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/buffer_head.h> 15 #include <linux/kthread.h> 16 #include <linux/parser.h> 17 #include <linux/mount.h> 18 #include <linux/seq_file.h> 19 #include <linux/proc_fs.h> 20 #include <linux/random.h> 21 #include <linux/exportfs.h> 22 #include <linux/blkdev.h> 23 #include <linux/quotaops.h> 24 #include <linux/f2fs_fs.h> 25 #include <linux/sysfs.h> 26 #include <linux/quota.h> 27 #include <linux/unicode.h> 28 #include <linux/part_stat.h> 29 #include <linux/zstd.h> 30 #include <linux/lz4.h> 31 32 #include "f2fs.h" 33 #include "node.h" 34 #include "segment.h" 35 #include "xattr.h" 36 #include "gc.h" 37 #include "iostat.h" 38 39 #define CREATE_TRACE_POINTS 40 #include <trace/events/f2fs.h> 41 42 static struct kmem_cache *f2fs_inode_cachep; 43 44 #ifdef CONFIG_F2FS_FAULT_INJECTION 45 46 const char *f2fs_fault_name[FAULT_MAX] = { 47 [FAULT_KMALLOC] = "kmalloc", 48 [FAULT_KVMALLOC] = "kvmalloc", 49 [FAULT_PAGE_ALLOC] = "page alloc", 50 [FAULT_PAGE_GET] = "page get", 51 [FAULT_ALLOC_NID] = "alloc nid", 52 [FAULT_ORPHAN] = "orphan", 53 [FAULT_BLOCK] = "no more block", 54 [FAULT_DIR_DEPTH] = "too big dir depth", 55 [FAULT_EVICT_INODE] = "evict_inode fail", 56 [FAULT_TRUNCATE] = "truncate fail", 57 [FAULT_READ_IO] = "read IO error", 58 [FAULT_CHECKPOINT] = "checkpoint error", 59 [FAULT_DISCARD] = "discard error", 60 [FAULT_WRITE_IO] = "write IO error", 61 [FAULT_SLAB_ALLOC] = "slab alloc", 62 [FAULT_DQUOT_INIT] = "dquot initialize", 63 [FAULT_LOCK_OP] = "lock_op", 64 [FAULT_BLKADDR] = "invalid blkaddr", 65 }; 66 67 void f2fs_build_fault_attr(struct f2fs_sb_info *sbi, unsigned int rate, 68 unsigned int type) 69 { 70 struct f2fs_fault_info *ffi = &F2FS_OPTION(sbi).fault_info; 71 72 if (rate) { 73 atomic_set(&ffi->inject_ops, 0); 74 ffi->inject_rate = rate; 75 } 76 77 if (type) 78 ffi->inject_type = type; 79 80 if (!rate && !type) 81 memset(ffi, 0, sizeof(struct f2fs_fault_info)); 82 } 83 #endif 84 85 /* f2fs-wide shrinker description */ 86 static struct shrinker f2fs_shrinker_info = { 87 .scan_objects = f2fs_shrink_scan, 88 .count_objects = f2fs_shrink_count, 89 .seeks = DEFAULT_SEEKS, 90 }; 91 92 enum { 93 Opt_gc_background, 94 Opt_disable_roll_forward, 95 Opt_norecovery, 96 Opt_discard, 97 Opt_nodiscard, 98 Opt_noheap, 99 Opt_heap, 100 Opt_user_xattr, 101 Opt_nouser_xattr, 102 Opt_acl, 103 Opt_noacl, 104 Opt_active_logs, 105 Opt_disable_ext_identify, 106 Opt_inline_xattr, 107 Opt_noinline_xattr, 108 Opt_inline_xattr_size, 109 Opt_inline_data, 110 Opt_inline_dentry, 111 Opt_noinline_dentry, 112 Opt_flush_merge, 113 Opt_noflush_merge, 114 Opt_barrier, 115 Opt_nobarrier, 116 Opt_fastboot, 117 Opt_extent_cache, 118 Opt_noextent_cache, 119 Opt_noinline_data, 120 Opt_data_flush, 121 Opt_reserve_root, 122 Opt_resgid, 123 Opt_resuid, 124 Opt_mode, 125 Opt_io_size_bits, 126 Opt_fault_injection, 127 Opt_fault_type, 128 Opt_lazytime, 129 Opt_nolazytime, 130 Opt_quota, 131 Opt_noquota, 132 Opt_usrquota, 133 Opt_grpquota, 134 Opt_prjquota, 135 Opt_usrjquota, 136 Opt_grpjquota, 137 Opt_prjjquota, 138 Opt_offusrjquota, 139 Opt_offgrpjquota, 140 Opt_offprjjquota, 141 Opt_jqfmt_vfsold, 142 Opt_jqfmt_vfsv0, 143 Opt_jqfmt_vfsv1, 144 Opt_alloc, 145 Opt_fsync, 146 Opt_test_dummy_encryption, 147 Opt_inlinecrypt, 148 Opt_checkpoint_disable, 149 Opt_checkpoint_disable_cap, 150 Opt_checkpoint_disable_cap_perc, 151 Opt_checkpoint_enable, 152 Opt_checkpoint_merge, 153 Opt_nocheckpoint_merge, 154 Opt_compress_algorithm, 155 Opt_compress_log_size, 156 Opt_compress_extension, 157 Opt_nocompress_extension, 158 Opt_compress_chksum, 159 Opt_compress_mode, 160 Opt_compress_cache, 161 Opt_atgc, 162 Opt_gc_merge, 163 Opt_nogc_merge, 164 Opt_discard_unit, 165 Opt_memory_mode, 166 Opt_age_extent_cache, 167 Opt_errors, 168 Opt_err, 169 }; 170 171 static match_table_t f2fs_tokens = { 172 {Opt_gc_background, "background_gc=%s"}, 173 {Opt_disable_roll_forward, "disable_roll_forward"}, 174 {Opt_norecovery, "norecovery"}, 175 {Opt_discard, "discard"}, 176 {Opt_nodiscard, "nodiscard"}, 177 {Opt_noheap, "no_heap"}, 178 {Opt_heap, "heap"}, 179 {Opt_user_xattr, "user_xattr"}, 180 {Opt_nouser_xattr, "nouser_xattr"}, 181 {Opt_acl, "acl"}, 182 {Opt_noacl, "noacl"}, 183 {Opt_active_logs, "active_logs=%u"}, 184 {Opt_disable_ext_identify, "disable_ext_identify"}, 185 {Opt_inline_xattr, "inline_xattr"}, 186 {Opt_noinline_xattr, "noinline_xattr"}, 187 {Opt_inline_xattr_size, "inline_xattr_size=%u"}, 188 {Opt_inline_data, "inline_data"}, 189 {Opt_inline_dentry, "inline_dentry"}, 190 {Opt_noinline_dentry, "noinline_dentry"}, 191 {Opt_flush_merge, "flush_merge"}, 192 {Opt_noflush_merge, "noflush_merge"}, 193 {Opt_barrier, "barrier"}, 194 {Opt_nobarrier, "nobarrier"}, 195 {Opt_fastboot, "fastboot"}, 196 {Opt_extent_cache, "extent_cache"}, 197 {Opt_noextent_cache, "noextent_cache"}, 198 {Opt_noinline_data, "noinline_data"}, 199 {Opt_data_flush, "data_flush"}, 200 {Opt_reserve_root, "reserve_root=%u"}, 201 {Opt_resgid, "resgid=%u"}, 202 {Opt_resuid, "resuid=%u"}, 203 {Opt_mode, "mode=%s"}, 204 {Opt_io_size_bits, "io_bits=%u"}, 205 {Opt_fault_injection, "fault_injection=%u"}, 206 {Opt_fault_type, "fault_type=%u"}, 207 {Opt_lazytime, "lazytime"}, 208 {Opt_nolazytime, "nolazytime"}, 209 {Opt_quota, "quota"}, 210 {Opt_noquota, "noquota"}, 211 {Opt_usrquota, "usrquota"}, 212 {Opt_grpquota, "grpquota"}, 213 {Opt_prjquota, "prjquota"}, 214 {Opt_usrjquota, "usrjquota=%s"}, 215 {Opt_grpjquota, "grpjquota=%s"}, 216 {Opt_prjjquota, "prjjquota=%s"}, 217 {Opt_offusrjquota, "usrjquota="}, 218 {Opt_offgrpjquota, "grpjquota="}, 219 {Opt_offprjjquota, "prjjquota="}, 220 {Opt_jqfmt_vfsold, "jqfmt=vfsold"}, 221 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"}, 222 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"}, 223 {Opt_alloc, "alloc_mode=%s"}, 224 {Opt_fsync, "fsync_mode=%s"}, 225 {Opt_test_dummy_encryption, "test_dummy_encryption=%s"}, 226 {Opt_test_dummy_encryption, "test_dummy_encryption"}, 227 {Opt_inlinecrypt, "inlinecrypt"}, 228 {Opt_checkpoint_disable, "checkpoint=disable"}, 229 {Opt_checkpoint_disable_cap, "checkpoint=disable:%u"}, 230 {Opt_checkpoint_disable_cap_perc, "checkpoint=disable:%u%%"}, 231 {Opt_checkpoint_enable, "checkpoint=enable"}, 232 {Opt_checkpoint_merge, "checkpoint_merge"}, 233 {Opt_nocheckpoint_merge, "nocheckpoint_merge"}, 234 {Opt_compress_algorithm, "compress_algorithm=%s"}, 235 {Opt_compress_log_size, "compress_log_size=%u"}, 236 {Opt_compress_extension, "compress_extension=%s"}, 237 {Opt_nocompress_extension, "nocompress_extension=%s"}, 238 {Opt_compress_chksum, "compress_chksum"}, 239 {Opt_compress_mode, "compress_mode=%s"}, 240 {Opt_compress_cache, "compress_cache"}, 241 {Opt_atgc, "atgc"}, 242 {Opt_gc_merge, "gc_merge"}, 243 {Opt_nogc_merge, "nogc_merge"}, 244 {Opt_discard_unit, "discard_unit=%s"}, 245 {Opt_memory_mode, "memory=%s"}, 246 {Opt_age_extent_cache, "age_extent_cache"}, 247 {Opt_errors, "errors=%s"}, 248 {Opt_err, NULL}, 249 }; 250 251 void f2fs_printk(struct f2fs_sb_info *sbi, const char *fmt, ...) 252 { 253 struct va_format vaf; 254 va_list args; 255 int level; 256 257 va_start(args, fmt); 258 259 level = printk_get_level(fmt); 260 vaf.fmt = printk_skip_level(fmt); 261 vaf.va = &args; 262 printk("%c%cF2FS-fs (%s): %pV\n", 263 KERN_SOH_ASCII, level, sbi->sb->s_id, &vaf); 264 265 va_end(args); 266 } 267 268 #if IS_ENABLED(CONFIG_UNICODE) 269 static const struct f2fs_sb_encodings { 270 __u16 magic; 271 char *name; 272 unsigned int version; 273 } f2fs_sb_encoding_map[] = { 274 {F2FS_ENC_UTF8_12_1, "utf8", UNICODE_AGE(12, 1, 0)}, 275 }; 276 277 static const struct f2fs_sb_encodings * 278 f2fs_sb_read_encoding(const struct f2fs_super_block *sb) 279 { 280 __u16 magic = le16_to_cpu(sb->s_encoding); 281 int i; 282 283 for (i = 0; i < ARRAY_SIZE(f2fs_sb_encoding_map); i++) 284 if (magic == f2fs_sb_encoding_map[i].magic) 285 return &f2fs_sb_encoding_map[i]; 286 287 return NULL; 288 } 289 290 struct kmem_cache *f2fs_cf_name_slab; 291 static int __init f2fs_create_casefold_cache(void) 292 { 293 f2fs_cf_name_slab = f2fs_kmem_cache_create("f2fs_casefolded_name", 294 F2FS_NAME_LEN); 295 return f2fs_cf_name_slab ? 0 : -ENOMEM; 296 } 297 298 static void f2fs_destroy_casefold_cache(void) 299 { 300 kmem_cache_destroy(f2fs_cf_name_slab); 301 } 302 #else 303 static int __init f2fs_create_casefold_cache(void) { return 0; } 304 static void f2fs_destroy_casefold_cache(void) { } 305 #endif 306 307 static inline void limit_reserve_root(struct f2fs_sb_info *sbi) 308 { 309 block_t limit = min((sbi->user_block_count >> 3), 310 sbi->user_block_count - sbi->reserved_blocks); 311 312 /* limit is 12.5% */ 313 if (test_opt(sbi, RESERVE_ROOT) && 314 F2FS_OPTION(sbi).root_reserved_blocks > limit) { 315 F2FS_OPTION(sbi).root_reserved_blocks = limit; 316 f2fs_info(sbi, "Reduce reserved blocks for root = %u", 317 F2FS_OPTION(sbi).root_reserved_blocks); 318 } 319 if (!test_opt(sbi, RESERVE_ROOT) && 320 (!uid_eq(F2FS_OPTION(sbi).s_resuid, 321 make_kuid(&init_user_ns, F2FS_DEF_RESUID)) || 322 !gid_eq(F2FS_OPTION(sbi).s_resgid, 323 make_kgid(&init_user_ns, F2FS_DEF_RESGID)))) 324 f2fs_info(sbi, "Ignore s_resuid=%u, s_resgid=%u w/o reserve_root", 325 from_kuid_munged(&init_user_ns, 326 F2FS_OPTION(sbi).s_resuid), 327 from_kgid_munged(&init_user_ns, 328 F2FS_OPTION(sbi).s_resgid)); 329 } 330 331 static inline int adjust_reserved_segment(struct f2fs_sb_info *sbi) 332 { 333 unsigned int sec_blks = sbi->blocks_per_seg * sbi->segs_per_sec; 334 unsigned int avg_vblocks; 335 unsigned int wanted_reserved_segments; 336 block_t avail_user_block_count; 337 338 if (!F2FS_IO_ALIGNED(sbi)) 339 return 0; 340 341 /* average valid block count in section in worst case */ 342 avg_vblocks = sec_blks / F2FS_IO_SIZE(sbi); 343 344 /* 345 * we need enough free space when migrating one section in worst case 346 */ 347 wanted_reserved_segments = (F2FS_IO_SIZE(sbi) / avg_vblocks) * 348 reserved_segments(sbi); 349 wanted_reserved_segments -= reserved_segments(sbi); 350 351 avail_user_block_count = sbi->user_block_count - 352 sbi->current_reserved_blocks - 353 F2FS_OPTION(sbi).root_reserved_blocks; 354 355 if (wanted_reserved_segments * sbi->blocks_per_seg > 356 avail_user_block_count) { 357 f2fs_err(sbi, "IO align feature can't grab additional reserved segment: %u, available segments: %u", 358 wanted_reserved_segments, 359 avail_user_block_count >> sbi->log_blocks_per_seg); 360 return -ENOSPC; 361 } 362 363 SM_I(sbi)->additional_reserved_segments = wanted_reserved_segments; 364 365 f2fs_info(sbi, "IO align feature needs additional reserved segment: %u", 366 wanted_reserved_segments); 367 368 return 0; 369 } 370 371 static inline void adjust_unusable_cap_perc(struct f2fs_sb_info *sbi) 372 { 373 if (!F2FS_OPTION(sbi).unusable_cap_perc) 374 return; 375 376 if (F2FS_OPTION(sbi).unusable_cap_perc == 100) 377 F2FS_OPTION(sbi).unusable_cap = sbi->user_block_count; 378 else 379 F2FS_OPTION(sbi).unusable_cap = (sbi->user_block_count / 100) * 380 F2FS_OPTION(sbi).unusable_cap_perc; 381 382 f2fs_info(sbi, "Adjust unusable cap for checkpoint=disable = %u / %u%%", 383 F2FS_OPTION(sbi).unusable_cap, 384 F2FS_OPTION(sbi).unusable_cap_perc); 385 } 386 387 static void init_once(void *foo) 388 { 389 struct f2fs_inode_info *fi = (struct f2fs_inode_info *) foo; 390 391 inode_init_once(&fi->vfs_inode); 392 } 393 394 #ifdef CONFIG_QUOTA 395 static const char * const quotatypes[] = INITQFNAMES; 396 #define QTYPE2NAME(t) (quotatypes[t]) 397 static int f2fs_set_qf_name(struct super_block *sb, int qtype, 398 substring_t *args) 399 { 400 struct f2fs_sb_info *sbi = F2FS_SB(sb); 401 char *qname; 402 int ret = -EINVAL; 403 404 if (sb_any_quota_loaded(sb) && !F2FS_OPTION(sbi).s_qf_names[qtype]) { 405 f2fs_err(sbi, "Cannot change journaled quota options when quota turned on"); 406 return -EINVAL; 407 } 408 if (f2fs_sb_has_quota_ino(sbi)) { 409 f2fs_info(sbi, "QUOTA feature is enabled, so ignore qf_name"); 410 return 0; 411 } 412 413 qname = match_strdup(args); 414 if (!qname) { 415 f2fs_err(sbi, "Not enough memory for storing quotafile name"); 416 return -ENOMEM; 417 } 418 if (F2FS_OPTION(sbi).s_qf_names[qtype]) { 419 if (strcmp(F2FS_OPTION(sbi).s_qf_names[qtype], qname) == 0) 420 ret = 0; 421 else 422 f2fs_err(sbi, "%s quota file already specified", 423 QTYPE2NAME(qtype)); 424 goto errout; 425 } 426 if (strchr(qname, '/')) { 427 f2fs_err(sbi, "quotafile must be on filesystem root"); 428 goto errout; 429 } 430 F2FS_OPTION(sbi).s_qf_names[qtype] = qname; 431 set_opt(sbi, QUOTA); 432 return 0; 433 errout: 434 kfree(qname); 435 return ret; 436 } 437 438 static int f2fs_clear_qf_name(struct super_block *sb, int qtype) 439 { 440 struct f2fs_sb_info *sbi = F2FS_SB(sb); 441 442 if (sb_any_quota_loaded(sb) && F2FS_OPTION(sbi).s_qf_names[qtype]) { 443 f2fs_err(sbi, "Cannot change journaled quota options when quota turned on"); 444 return -EINVAL; 445 } 446 kfree(F2FS_OPTION(sbi).s_qf_names[qtype]); 447 F2FS_OPTION(sbi).s_qf_names[qtype] = NULL; 448 return 0; 449 } 450 451 static int f2fs_check_quota_options(struct f2fs_sb_info *sbi) 452 { 453 /* 454 * We do the test below only for project quotas. 'usrquota' and 455 * 'grpquota' mount options are allowed even without quota feature 456 * to support legacy quotas in quota files. 457 */ 458 if (test_opt(sbi, PRJQUOTA) && !f2fs_sb_has_project_quota(sbi)) { 459 f2fs_err(sbi, "Project quota feature not enabled. Cannot enable project quota enforcement."); 460 return -1; 461 } 462 if (F2FS_OPTION(sbi).s_qf_names[USRQUOTA] || 463 F2FS_OPTION(sbi).s_qf_names[GRPQUOTA] || 464 F2FS_OPTION(sbi).s_qf_names[PRJQUOTA]) { 465 if (test_opt(sbi, USRQUOTA) && 466 F2FS_OPTION(sbi).s_qf_names[USRQUOTA]) 467 clear_opt(sbi, USRQUOTA); 468 469 if (test_opt(sbi, GRPQUOTA) && 470 F2FS_OPTION(sbi).s_qf_names[GRPQUOTA]) 471 clear_opt(sbi, GRPQUOTA); 472 473 if (test_opt(sbi, PRJQUOTA) && 474 F2FS_OPTION(sbi).s_qf_names[PRJQUOTA]) 475 clear_opt(sbi, PRJQUOTA); 476 477 if (test_opt(sbi, GRPQUOTA) || test_opt(sbi, USRQUOTA) || 478 test_opt(sbi, PRJQUOTA)) { 479 f2fs_err(sbi, "old and new quota format mixing"); 480 return -1; 481 } 482 483 if (!F2FS_OPTION(sbi).s_jquota_fmt) { 484 f2fs_err(sbi, "journaled quota format not specified"); 485 return -1; 486 } 487 } 488 489 if (f2fs_sb_has_quota_ino(sbi) && F2FS_OPTION(sbi).s_jquota_fmt) { 490 f2fs_info(sbi, "QUOTA feature is enabled, so ignore jquota_fmt"); 491 F2FS_OPTION(sbi).s_jquota_fmt = 0; 492 } 493 return 0; 494 } 495 #endif 496 497 static int f2fs_set_test_dummy_encryption(struct super_block *sb, 498 const char *opt, 499 const substring_t *arg, 500 bool is_remount) 501 { 502 struct f2fs_sb_info *sbi = F2FS_SB(sb); 503 struct fs_parameter param = { 504 .type = fs_value_is_string, 505 .string = arg->from ? arg->from : "", 506 }; 507 struct fscrypt_dummy_policy *policy = 508 &F2FS_OPTION(sbi).dummy_enc_policy; 509 int err; 510 511 if (!IS_ENABLED(CONFIG_FS_ENCRYPTION)) { 512 f2fs_warn(sbi, "test_dummy_encryption option not supported"); 513 return -EINVAL; 514 } 515 516 if (!f2fs_sb_has_encrypt(sbi)) { 517 f2fs_err(sbi, "Encrypt feature is off"); 518 return -EINVAL; 519 } 520 521 /* 522 * This mount option is just for testing, and it's not worthwhile to 523 * implement the extra complexity (e.g. RCU protection) that would be 524 * needed to allow it to be set or changed during remount. We do allow 525 * it to be specified during remount, but only if there is no change. 526 */ 527 if (is_remount && !fscrypt_is_dummy_policy_set(policy)) { 528 f2fs_warn(sbi, "Can't set test_dummy_encryption on remount"); 529 return -EINVAL; 530 } 531 532 err = fscrypt_parse_test_dummy_encryption(¶m, policy); 533 if (err) { 534 if (err == -EEXIST) 535 f2fs_warn(sbi, 536 "Can't change test_dummy_encryption on remount"); 537 else if (err == -EINVAL) 538 f2fs_warn(sbi, "Value of option \"%s\" is unrecognized", 539 opt); 540 else 541 f2fs_warn(sbi, "Error processing option \"%s\" [%d]", 542 opt, err); 543 return -EINVAL; 544 } 545 f2fs_warn(sbi, "Test dummy encryption mode enabled"); 546 return 0; 547 } 548 549 #ifdef CONFIG_F2FS_FS_COMPRESSION 550 /* 551 * 1. The same extension name cannot not appear in both compress and non-compress extension 552 * at the same time. 553 * 2. If the compress extension specifies all files, the types specified by the non-compress 554 * extension will be treated as special cases and will not be compressed. 555 * 3. Don't allow the non-compress extension specifies all files. 556 */ 557 static int f2fs_test_compress_extension(struct f2fs_sb_info *sbi) 558 { 559 unsigned char (*ext)[F2FS_EXTENSION_LEN]; 560 unsigned char (*noext)[F2FS_EXTENSION_LEN]; 561 int ext_cnt, noext_cnt, index = 0, no_index = 0; 562 563 ext = F2FS_OPTION(sbi).extensions; 564 ext_cnt = F2FS_OPTION(sbi).compress_ext_cnt; 565 noext = F2FS_OPTION(sbi).noextensions; 566 noext_cnt = F2FS_OPTION(sbi).nocompress_ext_cnt; 567 568 if (!noext_cnt) 569 return 0; 570 571 for (no_index = 0; no_index < noext_cnt; no_index++) { 572 if (!strcasecmp("*", noext[no_index])) { 573 f2fs_info(sbi, "Don't allow the nocompress extension specifies all files"); 574 return -EINVAL; 575 } 576 for (index = 0; index < ext_cnt; index++) { 577 if (!strcasecmp(ext[index], noext[no_index])) { 578 f2fs_info(sbi, "Don't allow the same extension %s appear in both compress and nocompress extension", 579 ext[index]); 580 return -EINVAL; 581 } 582 } 583 } 584 return 0; 585 } 586 587 #ifdef CONFIG_F2FS_FS_LZ4 588 static int f2fs_set_lz4hc_level(struct f2fs_sb_info *sbi, const char *str) 589 { 590 #ifdef CONFIG_F2FS_FS_LZ4HC 591 unsigned int level; 592 #endif 593 594 if (strlen(str) == 3) { 595 F2FS_OPTION(sbi).compress_level = 0; 596 return 0; 597 } 598 599 #ifdef CONFIG_F2FS_FS_LZ4HC 600 str += 3; 601 602 if (str[0] != ':') { 603 f2fs_info(sbi, "wrong format, e.g. <alg_name>:<compr_level>"); 604 return -EINVAL; 605 } 606 if (kstrtouint(str + 1, 10, &level)) 607 return -EINVAL; 608 609 if (level < LZ4HC_MIN_CLEVEL || level > LZ4HC_MAX_CLEVEL) { 610 f2fs_info(sbi, "invalid lz4hc compress level: %d", level); 611 return -EINVAL; 612 } 613 614 F2FS_OPTION(sbi).compress_level = level; 615 return 0; 616 #else 617 f2fs_info(sbi, "kernel doesn't support lz4hc compression"); 618 return -EINVAL; 619 #endif 620 } 621 #endif 622 623 #ifdef CONFIG_F2FS_FS_ZSTD 624 static int f2fs_set_zstd_level(struct f2fs_sb_info *sbi, const char *str) 625 { 626 unsigned int level; 627 int len = 4; 628 629 if (strlen(str) == len) { 630 F2FS_OPTION(sbi).compress_level = 0; 631 return 0; 632 } 633 634 str += len; 635 636 if (str[0] != ':') { 637 f2fs_info(sbi, "wrong format, e.g. <alg_name>:<compr_level>"); 638 return -EINVAL; 639 } 640 if (kstrtouint(str + 1, 10, &level)) 641 return -EINVAL; 642 643 if (!level || level > zstd_max_clevel()) { 644 f2fs_info(sbi, "invalid zstd compress level: %d", level); 645 return -EINVAL; 646 } 647 648 F2FS_OPTION(sbi).compress_level = level; 649 return 0; 650 } 651 #endif 652 #endif 653 654 static int parse_options(struct super_block *sb, char *options, bool is_remount) 655 { 656 struct f2fs_sb_info *sbi = F2FS_SB(sb); 657 substring_t args[MAX_OPT_ARGS]; 658 #ifdef CONFIG_F2FS_FS_COMPRESSION 659 unsigned char (*ext)[F2FS_EXTENSION_LEN]; 660 unsigned char (*noext)[F2FS_EXTENSION_LEN]; 661 int ext_cnt, noext_cnt; 662 #endif 663 char *p, *name; 664 int arg = 0; 665 kuid_t uid; 666 kgid_t gid; 667 int ret; 668 669 if (!options) 670 goto default_check; 671 672 while ((p = strsep(&options, ",")) != NULL) { 673 int token; 674 675 if (!*p) 676 continue; 677 /* 678 * Initialize args struct so we know whether arg was 679 * found; some options take optional arguments. 