node_inode = f2fs_iget(sb, F2FS_NODE_INO(sbi)); 5032 if (IS_ERR(sbi->node_inode)) { 5033 f2fs_err(sbi, "Failed to read node inode"); 5034 err = PTR_ERR(sbi->node_inode); 5035 goto free_stats; 5036 } 5037 5038 /* read root inode and dentry */ 5039 root = f2fs_iget(sb, F2FS_ROOT_INO(sbi)); 5040 if (IS_ERR(root)) { 5041 f2fs_err(sbi, "Failed to read root inode"); 5042 err = PTR_ERR(root); 5043 goto free_node_inode; 5044 } 5045 if (!S_ISDIR(root->i_mode) || !root->i_blocks || 5046 !root->i_size || !root->i_nlink) { 5047 iput(root); 5048 err = -EINVAL; 5049 goto free_node_inode; 5050 } 5051 5052 generic_set_sb_d_ops(sb); 5053 sb->s_root = d_make_root(root); /* allocate root dentry */ 5054 if (!sb->s_root) { 5055 err = -ENOMEM; 5056 goto free_node_inode; 5057 } 5058 5059 err = f2fs_init_compress_inode(sbi); 5060 if (err) 5061 goto free_root_inode; 5062 5063 err = f2fs_register_sysfs(sbi); 5064 if (err) 5065 goto free_compress_inode; 5066 5067 sbi->umount_lock_holder = current; 5068 #ifdef CONFIG_QUOTA 5069 /* Enable quota usage during mount */ 5070 if (f2fs_sb_has_quota_ino(sbi) && !f2fs_readonly(sb)) { 5071 err = f2fs_enable_quotas(sb); 5072 if (err) 5073 f2fs_err(sbi, "Cannot turn on quotas: error %d", err); 5074 } 5075 5076 quota_enabled = f2fs_recover_quota_begin(sbi); 5077 #endif 5078 /* if there are any orphan inodes, free them */ 5079 err = f2fs_recover_orphan_inodes(sbi); 5080 if (err) 5081 goto free_meta; 5082 5083 if (unlikely(is_set_ckpt_flags(sbi, CP_DISABLED_FLAG))) { 5084 skip_recovery = true; 5085 goto reset_checkpoint; 5086 } 5087 5088 /* recover fsynced data */ 5089 if (!test_opt(sbi, DISABLE_ROLL_FORWARD) && 5090 !test_opt(sbi, NORECOVERY)) { 5091 /* 5092 * mount should be failed, when device has readonly mode, and 5093 * previous checkpoint was not done by clean system shutdown. 5094 */ 5095 if (f2fs_hw_is_readonly(sbi)) { 5096 if (!is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG)) { 5097 err = f2fs_recover_fsync_data(sbi, true); 5098 if (err > 0) { 5099 err = -EROFS; 5100 f2fs_err(sbi, "Need to recover fsync data, but " 5101 "write access unavailable, please try " 5102 "mount w/ disable_roll_forward or norecovery"); 5103 } 5104 if (err < 0) 5105 goto free_meta; 5106 } 5107 f2fs_info(sbi, "write access unavailable, skipping recovery"); 5108 goto reset_checkpoint; 5109 } 5110 5111 if (need_fsck) 5112 set_sbi_flag(sbi, SBI_NEED_FSCK); 5113 5114 if (skip_recovery) 5115 goto reset_checkpoint; 5116 5117 err = f2fs_recover_fsync_data(sbi, false); 5118 if (err < 0) { 5119 if (err != -ENOMEM) 5120 skip_recovery = true; 5121 need_fsck = true; 5122 f2fs_err(sbi, "Cannot recover all fsync data errno=%d", 5123 err); 5124 goto free_meta; 5125 } 5126 } else { 5127 err = f2fs_recover_fsync_data(sbi, true); 5128 5129 if (!f2fs_readonly(sb) && err > 0) { 5130 