680 */ 681 args[0].to = args[0].from = NULL; 682 token = match_token(p, f2fs_tokens, args); 683 684 switch (token) { 685 case Opt_gc_background: 686 name = match_strdup(&args[0]); 687 688 if (!name) 689 return -ENOMEM; 690 if (!strcmp(name, "on")) { 691 F2FS_OPTION(sbi).bggc_mode = BGGC_MODE_ON; 692 } else if (!strcmp(name, "off")) { 693 F2FS_OPTION(sbi).bggc_mode = BGGC_MODE_OFF; 694 } else if (!strcmp(name, "sync")) { 695 F2FS_OPTION(sbi).bggc_mode = BGGC_MODE_SYNC; 696 } else { 697 kfree(name); 698 return -EINVAL; 699 } 700 kfree(name); 701 break; 702 case Opt_disable_roll_forward: 703 set_opt(sbi, DISABLE_ROLL_FORWARD); 704 break; 705 case Opt_norecovery: 706 /* this option mounts f2fs with ro */ 707 set_opt(sbi, NORECOVERY); 708 if (!f2fs_readonly(sb)) 709 return -EINVAL; 710 break; 711 case Opt_discard: 712 if (!f2fs_hw_support_discard(sbi)) { 713 f2fs_warn(sbi, "device does not support discard"); 714 break; 715 } 716 set_opt(sbi, DISCARD); 717 break; 718 case Opt_nodiscard: 719 if (f2fs_hw_should_discard(sbi)) { 720 f2fs_warn(sbi, "discard is required for zoned block devices"); 721 return -EINVAL; 722 } 723 clear_opt(sbi, DISCARD); 724 break; 725 case Opt_noheap: 726 set_opt(sbi, NOHEAP); 727 break; 728 case Opt_heap: 729 clear_opt(sbi, NOHEAP); 730 break; 731 #ifdef CONFIG_F2FS_FS_XATTR 732 case Opt_user_xattr: 733 set_opt(sbi, XATTR_USER); 734 break; 735 case Opt_nouser_xattr: 736 clear_opt(sbi, XATTR_USER); 737 break; 738 case Opt_inline_xattr: 739 set_opt(sbi, INLINE_XATTR); 740 break; 741 case Opt_noinline_xattr: 742 clear_opt(sbi, INLINE_XATTR); 743 break; 744 case Opt_inline_xattr_size: 745 if (args->from && match_int(args, &arg)) 746 return -EINVAL; 747 set_opt(sbi, INLINE_XATTR_SIZE); 748 F2FS_OPTION(sbi).inline_xattr_size = arg; 749 break; 750 #else 751 case Opt_user_xattr: 752 f2fs_info(sbi, "user_xattr options not supported"); 753 break; 754 case Opt_nouser_xattr: 755 f2fs_info(sbi, "nouser_xattr options not supported"); 756 break; 757 case Opt_inline_xattr: 758 f2fs_info(sbi, "inline_xattr options not supported"); 759 break; 760 case Opt_noinline_xattr: 761 f2fs_info(sbi, "noinline_xattr options not supported"); 762 break; 763 #endif 764 #ifdef CONFIG_F2FS_FS_POSIX_ACL 765 case Opt_acl: 766 set_opt(sbi, POSIX_ACL); 767 break; 768 case Opt_noacl: 769 clear_opt(sbi, POSIX_ACL); 770 break; 771 #else 772 case Opt_acl: 773 f2fs_info(sbi, "acl options not supported"); 774 break; 775 case Opt_noacl: 776 f2fs_info(sbi, "noacl options not supported"); 777 break; 778 #endif 779 case Opt_active_logs: 780 if (args->from && match_int(args, &arg)) 781 return -EINVAL; 782 if (arg != 2 && arg != 4 && 783 arg != NR_CURSEG_PERSIST_TYPE) 784 return -EINVAL; 785 F2FS_OPTION(sbi).active_logs = arg; 786 break; 787 case Opt_disable_ext_identify: 788 set_opt(sbi, DISABLE_EXT_IDENTIFY); 789 break; 790 case Opt_inline_data: 791 set_opt(sbi, INLINE_DATA); 792 break; 793 case Opt_inline_dentry: 794 set_opt(sbi, INLINE_DENTRY); 795 break; 796 case Opt_noinline_dentry: 797 clear_opt(sbi, INLINE_DENTRY); 798 break; 799 case Opt_flush_merge: 800 set_opt(sbi, FLUSH_MERGE); 801 break; 802 case Opt_noflush_merge: 803 clear_opt(sbi, FLUSH_MERGE); 804 break; 805 case Opt_nobarrier: 806 set_opt(sbi, NOBARRIER); 807 break; 808 case Opt_barrier: 809 clear_opt(sbi, NOBARRIER); 810 break; 811 case Opt_fastboot: 812 set_opt(sbi, FASTBOOT); 813 break; 814 case Opt_extent_cache: 815 set_opt(sbi, READ_EXTENT_CACHE); 816 break; 817 case Opt_noextent_cache: 818 clear_opt(sbi, READ_EXTENT_CACHE); 819 break; 820 case Opt_noinline_data: 821 clear_opt(sbi, INLINE_DATA); 822 break; 823 case Opt_data_flush: 824 set_opt(sbi, DATA_FLUSH); 825 break; 826 case Opt_reserve_root: 827 if (args->from && match_int(args, &arg)) 828 return -EINVAL; 829 if (test_opt(sbi, RESERVE_ROOT)) { 830 f2fs_info(sbi, "Preserve previous reserve_root=%u", 831 F2FS_OPTION(sbi).root_reserved_blocks); 832 } else { 833 F2FS_OPTION(sbi).root_reserved_blocks = arg; 834 set_opt(sbi, RESERVE_ROOT); 835 } 836 break; 837 case Opt_resuid: 838 if (args->from && match_int(args, &arg)) 839 return -EINVAL; 840 uid = make_kuid(current_user_ns(), arg); 841 if (!uid_valid(uid)) { 842 f2fs_err(sbi, "Invalid uid value %d", arg); 843 return -EINVAL; 844 } 845 F2FS_OPTION(sbi).s_resuid = uid; 846 break; 847 case Opt_resgid: 848 if (args->from && match_int(args, &arg)) 849 return -EINVAL; 850 gid = make_kgid(current_user_ns(), arg); 851 if (!gid_valid(gid)) { 852 f2fs_err(sbi, "Invalid gid value %d", arg); 853 return -EINVAL; 854 } 855 F2FS_OPTION(sbi).s_resgid = gid; 856 break; 857 case Opt_mode: 858 name = match_strdup(&args[0]); 859 860 if (!name) 861 return -ENOMEM; 862 if (!strcmp(name, "adaptive")) { 863 if (f2fs_sb_has_blkzoned(sbi)) { 864 f2fs_warn(sbi, "adaptive mode is not allowed with zoned block device feature"); 865 kfree(name); 866 return -EINVAL; 867 } 868 F2FS_OPTION(sbi).fs_mode = FS_MODE_ADAPTIVE; 869 } else if (!strcmp(name, "lfs")) { 870 F2FS_OPTION(sbi).fs_mode = FS_MODE_LFS; 871 } else if (!strcmp(name, "fragment:segment")) { 872 F2FS_OPTION(sbi).fs_mode = FS_MODE_FRAGMENT_SEG; 873 } else if (!strcmp(name, "fragment:block")) { 874 F2FS_OPTION(sbi).fs_mode = FS_MODE_FRAGMENT_BLK; 875 } else { 876 kfree(name); 877 return -EINVAL; 878 } 879 kfree(name); 880 break; 881 case Opt_io_size_bits: 882 if (args->from && match_int(args, &arg)) 883 return -EINVAL; 884 if (arg <= 0 || arg > __ilog2_u32(BIO_MAX_VECS)) { 885 f2fs_warn(sbi, "Not support %ld, larger than %d", 886 BIT(arg), BIO_MAX_VECS); 887 return -EINVAL; 888 } 889 F2FS_OPTION(sbi).write_io_size_bits = arg; 890 break; 891 #ifdef CONFIG_F2FS_FAULT_INJECTION 892 case Opt_fault_injection: 893 if (args->from && match_int(args, &arg)) 894 return -EINVAL; 895 f2fs_build_fault_attr(sbi, arg, F2FS_ALL_FAULT_TYPE); 896 set_opt(sbi, FAULT_INJECTION); 897 break; 898 899 case Opt_fault_type: 900 if (args->from && match_int(args, &arg)) 901 return -EINVAL; 902 f2fs_build_fault_attr(sbi, 0, arg); 903 set_opt(sbi, FAULT_INJECTION); 904 break; 905 #else 906 case Opt_fault_injection: 907 f2fs_info(sbi, "fault_injection options not supported"); 908 break; 909 910 case Opt_fault_type: 911 f2fs_info(sbi, "fault_type options not supported"); 912 break; 913 #endif 914 case Opt_lazytime: 915 sb->s_flags |= SB_LAZYTIME; 916 break; 917 case Opt_nolazytime: 918 sb->s_flags &= ~SB_LAZYTIME; 919 break; 920 #ifdef CONFIG_QUOTA 921 case Opt_quota: 922 case Opt_usrquota: 923 set_opt(sbi, USRQUOTA); 924 break; 925 case Opt_grpquota: 926 set_opt(sbi, GRPQUOTA); 927 break; 928 case Opt_prjquota: 929 set_opt(sbi, PRJQUOTA); 930 break; 931 case Opt_usrjquota: 932 ret = f2fs_set_qf_name(sb, USRQUOTA, &args[0]); 933 if (ret) 934 return ret; 935 break; 936 case Opt_grpjquota: 937 ret = f2fs_set_qf_name(sb, GRPQUOTA, &args[0]); 938 if (ret) 939 return ret; 940 break; 941 case Opt_prjjquota: 942 ret = f2fs_set_qf_name(sb, PRJQUOTA, &args[0]); 943 if (ret) 944 return ret; 945 break; 946 case Opt_offusrjquota: 947 ret = f2fs_clear_qf_name(sb, USRQUOTA); 948 if (ret) 949 return ret; 950 break; 951 case Opt_offgrpjquota: 952 ret = f2fs_clear_qf_name(sb, GRPQUOTA); 953 if (ret) 954 return ret; 955 break; 956 case Opt_offprjjquota: 957 ret = f2fs_clear_qf_name(sb, PRJQUOTA); 958 if (ret) 959 return ret; 960 break; 961 case Opt_jqfmt_vfsold: 962 F2FS_OPTION(sbi).s_jquota_fmt = QFMT_VFS_OLD; 963 break; 964 case Opt_jqfmt_vfsv0: 965 F2FS_OPTION(sbi).s_jquota_fmt = QFMT_VFS_V0; 966 break; 967 case Opt_jqfmt_vfsv1: 968 F2FS_OPTION(sbi).s_jquota_fmt = QFMT_VFS_V1; 969 break; 970 case Opt_noquota: 971 clear_opt(sbi, QUOTA); 972 clear_opt(sbi, USRQUOTA); 973 clear_opt(sbi, GRPQUOTA); 974 clear_opt(sbi, PRJQUOTA); 975 break; 976 #else 977 case Opt_quota: 978 case Opt_usrquota: 979 case Opt_grpquota: 980 case Opt_prjquota: 981 case Opt_usrjquota: 982 case Opt_grpjquota: 983 case Opt_prjjquota: 984 case Opt_offusrjquota: 985 case Opt_offgrpjquota: 986 case Opt_offprjjquota: 987 case Opt_jqfmt_vfsold: 988 case Opt_jqfmt_vfsv0: 989 case Opt_jqfmt_vfsv1: 990 case Opt_noquota: 991 f2fs_info(sbi, "quota operations not supported"); 992 break; 993 #endif 994 case Opt_alloc: 995 name = match_strdup(&args[0]); 996 if (!name) 997 return -ENOMEM; 998 999 if (!strcmp(name, "default")) { 1000 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_DEFAULT; 1001 } else if (!strcmp(name, "reuse")) { 1002 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_REUSE; 1003 } else { 1004 kfree(name); 1005 return -EINVAL; 1006 } 1007 kfree(name); 1008 break; 1009 case Opt_fsync: 1010 name = match_strdup(&args[0]); 1011 if (!name) 1012 return -ENOMEM; 1013 if (!strcmp(name, "posix")) { 1014 F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_POSIX; 1015 } else if (!strcmp(name, "strict")) { 1016 F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_STRICT; 1017 } else if (!strcmp(name, "nobarrier")) { 1018 F2FS_OPTION(sbi).fsync_mode = 1019 FSYNC_MODE_NOBARRIER; 1020 } else { 1021 kfree(name); 1022 return -EINVAL; 1023 } 1024 kfree(name); 1025 break; 1026 case Opt_test_dummy_encryption: 1027 ret = f2fs_set_test_dummy_encryption(sb, p, &args[0], 1028 is_remount); 1029 if (ret) 1030 return ret; 1031 break; 1032 case Opt_inlinecrypt: 1033 #ifdef CONFIG_FS_ENCRYPTION_INLINE_CRYPT 1034 sb->s_flags |= SB_INLINECRYPT; 1035 #else 1036 f2fs_info(sbi, "inline encryption not supported"); 1037 #endif 1038 break; 1039 case Opt_checkpoint_disable_cap_perc: 1040 if (args->from && match_int(args, &arg)) 1041 return -EINVAL; 1042 if (arg < 0 || arg > 100) 1043 return -EINVAL; 1044 F2FS_OPTION(sbi).unusable_cap_perc = arg; 1045 set_opt(sbi, DISABLE_CHECKPOINT); 1046 break; 1047 case Opt_checkpoint_disable_cap: 1048 if (args->from && match_int(args, &arg)) 1049 return -EINVAL; 1050 F2FS_OPTION(sbi).unusable_cap = arg; 1051 set_opt(sbi, DISABLE_CHECKPOINT); 1052 break; 1053 case Opt_checkpoint_disable: 1054 set_opt(sbi, DISABLE_CHECKPOINT); 1055 break; 1056 case Opt_checkpoint_enable: 1057 clear_opt(sbi, DISABLE_CHECKPOINT); 1058 break; 1059 case Opt_checkpoint_merge: 1060 set_opt(sbi, MERGE_CHECKPOINT); 1061 break; 1062 case Opt_nocheckpoint_merge: 1063 clear_opt(sbi, MERGE_CHECKPOINT); 1064 break; 1065 #ifdef CONFIG_F2FS_FS_COMPRESSION 1066 case Opt_compress_algorithm: 1067 if (!f2fs_sb_has_compression(sbi)) { 1068 f2fs_info(sbi, "Image doesn't support compression"); 1069 break; 1070 } 1071 name = match_strdup(&args[0]); 1072 if (!name) 1073 return -ENOMEM; 1074 if (!strcmp(name, "lzo")) { 1075 #ifdef CONFIG_F2FS_FS_LZO 1076 F2FS_OPTION(sbi).compress_level = 0; 1077 F2FS_OPTION(sbi).compress_algorithm = 1078 COMPRESS_LZO; 1079 #else 1080 f2fs_info(sbi, "kernel doesn't support lzo compression"); 1081 #endif 1082 } else if (!strncmp(name, "lz4", 3)) { 1083 #ifdef CONFIG_F2FS_FS_LZ4 1084 ret = f2fs_set_lz4hc_level(sbi, name); 1085 if (ret) { 1086 kfree(name); 1087 return -EINVAL; 1088 } 1089 F2FS_OPTION(sbi).compress_algorithm = 1090 COMPRESS_LZ4; 1091 #else 1092 f2fs_info(sbi, "kernel doesn't support lz4 compression"); 1093 #endif 1094 } else if (!strncmp(name, "zstd", 4)) { 1095 #ifdef CONFIG_F2FS_FS_ZSTD 1096 ret = f2fs_set_zstd_level(sbi, name); 1097 if (ret) { 1098 kfree(name); 1099 return -EINVAL; 1100 } 1101 F2FS_OPTION(sbi).compress_algorithm = 1102 COMPRESS_ZSTD; 1103 #else 1104 f2fs_info(sbi, "kernel doesn't support zstd compression"); 1105 #endif 1106 } else if (!strcmp(name, "lzo-rle")) { 1107 #ifdef CONFIG_F2FS_FS_LZORLE 1108 F2FS_OPTION(sbi).compress_level = 0; 1109 F2FS_OPTION(sbi).compress_algorithm = 1110 COMPRESS_LZORLE; 1111 #else 1112 f2fs_info(sbi, "kernel doesn't support lzorle compression"); 1113 #endif 1114 } else { 1115 kfree(name); 1116 return -EINVAL; 1117 } 1118 kfree(name); 1119 break; 1120 case Opt_compress_log_size: 1121 if (!f2fs_sb_has_compression(sbi)) { 1122 f2fs_info(sbi, "Image doesn't support compression"); 1123 break; 1124 } 1125 if (args->from && match_int(args, &arg)) 1126 return -EINVAL; 1127 if (arg < MIN_COMPRESS_LOG_SIZE || 1128 arg > MAX_COMPRESS_LOG_SIZE) { 1129 f2fs_err(sbi, 1130 "Compress cluster log size is out of range"); 1131 return -EINVAL; 1132 } 1133 F2FS_OPTION(sbi).compress_log_size = arg; 1134 break; 1135 case Opt_compress_extension: 1136 if (!f2fs_sb_has_compression(sbi)) { 1137 f2fs_info(sbi, "Image doesn't support compression"); 1138 break; 1139 } 1140 name = match_strdup(&args[0]); 1141 if (!name) 1142 return -ENOMEM; 1143 1144 ext = F2FS_OPTION(sbi).extensions; 1145 ext_cnt = F2FS_OPTION(sbi).compress_ext_cnt; 1146 1147 if (strlen(name) >= F2FS_EXTENSION_LEN || 1148 ext_cnt >= COMPRESS_EXT_NUM) { 1149 f2fs_err(sbi, 1150 "invalid extension length/number"); 1151 kfree(name); 1152 return -EINVAL; 1153 } 1154 1155 strcpy(ext[ext_cnt], name); 1156 F2FS_OPTION(sbi).compress_ext_cnt++; 1157 kfree(name); 1158 break; 1159 case Opt_nocompress_extension: 1160 if (!f2fs_sb_has_compression(sbi)) { 1161 f2fs_info(sbi, "Image doesn't support compression"); 1162 break; 1163 } 1164 name = match_strdup(&args[0]); 1165 if (!name) 1166 return -ENOMEM; 1167 1168 noext = F2FS_OPTION(sbi).noextensions; 1169 noext_cnt = F2FS_OPTION(sbi).nocompress_ext_cnt; 1170 1171 if (strlen(name) >= F2FS_EXTENSION_LEN || 1172 noext_cnt >= COMPRESS_EXT_NUM) { 1173 f2fs_err(sbi, 1174 "invalid extension length/number"); 1175 kfree(name); 1176 return -EINVAL; 1177 } 1178 1179 strcpy(noext[noext_cnt], name); 1180 F2FS_OPTION(sbi).nocompress_ext_cnt++; 1181 kfree(name); 1182 break; 1183 case Opt_compress_chksum: 1184 if (!f2fs_sb_has_compression(sbi)) { 1185 f2fs_info(sbi, "Image doesn't support compression"); 1186 break; 1187 } 1188 F2FS_OPTION(sbi).compress_chksum = true; 1189 break; 1190 case Opt_compress_mode: 1191 if (!f2fs_sb_has_compression(sbi)) { 1192 f2fs_info(sbi, "Image doesn't support compression"); 1193 break; 1194 } 1195 name = match_strdup(&args[0]); 1196 if (!name) 1197 return -ENOMEM; 1198 if (!strcmp(name, "fs")) { 1199 F2FS_OPTION(sbi).compress_mode = COMPR_MODE_FS; 1200 } else if (!strcmp(name, "user")) { 1201 F2FS_OPTION(sbi).compress_mode = COMPR_MODE_USER; 1202 } else { 1203 kfree(name); 1204 return -EINVAL; 1205 } 1206 kfree(name); 1207 break; 1208 case Opt_compress_cache: 1209 if (!f2fs_sb_has_compression(sbi)) { 1210 f2fs_info(sbi, "Image doesn't support compression"); 1211 break; 1212 } 1213 set_opt(sbi, COMPRESS_CACHE); 1214 break; 1215 #else 1216 case Opt_compress_algorithm: 1217 case Opt_compress_log_size: 1218 case Opt_compress_extension: 1219 case Opt_nocompress_extension: 1220 case Opt_compress_chksum: 1221 case Opt_compress_mode: 1222 case Opt_compress_cache: 1223 f2fs_info(sbi, "compression options not supported"); 1224 break; 1225 #endif 1226 case Opt_atgc: 1227 set_opt(sbi, ATGC); 1228 break; 1229 case Opt_gc_merge: 1230 set_opt(sbi, GC_MERGE); 1231 break; 1232 case Opt_nogc_merge: 1233 clear_opt(sbi, GC_MERGE); 1234 break; 1235 case Opt_discard_unit: 1236 name = match_strdup(&args[0]); 1237 if (!name) 1238 return -ENOMEM; 1239 if (!strcmp(name, "block")) { 1240 F2FS_OPTION(sbi).discard_unit = 1241 DISCARD_UNIT_BLOCK; 1242 } else if (!strcmp(name, "segment")) { 1243 F2FS_OPTION(sbi).discard_unit = 1244 DISCARD_UNIT_SEGMENT; 1245 } else if (!strcmp(name, "section")) { 1246 F2FS_OPTION(sbi).discard_unit = 1247 DISCARD_UNIT_SECTION; 1248 } else { 1249 kfree(name); 1250 return -EINVAL; 1251 } 1252 kfree(name); 1253 break; 1254 case Opt_memory_mode: 1255 name = match_strdup(&args[0]); 1256 if (!name) 1257 return -ENOMEM; 1258 if (!strcmp(name, "normal")) { 1259 F2FS_OPTION(sbi).memory_mode = 1260 MEMORY_MODE_NORMAL; 1261 } else if (!strcmp(name, "low")) { 1262 F2FS_OPTION(sbi).memory_mode = 1263 MEMORY_MODE_LOW; 1264 } else { 1265 kfree(name); 1266 return -EINVAL; 1267 } 1268 kfree(name); 1269 break; 1270 case Opt_age_extent_cache: 1271 set_opt(sbi, AGE_EXTENT_CACHE); 1272 break; 1273 case Opt_errors: 1274 name = match_strdup(&args[0]); 1275 if (!name) 1276 return -ENOMEM; 1277 if (!strcmp(name, "remount-ro")) { 1278 F2FS_OPTION(sbi).errors = 1279 MOUNT_ERRORS_READONLY; 1280 } else if (!strcmp(name, "continue")) { 1281 F2FS_OPTION(sbi).errors = 1282 MOUNT_ERRORS_CONTINUE; 1283 } else if (!strcmp(name, "panic")) { 1284 F2FS_OPTION(sbi).errors = 1285 MOUNT_ERRORS_PANIC; 1286 } else { 1287 kfree(name); 1288 return -EINVAL; 1289 } 1290 kfree(name); 1291 break; 1292 default: 1293 f2fs_err(sbi, "Unrecognized mount option \"%s\" or missing value", 1294 p); 1295 return -EINVAL; 1296 } 1297 } 1298 default_check: 1299 #ifdef CONFIG_QUOTA 1300 if (f2fs_check_quota_options(sbi)) 1301 return -EINVAL; 1302 #else 1303 if (f2fs_sb_has_quota_ino(sbi) && !f2fs_readonly(sbi->sb)) { 1304 f2fs_info(sbi, "Filesystem with quota feature cannot be mounted RDWR without CONFIG_QUOTA"); 1305 return -EINVAL; 1306 } 1307 if (f2fs_sb_has_project_quota(sbi) && !f2fs_readonly(sbi->sb)) { 1308 f2fs_err(sbi, "Filesystem with project quota feature cannot be mounted RDWR without CONFIG_QUOTA"); 1309 return -EINVAL; 1310 } 1311 #endif 1312 #if !IS_ENABLED(CONFIG_UNICODE) 1313 if (f2fs_sb_has_casefold(sbi)) { 1314 f2fs_err(sbi, 1315 "Filesystem with casefold feature cannot be mounted without CONFIG_UNICODE"); 1316 return -EINVAL; 1317 } 1318 #endif 1319 /* 1320 * The BLKZONED feature indicates that the drive was formatted with 1321 * zone alignment optimization. This is optional for host-aware 1322 * devices, but mandatory for host-managed zoned block devices. 1323 */ 1324 if (f2fs_sb_has_blkzoned(sbi)) { 1325 #ifdef CONFIG_BLK_DEV_ZONED 1326 if (F2FS_OPTION(sbi).discard_unit != 1327 DISCARD_UNIT_SECTION) { 1328 f2fs_info(sbi, "Zoned block device doesn't need small discard, set discard_unit=section by default"); 1329 F2FS_OPTION(sbi).discard_unit = 1330 DISCARD_UNIT_SECTION; 1331 } 1332 #else 1333 f2fs_err(sbi, "Zoned block device support is not enabled"); 1334 return -EINVAL; 1335 #endif 1336 } 1337 1338 #ifdef CONFIG_F2FS_FS_COMPRESSION 1339 if (f2fs_test_compress_extension(sbi)) { 1340 f2fs_err(sbi, "invalid compress or nocompress extension"); 1341 return -EINVAL; 1342 } 1343 #endif 1344 1345 if (F2FS_IO_SIZE_BITS(sbi) && !f2fs_lfs_mode(sbi)) { 1346 f2fs_err(sbi, "Should set mode=lfs with %luKB-sized IO", 1347 F2FS_IO_SIZE_KB(sbi)); 1348 return -EINVAL; 1349 } 1350 1351 if (test_opt(sbi, INLINE_XATTR_SIZE)) { 1352 int min_size, max_size; 1353 1354 if (!f2fs_sb_has_extra_attr(sbi) || 1355 !f2fs_sb_has_flexible_inline_xattr(sbi)) { 1356 f2fs_err(sbi, "extra_attr or flexible_inline_xattr feature is off"); 1357 return -EINVAL; 1358 } 1359 if (!test_opt(sbi, INLINE_XATTR)) { 1360 f2fs_err(sbi, "inline_xattr_size option should be set with inline_xattr option"); 1361 return -EINVAL; 1362 } 1363 1364 min_size = sizeof(struct f2fs_xattr_header) / sizeof(__le32); 1365 max_size = MAX_INLINE_XATTR_SIZE; 1366 1367 if (F2FS_OPTION(sbi).inline_xattr_size < min_size || 1368 F2FS_OPTION(sbi).inline_xattr_size > max_size) { 1369 f2fs_err(sbi, "inline xattr size is out of range: %d ~ %d", 1370 min_size, max_size); 1371 return -EINVAL; 1372 } 1373 } 1374 1375 if (test_opt(sbi, DISABLE_CHECKPOINT) && f2fs_lfs_mode(sbi)) { 1376 f2fs_err(sbi, "LFS is not compatible with checkpoint=disable"); 1377 return -EINVAL; 1378 } 1379 1380 if (test_opt(sbi, ATGC) && f2fs_lfs_mode(sbi)) { 1381 f2fs_err(sbi, "LFS is not compatible with ATGC"); 1382 return -EINVAL; 1383 } 1384 1385 if (f2fs_is_readonly(sbi) && test_opt(sbi, FLUSH_MERGE)) { 1386 f2fs_err(sbi, "FLUSH_MERGE not compatible with readonly mode"); 1387 return -EINVAL; 1388 } 1389 1390 if (f2fs_sb_has_readonly(sbi) && !f2fs_readonly(sbi->sb)) { 1391 f2fs_err(sbi, "Allow to mount readonly mode only"); 1392 return -EROFS; 1393 } 1394 return 0; 1395 } 1396 1397 static struct inode *f2fs_alloc_inode(struct super_block *sb) 1398 { 1399 struct f2fs_inode_info *fi; 1400 1401 if (time_to_inject(F2FS_SB(sb), FAULT_SLAB_ALLOC)) 1402 return NULL; 1403 1404 fi = alloc_inode_sb(sb, f2fs_inode_cachep, GFP_F2FS_ZERO); 1405 if (!fi) 1406 return NULL; 1407 1408 init_once((void *) fi); 1409 1410 /* Initialize f2fs-specific inode info */ 1411 atomic_set(&fi->dirty_pages, 0); 1412 atomic_set(&fi->i_compr_blocks, 0); 1413 init_f2fs_rwsem(&fi->i_sem); 1414 spin_lock_init(&fi->i_size_lock); 1415 INIT_LIST_HEAD(&fi->dirty_list); 1416 INIT_LIST_HEAD(&fi->gdirty_list); 1417 init_f2fs_rwsem(&fi->i_gc_rwsem[READ]); 1418 init_f2fs_rwsem(&fi->i_gc_rwsem[WRITE]); 1419 init_f2fs_rwsem(&fi->i_xattr_sem); 1420 1421 /* Will be used by directory only */ 1422 fi->i_dir_level = F2FS_SB(sb)->dir_level; 1423 1424 return &fi->vfs_inode; 1425 } 1426 1427 static int f2fs_drop_inode(struct inode *inode) 1428 { 1429 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1430 int ret; 1431 1432 /* 1433 * during filesystem shutdown, if checkpoint is disabled, 1434 * drop useless meta/node dirty pages. 1435 */ 1436 if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED))) { 1437 if (inode->i_ino == F2FS_NODE_INO(sbi) || 1438 inode->i_ino == F2FS_META_INO(sbi)) { 1439 trace_f2fs_drop_inode(inode, 1); 1440 return 1; 1441 } 1442 } 1443 1444 /* 1445 * This is to avoid a deadlock condition like below. 1446 * writeback_single_inode(inode) 1447 * - f2fs_write_data_page 1448 * - f2fs_gc -> iput -> evict 1449 * - inode_wait_for_writeback(inode) 1450 */ 1451 if ((!inode_unhashed(inode) && inode->i_state & I_SYNC)) { 1452 if (!inode->i_nlink && !is_bad_inode(inode)) { 1453 /* to avoid evict_inode call simultaneously */ 1454 atomic_inc(&inode->i_count); 1455 spin_unlock(&inode->i_lock); 1456 1457 /* should remain fi->extent_tree for writepage */ 1458 f2fs_destroy_extent_node(inode); 1459 1460 sb_start_intwrite(inode->i_sb); 1461 f2fs_i_size_write(inode, 0); 1462 1463 f2fs_submit_merged_write_cond(F2FS_I_SB(inode), 1464 inode, NULL, 0, DATA); 1465 truncate_inode_pages_final(inode->i_mapping); 1466 1467 if (F2FS_HAS_BLOCKS(inode)) 1468 f2fs_truncate(inode); 1469 1470 sb_end_intwrite(inode->i_sb); 1471 1472 spin_lock(&inode->i_lock); 1473 atomic_dec(&inode->i_count); 1474 } 1475 trace_f2fs_drop_inode(inode, 0); 1476 return 0; 1477 } 1478 ret = generic_drop_inode(inode); 1479 if (!ret) 1480 ret = fscrypt_drop_inode(inode); 1481 trace_f2fs_drop_inode(inode, ret); 1482 return ret; 1483 } 1484 1485 int f2fs_inode_dirtied(struct inode *inode, bool sync) 1486 { 1487 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1488 int ret = 0; 1489 1490 spin_lock(&sbi->inode_lock[DIRTY_META]); 1491 if (is_inode_flag_set(inode, FI_DIRTY_INODE)) { 1492 ret = 1; 1493 } else { 1494 set_inode_flag(inode, FI_DIRTY_INODE); 1495 stat_inc_dirty_inode(sbi, DIRTY_META); 1496 } 1497 if (sync && list_empty(&F2FS_I(inode)->gdirty_list)) { 1498 list_add_tail(&F2FS_I(inode)->gdirty_list, 1499 &sbi->inode_list[DIRTY_META]); 1500 inc_page_count(sbi, F2FS_DIRTY_IMETA); 1501 } 1502 spin_unlock(&sbi->inode_lock[DIRTY_META]); 1503 return ret; 1504 } 1505 1506 void f2fs_inode_synced(struct inode *inode) 1507 { 1508 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1509 1510 spin_lock(&sbi->inode_lock[DIRTY_META]); 1511 if (!is_inode_flag_set(inode, FI_DIRTY_INODE)) { 1512 spin_unlock(&sbi->inode_lock[DIRTY_META]); 1513 return; 1514 } 1515 if (!list_empty(&F2FS_I(inode)->gdirty_list)) { 1516 list_del_init(&F2FS_I(inode)->gdirty_list); 1517 dec_page_count(sbi, F2FS_DIRTY_IMETA); 1518 } 1519 clear_inode_flag(inode, FI_DIRTY_INODE); 1520 clear_inode_flag(inode, FI_AUTO_RECOVER); 1521 stat_dec_dirty_inode(F2FS_I_SB(inode), DIRTY_META); 1522 spin_unlock(&sbi->inode_lock[DIRTY_META]); 1523 } 1524 1525 /* 1526 * f2fs_dirty_inode() is called from __mark_inode_dirty() 1527 * 1528 * We should call set_dirty_inode to write the dirty inode through write_inode. 