err = -EINVAL; 5131 f2fs_err(sbi, "Need to recover fsync data"); 5132 goto free_meta; 5133 } 5134 } 5135 5136 reset_checkpoint: 5137 #ifdef CONFIG_QUOTA 5138 f2fs_recover_quota_end(sbi, quota_enabled); 5139 #endif 5140 /* 5141 * If the f2fs is not readonly and fsync data recovery succeeds, 5142 * write pointer consistency of cursegs and other zones are already 5143 * checked and fixed during recovery. However, if recovery fails, 5144 * write pointers are left untouched, and retry-mount should check 5145 * them here. 5146 */ 5147 if (skip_recovery) 5148 err = f2fs_check_and_fix_write_pointer(sbi); 5149 if (err) 5150 goto free_meta; 5151 5152 /* f2fs_recover_fsync_data() cleared this already */ 5153 clear_sbi_flag(sbi, SBI_POR_DOING); 5154 5155 err = f2fs_init_inmem_curseg(sbi); 5156 if (err) 5157 goto sync_free_meta; 5158 5159 if (test_opt(sbi, DISABLE_CHECKPOINT)) { 5160 err = f2fs_disable_checkpoint(sbi); 5161 if (err) 5162 goto sync_free_meta; 5163 } else if (is_set_ckpt_flags(sbi, CP_DISABLED_FLAG)) { 5164 f2fs_enable_checkpoint(sbi); 5165 } 5166 5167 /* 5168 * If filesystem is not mounted as read-only then 5169 * do start the gc_thread. 5170 */ 5171 if ((F2FS_OPTION(sbi).bggc_mode != BGGC_MODE_OFF || 5172 test_opt(sbi, GC_MERGE)) && !f2fs_readonly(sb)) { 5173 /* After POR, we can run background GC thread.*/ 5174 err = f2fs_start_gc_thread(sbi); 5175 if (err) 5176 goto sync_free_meta; 5177 } 5178 5179 /* recover broken superblock */ 5180 if (recovery) { 5181 err = f2fs_commit_super(sbi, true); 5182 f2fs_info(sbi, "Try to recover %dth superblock, ret: %d", 5183 sbi->valid_super_block ? 1 : 2, err); 5184 } 5185 5186 f2fs_join_shrinker(sbi); 5187 5188 f2fs_tuning_parameters(sbi); 5189 5190 f2fs_notice(sbi, "Mounted with checkpoint version = %llx", 5191 cur_cp_version(F2FS_CKPT(sbi))); 5192 f2fs_update_time(sbi, CP_TIME); 5193 f2fs_update_time(sbi, REQ_TIME); 5194 clear_sbi_flag(sbi, SBI_CP_DISABLED_QUICK); 5195 5196 sbi->umount_lock_holder = NULL; 5197 return 0; 5198 5199 sync_free_meta: 5200 /* safe to flush all the data */ 5201 sync_filesystem(sbi->sb); 5202 retry_cnt = 0; 5203 5204 free_meta: 5205 #ifdef CONFIG_QUOTA 5206 f2fs_truncate_quota_inode_pages(sb); 5207 if (f2fs_sb_has_quota_ino(sbi) && !f2fs_readonly(sb)) 5208 f2fs_quota_off_umount(sbi->sb); 5209 #endif 5210 /* 5211 * Some dirty meta pages can be produced by f2fs_recover_orphan_inodes() 5212 * failed by EIO. Then, iput(node_inode) can trigger balance_fs_bg() 5213 * followed by f2fs_write_checkpoint() through f2fs_write_node_pages(), which 5214 * falls into an infinite loop in f2fs_sync_meta_pages(). 