1529 */ 1530 static void f2fs_dirty_inode(struct inode *inode, int flags) 1531 { 1532 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1533 1534 if (inode->i_ino == F2FS_NODE_INO(sbi) || 1535 inode->i_ino == F2FS_META_INO(sbi)) 1536 return; 1537 1538 if (is_inode_flag_set(inode, FI_AUTO_RECOVER)) 1539 clear_inode_flag(inode, FI_AUTO_RECOVER); 1540 1541 f2fs_inode_dirtied(inode, false); 1542 } 1543 1544 static void f2fs_free_inode(struct inode *inode) 1545 { 1546 fscrypt_free_inode(inode); 1547 kmem_cache_free(f2fs_inode_cachep, F2FS_I(inode)); 1548 } 1549 1550 static void destroy_percpu_info(struct f2fs_sb_info *sbi) 1551 { 1552 percpu_counter_destroy(&sbi->total_valid_inode_count); 1553 percpu_counter_destroy(&sbi->rf_node_block_count); 1554 percpu_counter_destroy(&sbi->alloc_valid_block_count); 1555 } 1556 1557 static void destroy_device_list(struct f2fs_sb_info *sbi) 1558 { 1559 int i; 1560 1561 for (i = 0; i < sbi->s_ndevs; i++) { 1562 blkdev_put(FDEV(i).bdev, FMODE_EXCL); 1563 #ifdef CONFIG_BLK_DEV_ZONED 1564 kvfree(FDEV(i).blkz_seq); 1565 #endif 1566 } 1567 kvfree(sbi->devs); 1568 } 1569 1570 static void f2fs_put_super(struct super_block *sb) 1571 { 1572 struct f2fs_sb_info *sbi = F2FS_SB(sb); 1573 int i; 1574 bool done; 1575 1576 /* unregister procfs/sysfs entries in advance to avoid race case */ 1577 f2fs_unregister_sysfs(sbi); 1578 1579 f2fs_quota_off_umount(sb); 1580 1581 /* prevent remaining shrinker jobs */ 1582 mutex_lock(&sbi->umount_mutex); 1583 1584 /* 1585 * flush all issued checkpoints and stop checkpoint issue thread. 1586 * after then, all checkpoints should be done by each process context. 1587 */ 1588 f2fs_stop_ckpt_thread(sbi); 1589 1590 /* 1591 * We don't need to do checkpoint when superblock is clean. 1592 * But, the previous checkpoint was not done by umount, it needs to do 1593 * clean checkpoint again. 1594 */ 1595 if ((is_sbi_flag_set(sbi, SBI_IS_DIRTY) || 1596 !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG))) { 1597 struct cp_control cpc = { 1598 .reason = CP_UMOUNT, 1599 }; 1600 f2fs_write_checkpoint(sbi, &cpc); 1601 } 1602 1603 /* be sure to wait for any on-going discard commands */ 1604 done = f2fs_issue_discard_timeout(sbi); 1605 if (f2fs_realtime_discard_enable(sbi) && !sbi->discard_blks && done) { 1606 struct cp_control cpc = { 1607 .reason = CP_UMOUNT | CP_TRIMMED, 1608 }; 1609 f2fs_write_checkpoint(sbi, &cpc); 1610 } 1611 1612 /* 1613 * normally superblock is clean, so we need to release this. 1614 * In addition, EIO will skip do checkpoint, we need this as well. 1615 */ 1616 f2fs_release_ino_entry(sbi, true); 1617 1618 f2fs_leave_shrinker(sbi); 1619 mutex_unlock(&sbi->umount_mutex); 1620 1621 /* our cp_error case, we can wait for any writeback page */ 1622 f2fs_flush_merged_writes(sbi); 1623 1624 f2fs_wait_on_all_pages(sbi, F2FS_WB_CP_DATA); 1625 1626 f2fs_bug_on(sbi, sbi->fsync_node_num); 1627 1628 f2fs_destroy_compress_inode(sbi); 1629 1630 iput(sbi->node_inode); 1631 sbi->node_inode = NULL; 1632 1633 iput(sbi->meta_inode); 1634 sbi->meta_inode = NULL; 1635 1636 /* 1637 * iput() can update stat information, if f2fs_write_checkpoint() 1638 * above failed with error. 1639 */ 1640 f2fs_destroy_stats(sbi); 1641 1642 /* destroy f2fs internal modules */ 1643 f2fs_destroy_node_manager(sbi); 1644 f2fs_destroy_segment_manager(sbi); 1645 1646 /* flush s_error_work before sbi destroy */ 1647 flush_work(&sbi->s_error_work); 1648 1649 f2fs_destroy_post_read_wq(sbi); 1650 1651 kvfree(sbi->ckpt); 1652 1653 sb->s_fs_info = NULL; 1654 if (sbi->s_chksum_driver) 1655 crypto_free_shash(sbi->s_chksum_driver); 1656 kfree(sbi->raw_super); 1657 1658 destroy_device_list(sbi); 1659 f2fs_destroy_page_array_cache(sbi); 1660 f2fs_destroy_xattr_caches(sbi); 1661 mempool_destroy(sbi->write_io_dummy); 1662 #ifdef CONFIG_QUOTA 1663 for (i = 0; i < MAXQUOTAS; i++) 1664 kfree(F2FS_OPTION(sbi).s_qf_names[i]); 1665 #endif 1666 fscrypt_free_dummy_policy(&F2FS_OPTION(sbi).dummy_enc_policy); 1667 destroy_percpu_info(sbi); 1668 f2fs_destroy_iostat(sbi); 1669 for (i = 0; i < NR_PAGE_TYPE; i++) 1670 kvfree(sbi->write_io[i]); 1671 #if IS_ENABLED(CONFIG_UNICODE) 1672 utf8_unload(sb->s_encoding); 1673 #endif 1674 kfree(sbi); 1675 } 1676 1677 int f2fs_sync_fs(struct super_block *sb, int sync) 1678 { 1679 struct f2fs_sb_info *sbi = F2FS_SB(sb); 1680 int err = 0; 1681 1682 if (unlikely(f2fs_cp_error(sbi))) 1683 return 0; 1684 if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED))) 1685 return 0; 1686 1687 trace_f2fs_sync_fs(sb, sync); 1688 1689 if (unlikely(is_sbi_flag_set(sbi, SBI_POR_DOING))) 1690 return -EAGAIN; 1691 1692 if (sync) 1693 err = f2fs_issue_checkpoint(sbi); 1694 1695 return err; 1696 } 1697 1698 static int f2fs_freeze(struct super_block *sb) 1699 { 1700 if (f2fs_readonly(sb)) 1701 return 0; 1702 1703 /* IO error happened before */ 1704 if (unlikely(f2fs_cp_error(F2FS_SB(sb)))) 1705 return -EIO; 1706 1707 /* must be clean, since sync_filesystem() was already called */ 1708 if (is_sbi_flag_set(F2FS_SB(sb), SBI_IS_DIRTY)) 1709 return -EINVAL; 1710 1711 /* Let's flush checkpoints and stop the thread. */ 1712 f2fs_flush_ckpt_thread(F2FS_SB(sb)); 1713 1714 /* to avoid deadlock on f2fs_evict_inode->SB_FREEZE_FS */ 1715 set_sbi_flag(F2FS_SB(sb), SBI_IS_FREEZING); 1716 return 0; 1717 } 1718 1719 static int f2fs_unfreeze(struct super_block *sb) 1720 { 1721 clear_sbi_flag(F2FS_SB(sb), SBI_IS_FREEZING); 1722 return 0; 1723 } 1724 1725 #ifdef CONFIG_QUOTA 1726 static int f2fs_statfs_project(struct super_block *sb, 1727 kprojid_t projid, struct kstatfs *buf) 1728 { 1729 struct kqid qid; 1730 struct dquot *dquot; 1731 u64 limit; 1732 u64 curblock; 1733 1734 qid = make_kqid_projid(projid); 1735 dquot = dqget(sb, qid); 1736 if (IS_ERR(dquot)) 1737 return PTR_ERR(dquot); 1738 spin_lock(&dquot->dq_dqb_lock); 1739 1740 limit = min_not_zero(dquot->dq_dqb.dqb_bsoftlimit, 1741 dquot->dq_dqb.dqb_bhardlimit); 1742 if (limit) 1743 limit >>= sb->s_blocksize_bits; 1744 1745 if (limit && buf->f_blocks > limit) { 1746 curblock = (dquot->dq_dqb.dqb_curspace + 1747 dquot->dq_dqb.dqb_rsvspace) >> sb->s_blocksize_bits; 1748 buf->f_blocks = limit; 1749 buf->f_bfree = buf->f_bavail = 1750 (buf->f_blocks > curblock) ? 1751 (buf->f_blocks - curblock) : 0; 1752 } 1753 1754 limit = min_not_zero(dquot->dq_dqb.dqb_isoftlimit, 1755 dquot->dq_dqb.dqb_ihardlimit); 1756 1757 if (limit && buf->f_files > limit) { 1758 buf->f_files = limit; 1759 buf->f_ffree = 1760 (buf->f_files > dquot->dq_dqb.dqb_curinodes) ? 1761 (buf->f_files - dquot->dq_dqb.dqb_curinodes) : 0; 1762 } 1763 1764 spin_unlock(&dquot->dq_dqb_lock); 1765 dqput(dquot); 1766 return 0; 1767 } 1768 #endif 1769 1770 static int f2fs_statfs(struct dentry *dentry, struct kstatfs *buf) 1771 { 1772 struct super_block *sb = dentry->d_sb; 1773 struct f2fs_sb_info *sbi = F2FS_SB(sb); 1774 u64 id = huge_encode_dev(sb->s_bdev->bd_dev); 1775 block_t total_count, user_block_count, start_count; 1776 u64 avail_node_count; 1777 unsigned int total_valid_node_count; 1778 1779 total_count = le64_to_cpu(sbi->raw_super->block_count); 1780 start_count = le32_to_cpu(sbi->raw_super->segment0_blkaddr); 1781 buf->f_type = F2FS_SUPER_MAGIC; 1782 buf->f_bsize = sbi->blocksize; 1783 1784 buf->f_blocks = total_count - start_count; 1785 1786 spin_lock(&sbi->stat_lock); 1787 1788 user_block_count = sbi->user_block_count; 1789 total_valid_node_count = valid_node_count(sbi); 1790 avail_node_count = sbi->total_node_count - F2FS_RESERVED_NODE_NUM; 1791 buf->f_bfree = user_block_count - valid_user_blocks(sbi) - 1792 sbi->current_reserved_blocks; 1793 1794 if (unlikely(buf->f_bfree <= sbi->unusable_block_count)) 1795 buf->f_bfree = 0; 1796 else 1797 buf->f_bfree -= sbi->unusable_block_count; 1798 spin_unlock(&sbi->stat_lock); 1799 1800 if (buf->f_bfree > F2FS_OPTION(sbi).root_reserved_blocks) 1801 buf->f_bavail = buf->f_bfree - 1802 F2FS_OPTION(sbi).root_reserved_blocks; 1803 else 1804 buf->f_bavail = 0; 1805 1806 if (avail_node_count > user_block_count) { 1807 buf->f_files = user_block_count; 1808 buf->f_ffree = buf->f_bavail; 1809 } else { 1810 buf->f_files = avail_node_count; 1811 buf->f_ffree = min(avail_node_count - total_valid_node_count, 1812 buf->f_bavail); 1813 } 1814 1815 buf->f_namelen = F2FS_NAME_LEN; 1816 buf->f_fsid = u64_to_fsid(id); 1817 1818 #ifdef CONFIG_QUOTA 1819 if (is_inode_flag_set(dentry->d_inode, FI_PROJ_INHERIT) && 1820 sb_has_quota_limits_enabled(sb, PRJQUOTA)) { 1821 f2fs_statfs_project(sb, F2FS_I(dentry->d_inode)->i_projid, buf); 1822 } 1823 #endif 1824 return 0; 1825 } 1826 1827 static inline void f2fs_show_quota_options(struct seq_file *seq, 1828 struct super_block *sb) 1829 { 1830 #ifdef CONFIG_QUOTA 1831 struct f2fs_sb_info *sbi = F2FS_SB(sb); 1832 1833 if (F2FS_OPTION(sbi).s_jquota_fmt) { 1834 char *fmtname = ""; 1835 1836 switch (F2FS_OPTION(sbi).s_jquota_fmt) { 1837 case QFMT_VFS_OLD: 1838 fmtname = "vfsold"; 1839 break; 1840 case QFMT_VFS_V0: 1841 fmtname = "vfsv0"; 1842 break; 1843 case QFMT_VFS_V1: 1844 fmtname = "vfsv1"; 1845 break; 1846 } 1847 seq_printf(seq, ",jqfmt=%s", fmtname); 1848 } 1849 1850 if (F2FS_OPTION(sbi).s_qf_names[USRQUOTA]) 1851 seq_show_option(seq, "usrjquota", 1852 F2FS_OPTION(sbi).s_qf_names[USRQUOTA]); 1853 1854 if (F2FS_OPTION(sbi).s_qf_names[GRPQUOTA]) 1855 seq_show_option(seq, "grpjquota", 1856 F2FS_OPTION(sbi).s_qf_names[GRPQUOTA]); 1857 1858 if (F2FS_OPTION(sbi).s_qf_names[PRJQUOTA]) 1859 seq_show_option(seq, "prjjquota", 1860 F2FS_OPTION(sbi).s_qf_names[PRJQUOTA]); 1861 #endif 1862 } 1863 1864 #ifdef CONFIG_F2FS_FS_COMPRESSION 1865 static inline void f2fs_show_compress_options(struct seq_file *seq, 1866 struct super_block *sb) 1867 { 1868 struct f2fs_sb_info *sbi = F2FS_SB(sb); 1869 char *algtype = ""; 1870 int i; 1871 1872 if (!f2fs_sb_has_compression(sbi)) 1873 return; 1874 1875 switch (F2FS_OPTION(sbi).compress_algorithm) { 1876 case COMPRESS_LZO: 1877 algtype = "lzo"; 1878 break; 1879 case COMPRESS_LZ4: 1880 algtype = "lz4"; 1881 break; 1882 case COMPRESS_ZSTD: 1883 algtype = "zstd"; 1884 break; 1885 case COMPRESS_LZORLE: 1886 algtype = "lzo-rle"; 1887 break; 1888 } 1889 seq_printf(seq, ",compress_algorithm=%s", algtype); 1890 1891 if (F2FS_OPTION(sbi).compress_level) 1892 seq_printf(seq, ":%d", F2FS_OPTION(sbi).compress_level); 1893 1894 seq_printf(seq, ",compress_log_size=%u", 1895 F2FS_OPTION(sbi).compress_log_size); 1896 1897 for (i = 0; i < F2FS_OPTION(sbi).compress_ext_cnt; i++) { 1898 seq_printf(seq, ",compress_extension=%s", 1899 F2FS_OPTION(sbi).extensions[i]); 1900 } 1901 1902 for (i = 0; i < F2FS_OPTION(sbi).nocompress_ext_cnt; i++) { 1903 seq_printf(seq, ",nocompress_extension=%s", 1904 F2FS_OPTION(sbi).noextensions[i]); 1905 } 1906 1907 if (F2FS_OPTION(sbi).compress_chksum) 1908 seq_puts(seq, ",compress_chksum"); 1909 1910 if (F2FS_OPTION(sbi).compress_mode == COMPR_MODE_FS) 1911 seq_printf(seq, ",compress_mode=%s", "fs"); 1912 else if (F2FS_OPTION(sbi).compress_mode == COMPR_MODE_USER) 1913 seq_printf(seq, ",compress_mode=%s", "user"); 1914 1915 if (test_opt(sbi, COMPRESS_CACHE)) 1916 seq_puts(seq, ",compress_cache"); 1917 } 1918 #endif 1919 1920 static int f2fs_show_options(struct seq_file *seq, struct dentry *root) 1921 { 1922 struct f2fs_sb_info *sbi = F2FS_SB(root->d_sb); 1923 1924 if (F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_SYNC) 1925 seq_printf(seq, ",background_gc=%s", "sync"); 1926 else if (F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_ON) 1927 seq_printf(seq, ",background_gc=%s", "on"); 1928 else if (F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_OFF) 1929 seq_printf(seq, ",background_gc=%s", "off"); 1930 1931 if (test_opt(sbi, GC_MERGE)) 1932 seq_puts(seq, ",gc_merge"); 1933 else 1934 seq_puts(seq, ",nogc_merge"); 1935 1936 if (test_opt(sbi, DISABLE_ROLL_FORWARD)) 1937 seq_puts(seq, ",disable_roll_forward"); 1938 if (test_opt(sbi, NORECOVERY)) 1939 seq_puts(seq, ",norecovery"); 1940 if (test_opt(sbi, DISCARD)) { 1941 seq_puts(seq, ",discard"); 1942 if (F2FS_OPTION(sbi).discard_unit == DISCARD_UNIT_BLOCK) 1943 seq_printf(seq, ",discard_unit=%s", "block"); 1944 else if (F2FS_OPTION(sbi).discard_unit == DISCARD_UNIT_SEGMENT) 1945 seq_printf(seq, ",discard_unit=%s", "segment"); 1946 else if (F2FS_OPTION(sbi).discard_unit == DISCARD_UNIT_SECTION) 1947 seq_printf(seq, ",discard_unit=%s", "section"); 1948 } else { 1949 seq_puts(seq, ",nodiscard"); 1950 } 1951 if (test_opt(sbi, NOHEAP)) 1952 seq_puts(seq, ",no_heap"); 1953 else 1954 seq_puts(seq, ",heap"); 1955 #ifdef CONFIG_F2FS_FS_XATTR 1956 if (test_opt(sbi, XATTR_USER)) 1957 seq_puts(seq, ",user_xattr"); 1958 else 1959 seq_puts(seq, ",nouser_xattr"); 1960 if (test_opt(sbi, INLINE_XATTR)) 1961 seq_puts(seq, ",inline_xattr"); 1962 else 1963 seq_puts(seq, ",noinline_xattr"); 1964 if (test_opt(sbi, INLINE_XATTR_SIZE)) 1965 seq_printf(seq, ",inline_xattr_size=%u", 1966 F2FS_OPTION(sbi).inline_xattr_size); 1967 #endif 1968 #ifdef CONFIG_F2FS_FS_POSIX_ACL 1969 if (test_opt(sbi, POSIX_ACL)) 1970 seq_puts(seq, ",acl"); 1971 else 1972 seq_puts(seq, ",noacl"); 1973 #endif 1974 if (test_opt(sbi, DISABLE_EXT_IDENTIFY)) 1975 seq_puts(seq, ",disable_ext_identify"); 1976 if (test_opt(sbi, INLINE_DATA)) 1977 seq_puts(seq, ",inline_data"); 1978 else 1979 seq_puts(seq, ",noinline_data"); 1980 if (test_opt(sbi, INLINE_DENTRY)) 1981 seq_puts(seq, ",inline_dentry"); 1982 else 1983 seq_puts(seq, ",noinline_dentry"); 1984 if (test_opt(sbi, FLUSH_MERGE)) 1985 seq_puts(seq, ",flush_merge"); 1986 else 1987 seq_puts(seq, ",noflush_merge"); 1988 if (test_opt(sbi, NOBARRIER)) 1989 seq_puts(seq, ",nobarrier"); 1990 else 1991 seq_puts(seq, ",barrier"); 1992 if (test_opt(sbi, FASTBOOT)) 1993 seq_puts(seq, ",fastboot"); 1994 if (test_opt(sbi, READ_EXTENT_CACHE)) 1995 seq_puts(seq, ",extent_cache"); 1996 else 1997 seq_puts(seq, ",noextent_cache"); 1998 if (test_opt(sbi, AGE_EXTENT_CACHE)) 1999 seq_puts(seq, ",age_extent_cache"); 2000 if (test_opt(sbi, DATA_FLUSH)) 2001 seq_puts(seq, ",data_flush"); 2002 2003 seq_puts(seq, ",mode="); 2004 if (F2FS_OPTION(sbi).fs_mode == FS_MODE_ADAPTIVE) 2005 seq_puts(seq, "adaptive"); 2006 else if (F2FS_OPTION(sbi).fs_mode == FS_MODE_LFS) 2007 seq_puts(seq, "lfs"); 2008 else if (F2FS_OPTION(sbi).fs_mode == FS_MODE_FRAGMENT_SEG) 2009 seq_puts(seq, "fragment:segment"); 2010 else if (F2FS_OPTION(sbi).fs_mode == FS_MODE_FRAGMENT_BLK) 2011 seq_puts(seq, "fragment:block"); 2012 seq_printf(seq, ",active_logs=%u", F2FS_OPTION(sbi).active_logs); 2013 if (test_opt(sbi, RESERVE_ROOT)) 2014 seq_printf(seq, ",reserve_root=%u,resuid=%u,resgid=%u", 2015 F2FS_OPTION(sbi).root_reserved_blocks, 2016 from_kuid_munged(&init_user_ns, 2017 F2FS_OPTION(sbi).s_resuid), 2018 from_kgid_munged(&init_user_ns, 2019 F2FS_OPTION(sbi).s_resgid)); 2020 if (F2FS_IO_SIZE_BITS(sbi)) 2021 seq_printf(seq, ",io_bits=%u", 2022 F2FS_OPTION(sbi).write_io_size_bits); 2023 #ifdef CONFIG_F2FS_FAULT_INJECTION 2024 if (test_opt(sbi, FAULT_INJECTION)) { 2025 seq_printf(seq, ",fault_injection=%u", 2026 F2FS_OPTION(sbi).fault_info.inject_rate); 2027 seq_printf(seq, ",fault_type=%u", 2028 F2FS_OPTION(sbi).fault_info.inject_type); 2029 } 2030 #endif 2031 #ifdef CONFIG_QUOTA 2032 if (test_opt(sbi, QUOTA)) 2033 seq_puts(seq, ",quota"); 2034 if (test_opt(sbi, USRQUOTA)) 2035 seq_puts(seq, ",usrquota"); 2036 if (test_opt(sbi, GRPQUOTA)) 2037 seq_puts(seq, ",grpquota"); 2038 if (test_opt(sbi, PRJQUOTA)) 2039 seq_puts(seq, ",prjquota"); 2040 #endif 2041 f2fs_show_quota_options(seq, sbi->sb); 2042 2043 fscrypt_show_test_dummy_encryption(seq, ',', sbi->sb); 2044 2045 if (sbi->sb->s_flags & SB_INLINECRYPT) 2046 seq_puts(seq, ",inlinecrypt"); 2047 2048 if (F2FS_OPTION(sbi).alloc_mode == ALLOC_MODE_DEFAULT) 2049 seq_printf(seq, ",alloc_mode=%s", "default"); 2050 else if (F2FS_OPTION(sbi).alloc_mode == ALLOC_MODE_REUSE) 2051 seq_printf(seq, ",alloc_mode=%s", "reuse"); 2052 2053 if (test_opt(sbi, DISABLE_CHECKPOINT)) 2054 seq_printf(seq, ",checkpoint=disable:%u", 2055 F2FS_OPTION(sbi).unusable_cap); 2056 if (test_opt(sbi, MERGE_CHECKPOINT)) 2057 seq_puts(seq, ",checkpoint_merge"); 2058 else 2059 seq_puts(seq, ",nocheckpoint_merge"); 2060 if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_POSIX) 2061 seq_printf(seq, ",fsync_mode=%s", "posix"); 2062 else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_STRICT) 2063 seq_printf(seq, ",fsync_mode=%s", "strict"); 2064 else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_NOBARRIER) 2065 seq_printf(seq, ",fsync_mode=%s", "nobarrier"); 2066 2067 #ifdef CONFIG_F2FS_FS_COMPRESSION 2068 f2fs_show_compress_options(seq, sbi->sb); 2069 #endif 2070 2071 if (test_opt(sbi, ATGC)) 2072 seq_puts(seq, ",atgc"); 2073 2074 if (F2FS_OPTION(sbi).memory_mode == MEMORY_MODE_NORMAL) 2075 seq_printf(seq, ",memory=%s", "normal"); 2076 else if (F2FS_OPTION(sbi).memory_mode == MEMORY_MODE_LOW) 2077 seq_printf(seq, ",memory=%s", "low"); 2078 2079 if (F2FS_OPTION(sbi).errors == MOUNT_ERRORS_READONLY) 2080 seq_printf(seq, ",errors=%s", "remount-ro"); 2081 else if (F2FS_OPTION(sbi).errors == MOUNT_ERRORS_CONTINUE) 2082 seq_printf(seq, ",errors=%s", "continue"); 2083 else if (F2FS_OPTION(sbi).errors == MOUNT_ERRORS_PANIC) 2084 seq_printf(seq, ",errors=%s", "panic"); 2085 2086 return 0; 2087 } 2088 2089 static void default_options(struct f2fs_sb_info *sbi) 2090 { 2091 /* init some FS parameters */ 2092 if (f2fs_sb_has_readonly(sbi)) 2093 F2FS_OPTION(sbi).active_logs = NR_CURSEG_RO_TYPE; 2094 else 2095 F2FS_OPTION(sbi).active_logs = NR_CURSEG_PERSIST_TYPE; 2096 2097 F2FS_OPTION(sbi).inline_xattr_size = DEFAULT_INLINE_XATTR_ADDRS; 2098 if (le32_to_cpu(F2FS_RAW_SUPER(sbi)->segment_count_main) <= 2099 SMALL_VOLUME_SEGMENTS) 2100 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_REUSE; 2101 else 2102 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_DEFAULT; 2103 F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_POSIX; 2104 F2FS_OPTION(sbi).s_resuid = make_kuid(&init_user_ns, F2FS_DEF_RESUID); 2105 F2FS_OPTION(sbi).s_resgid = make_kgid(&init_user_ns, F2FS_DEF_RESGID); 2106 if (f2fs_sb_has_compression(sbi)) { 2107 F2FS_OPTION(sbi).compress_algorithm = COMPRESS_LZ4; 2108 F2FS_OPTION(sbi).compress_log_size = MIN_COMPRESS_LOG_SIZE; 2109 F2FS_OPTION(sbi).compress_ext_cnt = 0; 2110 F2FS_OPTION(sbi).compress_mode = COMPR_MODE_FS; 2111 } 2112 F2FS_OPTION(sbi).bggc_mode = BGGC_MODE_ON; 2113 F2FS_OPTION(sbi).memory_mode = MEMORY_MODE_NORMAL; 2114 F2FS_OPTION(sbi).errors = MOUNT_ERRORS_CONTINUE; 2115 2116 sbi->sb->s_flags &= ~SB_INLINECRYPT; 2117 2118 set_opt(sbi, INLINE_XATTR); 2119 set_opt(sbi, INLINE_DATA); 2120 set_opt(sbi, INLINE_DENTRY); 2121 set_opt(sbi, READ_EXTENT_CACHE); 2122 set_opt(sbi, NOHEAP); 2123 clear_opt(sbi, DISABLE_CHECKPOINT); 2124 set_opt(sbi, MERGE_CHECKPOINT); 2125 F2FS_OPTION(sbi).unusable_cap = 0; 2126 sbi->sb->s_flags |= SB_LAZYTIME; 2127 if (!f2fs_is_readonly(sbi)) 2128 set_opt(sbi, FLUSH_MERGE); 2129 if (f2fs_hw_support_discard(sbi) || f2fs_hw_should_discard(sbi)) 2130 set_opt(sbi, DISCARD); 2131 if (f2fs_sb_has_blkzoned(sbi)) { 2132 F2FS_OPTION(sbi).fs_mode = FS_MODE_LFS; 2133 F2FS_OPTION(sbi).discard_unit = DISCARD_UNIT_SECTION; 2134 } else { 2135 F2FS_OPTION(sbi).fs_mode = FS_MODE_ADAPTIVE; 2136 F2FS_OPTION(sbi).discard_unit = DISCARD_UNIT_BLOCK; 2137 } 2138 2139 #ifdef CONFIG_F2FS_FS_XATTR 2140 set_opt(sbi, XATTR_USER); 2141 #endif 2142 #ifdef CONFIG_F2FS_FS_POSIX_ACL 2143 set_opt(sbi, POSIX_ACL); 2144 #endif 2145 2146 f2fs_build_fault_attr(sbi, 0, 0); 2147 } 2148 2149 #ifdef CONFIG_QUOTA 2150 static int f2fs_enable_quotas(struct super_block *sb); 2151 #endif 2152 2153 static int f2fs_disable_checkpoint(struct f2fs_sb_info *sbi) 2154 { 2155 unsigned int s_flags = sbi->sb->s_flags; 2156 struct