5215 */ 5216 truncate_inode_pages_final(META_MAPPING(sbi)); 5217 /* evict some inodes being cached by GC */ 5218 evict_inodes(sb); 5219 f2fs_unregister_sysfs(sbi); 5220 free_compress_inode: 5221 f2fs_destroy_compress_inode(sbi); 5222 free_root_inode: 5223 dput(sb->s_root); 5224 sb->s_root = NULL; 5225 free_node_inode: 5226 f2fs_release_ino_entry(sbi, true); 5227 truncate_inode_pages_final(NODE_MAPPING(sbi)); 5228 iput(sbi->node_inode); 5229 sbi->node_inode = NULL; 5230 free_stats: 5231 f2fs_destroy_stats(sbi); 5232 free_nm: 5233 /* stop discard thread before destroying node manager */ 5234 f2fs_stop_discard_thread(sbi); 5235 f2fs_destroy_node_manager(sbi); 5236 free_sm: 5237 f2fs_destroy_segment_manager(sbi); 5238 stop_ckpt_thread: 5239 f2fs_stop_ckpt_thread(sbi); 5240 /* flush s_error_work before sbi destroy */ 5241 flush_work(&sbi->s_error_work); 5242 f2fs_destroy_post_read_wq(sbi); 5243 free_devices: 5244 destroy_device_list(sbi); 5245 kvfree(sbi->ckpt); 5246 free_meta_inode: 5247 make_bad_inode(sbi->meta_inode); 5248 iput(sbi->meta_inode); 5249 sbi->meta_inode = NULL; 5250 free_page_array_cache: 5251 f2fs_destroy_page_array_cache(sbi); 5252 free_xattr_cache: 5253 f2fs_destroy_xattr_caches(sbi); 5254 free_percpu: 5255 destroy_percpu_info(sbi); 5256 free_iostat: 5257 f2fs_destroy_iostat(sbi); 5258 free_bio_info: 5259 for (i = 0; i < NR_PAGE_TYPE; i++) 5260 kfree(sbi->write_io[i]); 5261 5262 #if IS_ENABLED(CONFIG_UNICODE) 5263 utf8_unload(sb->s_encoding); 5264 sb->s_encoding = NULL; 5265 #endif 5266 free_options: 5267 #ifdef CONFIG_QUOTA 5268 for (i = 0; i < MAXQUOTAS; i++) 5269 kfree(F2FS_OPTION(sbi).s_qf_names[i]); 5270 #endif 5271 /* no need to free dummy_enc_policy, we just keep it in ctx when failed */ 5272 swap(F2FS_CTX_INFO(ctx).dummy_enc_policy, F2FS_OPTION(sbi).dummy_enc_policy); 5273 free_sb_buf: 5274 kfree(raw_super); 5275 free_sbi: 5276 kfree(sbi); 5277 sb->s_fs_info = NULL; 5278 5279 /* give only one another chance */ 5280 if (retry_cnt > 0 && skip_recovery) { 5281 retry_cnt--; 5282 shrink_dcache_sb(sb); 5283 goto try_onemore; 5284 } 5285 return err; 5286 } 5287 5288 static int f2fs_get_tree(struct fs_context *fc) 5289 { 5290 return get_tree_bdev(fc, f2fs_fill_super); 5291 } 5292 5293 static int f2fs_reconfigure(struct fs_context *fc) 5294 { 5295 struct super_block *sb = fc->root->d_sb; 5296 5297 return __f2fs_remount(fc, sb); 5298 } 5299 5300 static void f2fs_fc_free(struct fs_context *fc) 5301 { 5302 struct f2fs_fs_context *ctx = fc->fs_private; 5303 5304 if (!ctx) 5305 return; 5306 5307 #ifdef CONFIG_QUOTA 5308 f2fs_unnote_qf_name_all(fc); 5309 #endif 5310 fscrypt_free_dummy_policy(&F2FS_CTX_INFO(ctx).dummy_enc_policy); 5311 kfree(ctx); 5312 } 5313 5314 static const struct fs_context_operations f2fs_context_ops = { 5315 .parse_param = f2fs_parse_param, 5316 .get_tree = f2fs_get_tree, 5317 .reconfigure = f2fs_reconfigure, 5318 .free = f2fs_fc_free, 5319 }; 5320 5321 static void kill_f2fs_super(struct super_block *sb) 5322 { 5323 struct f2fs_sb_info *sbi = F2FS_SB(sb); 5324 5325 if (sb->s_root) { 5326 sbi->umount_lock_holder = current; 5327 5328 set_sbi_flag(sbi, SBI_IS_CLOSE); 5329 f2fs_stop_gc_thread(sbi); 5330 f2fs_stop_discard_thread(sbi); 5331 5332 #ifdef CONFIG_F2FS_FS_COMPRESSION 5333 /* 5334 * latter evict_inode() can bypass checking and invalidating 5335 * compress inode cache. 