cp_control cpc; 2157 unsigned int gc_mode = sbi->gc_mode; 2158 int err = 0; 2159 int ret; 2160 block_t unusable; 2161 2162 if (s_flags & SB_RDONLY) { 2163 f2fs_err(sbi, "checkpoint=disable on readonly fs"); 2164 return -EINVAL; 2165 } 2166 sbi->sb->s_flags |= SB_ACTIVE; 2167 2168 /* check if we need more GC first */ 2169 unusable = f2fs_get_unusable_blocks(sbi); 2170 if (!f2fs_disable_cp_again(sbi, unusable)) 2171 goto skip_gc; 2172 2173 f2fs_update_time(sbi, DISABLE_TIME); 2174 2175 sbi->gc_mode = GC_URGENT_HIGH; 2176 2177 while (!f2fs_time_over(sbi, DISABLE_TIME)) { 2178 struct f2fs_gc_control gc_control = { 2179 .victim_segno = NULL_SEGNO, 2180 .init_gc_type = FG_GC, 2181 .should_migrate_blocks = false, 2182 .err_gc_skipped = true, 2183 .nr_free_secs = 1 }; 2184 2185 f2fs_down_write(&sbi->gc_lock); 2186 err = f2fs_gc(sbi, &gc_control); 2187 if (err == -ENODATA) { 2188 err = 0; 2189 break; 2190 } 2191 if (err && err != -EAGAIN) 2192 break; 2193 } 2194 2195 ret = sync_filesystem(sbi->sb); 2196 if (ret || err) { 2197 err = ret ? ret : err; 2198 goto restore_flag; 2199 } 2200 2201 unusable = f2fs_get_unusable_blocks(sbi); 2202 if (f2fs_disable_cp_again(sbi, unusable)) { 2203 err = -EAGAIN; 2204 goto restore_flag; 2205 } 2206 2207 skip_gc: 2208 f2fs_down_write(&sbi->gc_lock); 2209 cpc.reason = CP_PAUSE; 2210 set_sbi_flag(sbi, SBI_CP_DISABLED); 2211 err = f2fs_write_checkpoint(sbi, &cpc); 2212 if (err) 2213 goto out_unlock; 2214 2215 spin_lock(&sbi->stat_lock); 2216 sbi->unusable_block_count = unusable; 2217 spin_unlock(&sbi->stat_lock); 2218 2219 out_unlock: 2220 f2fs_up_write(&sbi->gc_lock); 2221 restore_flag: 2222 sbi->gc_mode = gc_mode; 2223 sbi->sb->s_flags = s_flags; /* Restore SB_RDONLY status */ 2224 return err; 2225 } 2226 2227 static void f2fs_enable_checkpoint(struct f2fs_sb_info *sbi) 2228 { 2229 int retry = DEFAULT_RETRY_IO_COUNT; 2230 2231 /* we should flush all the data to keep data consistency */ 2232 do { 2233 sync_inodes_sb(sbi->sb); 2234 f2fs_io_schedule_timeout(DEFAULT_IO_TIMEOUT); 2235 } while (get_pages(sbi, F2FS_DIRTY_DATA) && retry--); 2236 2237 if (unlikely(retry < 0)) 2238 f2fs_warn(sbi, "checkpoint=enable has some unwritten data."); 2239 2240 f2fs_down_write(&sbi->gc_lock); 2241 f2fs_dirty_to_prefree(sbi); 2242 2243 clear_sbi_flag(sbi, SBI_CP_DISABLED); 2244 set_sbi_flag(sbi, SBI_IS_DIRTY); 2245 f2fs_up_write(&sbi->gc_lock); 2246 2247 f2fs_sync_fs(sbi->sb, 1); 2248 2249 /* Let's ensure there's no pending checkpoint anymore */ 2250 f2fs_flush_ckpt_thread(sbi); 2251 } 2252 2253 static int f2fs_remount(struct super_block *sb, int *flags, char *data) 2254 { 2255 struct f2fs_sb_info *sbi = F2FS_SB(sb); 2256 struct f2fs_mount_info org_mount_opt; 2257 unsigned long old_sb_flags; 2258 int err; 2259 bool need_restart_gc = false, need_stop_gc = false; 2260 bool need_restart_ckpt = false, need_stop_ckpt = false; 2261 bool need_restart_flush = false, need_stop_flush = false; 2262 bool need_restart_discard = false, need_stop_discard = false; 2263 bool no_read_extent_cache = !test_opt(sbi, READ_EXTENT_CACHE); 2264 bool no_age_extent_cache = !test_opt(sbi, AGE_EXTENT_CACHE); 2265 bool enable_checkpoint = !test_opt(sbi, DISABLE_CHECKPOINT); 2266 bool no_io_align = !F2FS_IO_ALIGNED(sbi); 2267 bool no_atgc = !test_opt(sbi, ATGC); 2268 bool no_discard = !test_opt(sbi, DISCARD); 2269 bool no_compress_cache = !test_opt(sbi, COMPRESS_CACHE); 2270 bool block_unit_discard = f2fs_block_unit_discard(sbi); 2271 #ifdef CONFIG_QUOTA 2272 int i, j; 2273 #endif 2274 2275 /* 2276 * Save the old mount options in case we 2277 * need to restore them. 2278 */ 2279 org_mount_opt = sbi->mount_opt; 2280 old_sb_flags = sb->s_flags; 2281 2282 #ifdef CONFIG_QUOTA 2283 org_mount_opt.s_jquota_fmt = F2FS_OPTION(sbi).s_jquota_fmt; 2284 for (i = 0; i < MAXQUOTAS; i++) { 2285 if (F2FS_OPTION(sbi).s_qf_names[i]) { 2286 org_mount_opt.s_qf_names[i] = 2287 kstrdup(F2FS_OPTION(sbi).s_qf_names[i], 2288 GFP_KERNEL); 2289 if (!org_mount_opt.s_qf_names[i]) { 2290 for (j = 0; j < i; j++) 2291 kfree(org_mount_opt.s_qf_names[j]); 2292 return -ENOMEM; 2293 } 2294 } else { 2295 org_mount_opt.s_qf_names[i] = NULL; 2296 } 2297 } 2298 #endif 2299 2300 /* recover superblocks we couldn't write due to previous RO mount */ 2301 if (!(*flags & SB_RDONLY) && is_sbi_flag_set(sbi, SBI_NEED_SB_WRITE)) { 2302 err = f2fs_commit_super(sbi, false); 2303 f2fs_info(sbi, "Try to recover all the superblocks, ret: %d", 2304 err); 2305 if (!err) 2306 clear_sbi_flag(sbi, SBI_NEED_SB_WRITE); 2307 } 2308 2309 default_options(sbi); 2310 2311 /* parse mount options */ 2312 err = parse_options(sb, data, true); 2313 if (err) 2314 goto restore_opts; 2315 2316 /* flush outstanding errors before changing fs state */ 2317 flush_work(&sbi->s_error_work); 2318 2319 /* 2320 * Previous and new state of filesystem is RO, 2321 * so skip checking GC and FLUSH_MERGE conditions. 2322 */ 2323 if (f2fs_readonly(sb) && (*flags & SB_RDONLY)) 2324 goto skip; 2325 2326 if (f2fs_dev_is_readonly(sbi) && !(*flags & SB_RDONLY)) { 2327 err = -EROFS; 2328 goto restore_opts; 2329 } 2330 2331 #ifdef CONFIG_QUOTA 2332 if (!f2fs_readonly(sb) && (*flags & SB_RDONLY)) { 2333 err = dquot_suspend(sb, -1); 2334 if (err < 0) 2335 goto restore_opts; 2336 } else if (f2fs_readonly(sb) && !(*flags & SB_RDONLY)) { 2337 /* dquot_resume needs RW */ 2338 sb->s_flags &= ~SB_RDONLY; 2339 if (sb_any_quota_suspended(sb)) { 2340 dquot_resume(sb, -1); 2341 } else if (f2fs_sb_has_quota_ino(sbi)) { 2342 err = f2fs_enable_quotas(sb); 2343 if (err) 2344 goto restore_opts; 2345 } 2346 } 2347 #endif 2348 if (f2fs_lfs_mode(sbi) && !IS_F2FS_IPU_DISABLE(sbi)) { 2349 err = -EINVAL; 2350 f2fs_warn(sbi, "LFS is not compatible with IPU"); 2351 goto restore_opts; 2352 } 2353 2354 /* disallow enable atgc dynamically */ 2355 if (no_atgc == !!test_opt(sbi, ATGC)) { 2356 err = -EINVAL; 2357 f2fs_warn(sbi, "switch atgc option is not allowed"); 2358 goto restore_opts; 2359 } 2360 2361 /* disallow enable/disable extent_cache dynamically */ 2362 if (no_read_extent_cache == !!test_opt(sbi, READ_EXTENT_CACHE)) { 2363 err = -EINVAL; 2364 f2fs_warn(sbi, "switch extent_cache option is not allowed"); 2365 goto restore_opts; 2366 } 2367 /* disallow enable/disable age extent_cache dynamically */ 2368 if (no_age_extent_cache == !!test_opt(sbi, AGE_EXTENT_CACHE)) { 2369 err = -EINVAL; 2370 f2fs_warn(sbi, "switch age_extent_cache option is not allowed"); 2371 goto restore_opts; 2372 } 2373 2374 if (no_io_align == !!F2FS_IO_ALIGNED(sbi)) { 2375 err = -EINVAL; 2376 f2fs_warn(sbi, "switch io_bits option is not allowed"); 2377 goto restore_opts; 2378 } 2379 2380 if (no_compress_cache == !!test_opt(sbi, COMPRESS_CACHE)) { 2381 err = -EINVAL; 2382 f2fs_warn(sbi, "switch compress_cache option is not allowed"); 2383 goto restore_opts; 2384 } 2385 2386 if (block_unit_discard != f2fs_block_unit_discard(sbi)) { 2387 err = -EINVAL; 2388 f2fs_warn(sbi, "switch discard_unit option is not allowed"); 2389 goto restore_opts; 2390 } 2391 2392 if ((*flags & SB_RDONLY) && test_opt(sbi, DISABLE_CHECKPOINT)) { 2393 err = -EINVAL; 2394 f2fs_warn(sbi, "disabling checkpoint not compatible with read-only"); 2395 goto restore_opts; 2396 } 2397 2398 /* 2399 * We stop the GC thread if FS is mounted as RO 2400 * or if background_gc = off is passed in mount 2401 * option. Also sync the filesystem. 2402 */ 2403 if ((*flags & SB_RDONLY) || 2404 (F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_OFF && 2405 !test_opt(sbi, GC_MERGE))) { 2406 if (sbi->gc_thread) { 2407 f2fs_stop_gc_thread(sbi); 2408 need_restart_gc = true; 2409 } 2410 } else if (!sbi->gc_thread) { 2411 err = f2fs_start_gc_thread(sbi); 2412 if (err) 2413 goto restore_opts; 2414 need_stop_gc = true; 2415 } 2416 2417 if (*flags & SB_RDONLY) { 2418 sync_inodes_sb(sb); 2419 2420 set_sbi_flag(sbi, SBI_IS_DIRTY); 2421 set_sbi_flag(sbi, SBI_IS_CLOSE); 2422 f2fs_sync_fs(sb, 1); 2423 clear_sbi_flag(sbi, SBI_IS_CLOSE); 2424 } 2425 2426 if ((*flags & SB_RDONLY) || test_opt(sbi, DISABLE_CHECKPOINT) || 2427 !test_opt(sbi, MERGE_CHECKPOINT)) { 2428 f2fs_stop_ckpt_thread(sbi); 2429 need_restart_ckpt = true; 2430 } else { 2431 /* Flush if the prevous checkpoint, if exists. */ 2432 f2fs_flush_ckpt_thread(sbi); 2433 2434 err = f2fs_start_ckpt_thread(sbi); 2435 if (err) { 2436 f2fs_err(sbi, 2437 "Failed to start F2FS issue_checkpoint_thread (%d)", 2438 err); 2439 goto restore_gc; 2440 } 2441 need_stop_ckpt = true; 2442 } 2443 2444 /* 2445 * We stop issue flush thread if FS is mounted as RO 2446 * or if flush_merge is not passed in mount option. 2447 */ 2448 if ((*flags & SB_RDONLY) || !test_opt(sbi, FLUSH_MERGE)) { 2449 clear_opt(sbi, FLUSH_MERGE); 2450 f2fs_destroy_flush_cmd_control(sbi, false); 2451 need_restart_flush = true; 2452 } else { 2453 err = f2fs_create_flush_cmd_control(sbi); 2454 if (err) 2455 goto restore_ckpt; 2456 need_stop_flush = true; 2457 } 2458 2459 if (no_discard == !!test_opt(sbi, DISCARD)) { 2460 if (test_opt(sbi, DISCARD)) { 2461 err = f2fs_start_discard_thread(sbi); 2462 if (err) 2463 goto restore_flush; 2464 need_stop_discard = true; 2465 } else { 2466 f2fs_stop_discard_thread(sbi); 2467 f2fs_issue_discard_timeout(sbi); 2468 need_restart_discard = true; 2469 } 2470 } 2471 2472 if (enable_checkpoint == !!test_opt(sbi, DISABLE_CHECKPOINT)) { 2473 if (test_opt(sbi, DISABLE_CHECKPOINT)) { 2474 err = f2fs_disable_checkpoint(sbi); 2475 if (err) 2476 goto restore_discard; 2477 } else { 2478 f2fs_enable_checkpoint(sbi); 2479 } 2480 } 2481 2482 skip: 2483 #ifdef CONFIG_QUOTA 2484 /* Release old quota file names */ 2485 for (i = 0; i < MAXQUOTAS; i++) 2486 kfree(org_mount_opt.s_qf_names[i]); 2487 #endif 2488 /* Update the POSIXACL Flag */ 2489 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) | 2490 (test_opt(sbi, POSIX_ACL) ? SB_POSIXACL : 0); 2491 2492 limit_reserve_root(sbi); 2493 adjust_unusable_cap_perc(sbi); 2494 *flags = (*flags & ~SB_LAZYTIME) | (sb->s_flags & SB_LAZYTIME); 2495 return 0; 2496 restore_discard: 2497 if (need_restart_discard) { 2498 if (f2fs_start_discard_thread(sbi)) 2499 f2fs_warn(sbi, "discard has been stopped"); 2500 } else if (need_stop_discard) { 2501 f2fs_stop_discard_thread(sbi); 2502 } 2503 restore_flush: 2504 if (need_restart_flush) { 2505 if (f2fs_create_flush_cmd_control(sbi)) 2506 f2fs_warn(sbi, "background flush thread has stopped"); 2507 } else if (need_stop_flush) { 2508 clear_opt(sbi, FLUSH_MERGE); 2509 f2fs_destroy_flush_cmd_control(sbi, false); 2510 } 2511 restore_ckpt: 2512 if (need_restart_ckpt) { 2513 if (f2fs_start_ckpt_thread(sbi)) 2514 f2fs_warn(sbi, "background ckpt thread has stopped"); 2515 } else if (need_stop_ckpt) { 2516 f2fs_stop_ckpt_thread(sbi); 2517 } 2518 restore_gc: 2519 if (need_restart_gc) { 2520 if (f2fs_start_gc_thread(sbi)) 2521 f2fs_warn(sbi, "background gc thread has stopped"); 2522 } else if (need_stop_gc) { 2523 f2fs_stop_gc_thread(sbi); 2524 } 2525 restore_opts: 2526 #ifdef CONFIG_QUOTA 2527 F2FS_OPTION(sbi).s_jquota_fmt = org_mount_opt.s_jquota_fmt; 2528 for (i = 0; i < MAXQUOTAS; i++) { 2529 kfree(F2FS_OPTION(sbi).s_qf_names[i]); 2530 F2FS_OPTION(sbi).s_qf_names[i] = org_mount_opt.s_qf_names[i]; 2531 } 2532 #endif 2533 sbi->mount_opt = org_mount_opt; 2534 sb->s_flags = old_sb_flags; 2535 return err; 2536 } 2537 2538 #ifdef CONFIG_QUOTA 2539 static bool f2fs_need_recovery(struct f2fs_sb_info *sbi) 2540 { 2541 /* need to recovery orphan */ 2542 if (is_set_ckpt_flags(sbi, CP_ORPHAN_PRESENT_FLAG)) 2543 return true; 2544 /* need to recovery data */ 2545 if (test_opt(sbi, DISABLE_ROLL_FORWARD)) 2546 return false; 2547 if (test_opt(sbi, NORECOVERY)) 2548 return false; 2549 return !