5336 */ 5337 if (test_opt(sbi, COMPRESS_CACHE)) 5338 truncate_inode_pages_final(COMPRESS_MAPPING(sbi)); 5339 #endif 5340 5341 if (is_sbi_flag_set(sbi, SBI_IS_DIRTY) || 5342 !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG)) { 5343 struct cp_control cpc = { 5344 .reason = CP_UMOUNT, 5345 }; 5346 stat_inc_cp_call_count(sbi, TOTAL_CALL); 5347 f2fs_write_checkpoint(sbi, &cpc); 5348 } 5349 5350 if (is_sbi_flag_set(sbi, SBI_IS_RECOVERED) && f2fs_readonly(sb)) 5351 sb->s_flags &= ~SB_RDONLY; 5352 } 5353 kill_block_super(sb); 5354 /* Release block devices last, after fscrypt_destroy_keyring(). */ 5355 if (sbi) { 5356 destroy_device_list(sbi); 5357 kfree(sbi); 5358 sb->s_fs_info = NULL; 5359 } 5360 } 5361 5362 static int f2fs_init_fs_context(struct fs_context *fc) 5363 { 5364 struct f2fs_fs_context *ctx; 5365 5366 ctx = kzalloc(sizeof(struct f2fs_fs_context), GFP_KERNEL); 5367 if (!ctx) 5368 return -ENOMEM; 5369 5370 fc->fs_private = ctx; 5371 fc->ops = &f2fs_context_ops; 5372 5373 return 0; 5374 } 5375 5376 static struct file_system_type f2fs_fs_type = { 5377 .owner = THIS_MODULE, 5378 .name = "f2fs", 5379 .init_fs_context = f2fs_init_fs_context, 5380 .kill_sb = kill_f2fs_super, 5381 .fs_flags = FS_REQUIRES_DEV | FS_ALLOW_IDMAP, 5382 }; 5383 MODULE_ALIAS_FS("f2fs"); 5384 5385 static int __init init_inodecache(void) 5386 { 5387 f2fs_inode_cachep = kmem_cache_create("f2fs_inode_cache", 5388 sizeof(struct f2fs_inode_info), 0, 5389 SLAB_RECLAIM_ACCOUNT|SLAB_ACCOUNT, NULL); 5390 return f2fs_inode_cachep ? 0 : -ENOMEM; 5391 } 5392 5393 static void destroy_inodecache(void) 5394 { 5395 /* 5396 * Make sure all delayed rcu free inodes are flushed before we 5397 * destroy cache. 5398 */ 5399 rcu_barrier(); 5400 kmem_cache_destroy(f2fs_inode_cachep); 5401 } 5402 5403 static int __init init_f2fs_fs(void) 5404 { 5405 int err; 5406 5407 err = init_inodecache(); 5408 if (err) 5409 goto fail; 5410 err = f2fs_create_node_manager_caches(); 5411 if (err) 5412 goto free_inodecache; 5413 err = f2fs_create_segment_manager_caches(); 5414 if (err) 5415 goto free_node_manager_caches; 5416 err = f2fs_create_checkpoint_caches(); 5417 if (err) 5418 goto free_segment_manager_caches; 5419 err = f2fs_create_recovery_cache(); 5420 if (err) 5421 goto free_checkpoint_caches; 5422 err = f2fs_create_extent_cache(); 5423 if (err) 5424 goto free_recovery_cache; 5425 err = f2fs_create_garbage_collection_cache(); 5426 if (err) 5427 goto free_extent_cache; 5428 err = f2fs_init_sysfs(); 5429 if (err) 5430 goto free_garbage_collection_cache; 5431 err = f2fs_init_shrinker(); 5432 if (err) 5433 goto free_sysfs; 5434 f2fs_create_root_stats(); 5435 err = f2fs_init_post_read_processing(); 5436 if (err) 5437 goto free_root_stats; 5438 err = f2fs_init_iostat_processing(); 5439 if (err) 5440 goto free_post_read; 5441 err = f2fs_init_bio_entry_cache(); 5442 if (err) 5443 goto free_iostat; 5444 err = f2fs_init_bioset(); 5445 if (err) 5446 goto free_bio_entry_cache; 5447 err = f2fs_init_compress_mempool(); 5448 if (err) 5449 goto free_bioset; 5450 err = f2fs_init_compress_cache(); 5451 if (err) 5452 goto free_compress_mempool; 5453 err = f2fs_create_casefold_cache(); 5454 if (err) 5455 goto free_compress_cache; 5456 err = register_filesystem(&f2fs_fs_type); 5457 if (err) 5458 goto free_casefold_cache; 5459 return 0; 5460 free_casefold_cache: 5461 f2fs_destroy_casefold_cache(); 5462 free_compress_cache: 5463 f2fs_destroy_compress_cache(); 5464 free_compress_mempool: 5465 f2fs_destroy_compress_mempool(); 5466 free_bioset: 5467 f2fs_destroy_bioset(); 5468 free_bio_entry_cache: 5469 f2fs_destroy_bio_entry_cache(); 5470 free_iostat: 5471 f2fs_destroy_iostat_processing(); 5472 free_post_read: 5473 f2fs_destroy_post_read_processing(); 5474 free_root_stats: 5475 f2fs_destroy_root_stats(); 5476 f2fs_exit_shrinker(); 5477 free_sysfs: 5478 f2fs_exit_sysfs(); 5479 free_garbage_collection_cache: 5480 f2fs_destroy_garbage_collection_cache(); 5481 free_extent_cache: 5482 f2fs_destroy_extent_cache(); 5483 free_recovery_cache: 5484 f2fs_destroy_recovery_cache(); 5485 free_checkpoint_caches: 5486 f2fs_destroy_checkpoint_caches(); 5487 free_segment_manager_caches: 5488 f2fs_destroy_segment_manager_caches(); 5489 free_node_manager_caches: 5490 f2fs_destroy_node_manager_caches(); 5491 free_inodecache: 5492 destroy_inodecache(); 5493 fail: 5494 return err; 5495 } 5496 5497 static void __exit exit_f2fs_fs(void) 5498 { 5499 unregister_filesystem(&f2fs_fs_type); 5500 f2fs_destroy_casefold_cache(); 5501 f2fs_destroy_compress_cache(); 5502 f2fs_destroy_compress_mempool(); 5503 f2fs_destroy_bioset(); 5504 f2fs_destroy_bio_entry_cache(); 5505 f2fs_destroy_iostat_processing(); 5506 f2fs_destroy_post_read_processing(); 5507 f2fs_destroy_root_stats(); 5508 f2fs_exit_shrinker(); 5509 f2fs_exit_sysfs(); 5510 f2fs_destroy_garbage_collection_cache(); 5511 f2fs_destroy_extent_cache(); 5512 f2fs_destroy_recovery_cache(); 5513 f2fs_destroy_checkpoint_caches(); 5514 f2fs_destroy_segment_manager_caches(); 5515 f2fs_destroy_node_manager_caches(); 5516 destroy_inodecache(); 5517 } 5518 5519 module_init(init_f2fs_fs) 5520 module_exit(exit_f2fs_fs) 5521 5522 MODULE_AUTHOR("Samsung Electronics's Praesto Team"); 5523 MODULE_DESCRIPTION("Flash Friendly File System"); 5524 MODULE_LICENSE("GPL"); 5525