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG); 2550 } 2551 2552 static bool f2fs_recover_quota_begin(struct f2fs_sb_info *sbi) 2553 { 2554 bool readonly = f2fs_readonly(sbi->sb); 2555 2556 if (!f2fs_need_recovery(sbi)) 2557 return false; 2558 2559 /* it doesn't need to check f2fs_sb_has_readonly() */ 2560 if (f2fs_hw_is_readonly(sbi)) 2561 return false; 2562 2563 if (readonly) { 2564 sbi->sb->s_flags &= ~SB_RDONLY; 2565 set_sbi_flag(sbi, SBI_IS_WRITABLE); 2566 } 2567 2568 /* 2569 * Turn on quotas which were not enabled for read-only mounts if 2570 * filesystem has quota feature, so that they are updated correctly. 2571 */ 2572 return f2fs_enable_quota_files(sbi, readonly); 2573 } 2574 2575 static void f2fs_recover_quota_end(struct f2fs_sb_info *sbi, 2576 bool quota_enabled) 2577 { 2578 if (quota_enabled) 2579 f2fs_quota_off_umount(sbi->sb); 2580 2581 if (is_sbi_flag_set(sbi, SBI_IS_WRITABLE)) { 2582 clear_sbi_flag(sbi, SBI_IS_WRITABLE); 2583 sbi->sb->s_flags |= SB_RDONLY; 2584 } 2585 } 2586 2587 /* Read data from quotafile */ 2588 static ssize_t f2fs_quota_read(struct super_block *sb, int type, char *data, 2589 size_t len, loff_t off) 2590 { 2591 struct inode *inode = sb_dqopt(sb)->files[type]; 2592 struct address_space *mapping = inode->i_mapping; 2593 block_t blkidx = F2FS_BYTES_TO_BLK(off); 2594 int offset = off & (sb->s_blocksize - 1); 2595 int tocopy; 2596 size_t toread; 2597 loff_t i_size = i_size_read(inode); 2598 struct page *page; 2599 2600 if (off > i_size) 2601 return 0; 2602 2603 if (off + len > i_size) 2604 len = i_size - off; 2605 toread = len; 2606 while (toread > 0) { 2607 tocopy = min_t(unsigned long, sb->s_blocksize - offset, toread); 2608 repeat: 2609 page = read_cache_page_gfp(mapping, blkidx, GFP_NOFS); 2610 if (IS_ERR(page)) { 2611 if (PTR_ERR(page) == -ENOMEM) { 2612 memalloc_retry_wait(GFP_NOFS); 2613 goto repeat; 2614 } 2615 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR); 2616 return PTR_ERR(page); 2617 } 2618 2619 lock_page(page); 2620 2621 if (unlikely(page->mapping != mapping)) { 2622 f2fs_put_page(page, 1); 2623 goto repeat; 2624 } 2625 if (unlikely(!PageUptodate(page))) { 2626 f2fs_put_page(page, 1); 2627 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR); 2628 return -EIO; 2629 } 2630 2631 memcpy_from_page(data, page, offset, tocopy); 2632 f2fs_put_page(page, 1); 2633 2634 offset = 0; 2635 toread -= tocopy; 2636 data += tocopy; 2637 blkidx++; 2638 } 2639 return len; 2640 } 2641 2642 /* Write to quotafile */ 2643 static ssize_t f2fs_quota_write(struct super_block *sb, int type, 2644 const char *data, size_t len, loff_t off) 2645 { 2646 struct inode *inode = sb_dqopt(sb)->files[type]; 2647 struct address_space *mapping = inode->i_mapping; 2648 const struct address_space_operations *a_ops = mapping->a_ops; 2649 int offset = off & (sb->s_blocksize - 1); 2650 size_t towrite = len; 2651 struct page *page; 2652 void *fsdata = NULL; 2653 int err = 0; 2654 int tocopy; 2655 2656 while (towrite > 0) { 2657 tocopy = min_t(unsigned long, sb->s_blocksize - offset, 2658 towrite); 2659 retry: 2660 err = a_ops->write_begin(NULL, mapping, off, tocopy, 2661 &page, &fsdata); 2662 if (unlikely(err)) { 2663 if (err == -ENOMEM) { 2664 f2fs_io_schedule_timeout(DEFAULT_IO_TIMEOUT); 2665 goto retry; 2666 } 2667 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR); 2668 break; 2669 } 2670 2671 memcpy_to_page(page, offset, data, tocopy); 2672 2673 a_ops->write_end(NULL, mapping, off, tocopy, tocopy, 2674 page, fsdata); 2675 offset = 0; 2676 towrite -= tocopy; 2677 off += tocopy; 2678 data += tocopy; 2679 cond_resched(); 2680 } 2681 2682 if (len == towrite) 2683 return err; 2684 inode->i_mtime = inode->i_ctime = current_time(inode); 2685 f2fs_mark_inode_dirty_sync(inode, false); 2686 return len - towrite; 2687 } 2688 2689 int f2fs_dquot_initialize(struct inode *inode) 2690 { 2691 if (time_to_inject(F2FS_I_SB(inode), FAULT_DQUOT_INIT)) 2692 return -ESRCH; 2693 2694 return dquot_initialize(inode); 2695 } 2696 2697 static struct dquot **f2fs_get_dquots(struct inode *inode) 2698 { 2699 return F2FS_I(inode)->i_dquot; 2700 } 2701 2702 static qsize_t *f2fs_get_reserved_space(struct inode *inode) 2703 { 2704 return &F2FS_I(inode)->i_reserved_quota; 2705 } 2706 2707 static int f2fs_quota_on_mount(struct f2fs_sb_info *sbi, int type) 2708 { 2709 if (is_set_ckpt_flags(sbi, CP_QUOTA_NEED_FSCK_FLAG)) { 2710 f2fs_err(sbi, "quota sysfile may be corrupted, skip loading it"); 2711 return 0; 2712 } 2713 2714 return dquot_quota_on_mount(sbi->sb, F2FS_OPTION(sbi).s_qf_names[type], 2715 F2FS_OPTION(sbi).s_jquota_fmt, type); 2716 } 2717 2718 int f2fs_enable_quota_files(struct f2fs_sb_info *sbi, bool rdonly) 2719 { 2720 int enabled = 0; 2721 int i, err; 2722 2723 if (f2fs_sb_has_quota_ino(sbi) && rdonly) { 2724 err = f2fs_enable_quotas(sbi->sb); 2725 if (err) { 2726 f2fs_err(sbi, "Cannot turn on quota_ino: %d", err); 2727 return 0; 2728 } 2729 return 1; 2730 } 2731 2732 for (i = 0; i < MAXQUOTAS; i++) { 2733 if (F2FS_OPTION(sbi).s_qf_names[i]) { 2734 err = f2fs_quota_on_mount(sbi, i); 2735 if (!err) { 2736 enabled = 1; 2737 continue; 2738 } 2739 f2fs_err(sbi, "Cannot turn on quotas: %d on %d", 2740 err, i); 2741 } 2742 } 2743 return enabled; 2744 } 2745 2746 static int f2fs_quota_enable(struct super_block *sb, int type, int format_id, 2747 unsigned int flags) 2748 { 2749 struct inode *qf_inode; 2750 unsigned long qf_inum; 2751 int err; 2752 2753 BUG_ON(!f2fs_sb_has_quota_ino(F2FS_SB(sb))); 2754 2755 qf_inum = f2fs_qf_ino(sb, type); 2756 if (!qf_inum) 2757 return -EPERM; 2758 2759 qf_inode = f2fs_iget(sb, qf_inum); 2760 if (IS_ERR(qf_inode)) { 2761 f2fs_err(F2FS_SB(sb), "Bad quota inode %u:%lu", type, qf_inum); 2762 return PTR_ERR(qf_inode); 2763 } 2764 2765 /* Don't account quota for quota files to avoid recursion */ 2766 qf_inode->i_flags |= S_NOQUOTA; 2767 err = dquot_load_quota_inode(qf_inode, type, format_id, flags); 2768 iput(qf_inode); 2769 return err; 2770 } 2771 2772 static int f2fs_enable_quotas(struct super_block *sb) 2773 { 2774 struct f2fs_sb_info *sbi = F2FS_SB(sb); 2775 int type, err = 0; 2776 unsigned long qf_inum; 2777 bool quota_mopt[MAXQUOTAS] = { 2778 test_opt(sbi, USRQUOTA), 2779 test_opt(sbi, GRPQUOTA), 2780 test_opt(sbi, PRJQUOTA), 2781 }; 2782 2783 if (is_set_ckpt_flags(F2FS_SB(sb), CP_QUOTA_NEED_FSCK_FLAG)) { 2784 f2fs_err(sbi, "quota file may be corrupted, skip loading it"); 2785 return 0; 2786 } 2787 2788 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE; 2789 2790 for (type = 0; type < MAXQUOTAS; type++) { 2791 qf_inum = f2fs_qf_ino(sb, type); 2792 if (qf_inum) { 2793 err = f2fs_quota_enable(sb, type, QFMT_VFS_V1, 2794 DQUOT_USAGE_ENABLED | 2795 (quota_mopt[type] ? DQUOT_LIMITS_ENABLED : 0)); 2796 if (err) { 2797 f2fs_err(sbi, "Failed to enable quota tracking (type=%d, err=%d). Please run fsck to fix.", 2798 type, err); 2799 for (type--; type >= 0; type--) 2800 dquot_quota_off(sb, type); 2801 set_sbi_flag(F2FS_SB(sb), 2802 SBI_QUOTA_NEED_REPAIR); 2803 return err; 2804 } 2805 } 2806 } 2807 return 0; 2808 } 2809 2810 static int f2fs_quota_sync_file(struct f2fs_sb_info *sbi, int type) 2811 { 2812 struct quota_info *dqopt = sb_dqopt(sbi->sb); 2813 struct address_space *mapping = dqopt->files[type]->i_mapping; 2814 int ret = 0; 2815 2816 ret = dquot_writeback_dquots(sbi->sb, type); 2817 if (ret) 2818 goto out; 2819 2820 ret = filemap_fdatawrite(mapping); 2821 if (ret) 2822 goto out; 2823 2824 /* if we are using journalled quota */ 2825 if (is_journalled_quota(sbi)) 2826 goto out; 2827 2828 ret = filemap_fdatawait(mapping); 2829 2830 truncate_inode_pages(&dqopt->files[type]->i_data, 0); 2831 out: 2832 if (ret) 2833 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR); 2834 return ret; 2835 } 2836 2837 int f2fs_quota_sync(struct super_block *sb, int type) 2838 { 2839 struct f2fs_sb_info *sbi = F2FS_SB(sb); 2840 struct quota_info *dqopt = sb_dqopt(sb); 2841 int cnt; 2842 int ret = 0; 2843 2844 /* 2845 * Now when everything is written we can discard the pagecache so 2846 * that userspace sees the changes. 2847 */ 2848 for (cnt = 0; cnt < MAXQUOTAS; cnt++) { 2849 2850 if (type != -1 && cnt != type) 2851 continue; 2852 2853 if (!sb_has_quota_active(sb, cnt)) 2854 continue; 2855 2856 if (!f2fs_sb_has_quota_ino(sbi)) 2857 inode_lock(dqopt->files[cnt]); 2858 2859 /* 2860 * do_quotactl 2861 * f2fs_quota_sync 2862 * f2fs_down_read(quota_sem) 2863 * dquot_writeback_dquots() 2864 * f2fs_dquot_commit 2865 * block_operation 2866 * f2fs_down_read(quota_sem) 2867 */ 2868 f2fs_lock_op(sbi); 2869 f2fs_down_read(&sbi->quota_sem); 2870 2871 ret = f2fs_quota_sync_file(sbi, cnt); 2872 2873 f2fs_up_read(&sbi->quota_sem); 2874 f2fs_unlock_op(sbi); 2875 2876 if (!f2fs_sb_has_quota_ino(sbi)) 2877 inode_unlock(dqopt->files[cnt]); 2878 2879 if (ret) 2880 break; 2881 } 2882 return ret; 2883 } 2884 2885 static int f2fs_quota_on(struct super_block *sb, int type, int format_id, 2886 const struct path *path) 2887 { 2888 struct inode *inode; 2889 int err; 2890 2891 /* if quota sysfile exists, deny enabling quota with specific file */ 2892 if (f2fs_sb_has_quota_ino(F2FS_SB(sb))) { 2893 f2fs_err(F2FS_SB(sb), "quota sysfile already exists"); 2894 return -EBUSY; 2895 } 2896 2897 err = f2fs_quota_sync(sb, type); 2898 if (err) 2899 return err; 2900 2901 err = dquot_quota_on(sb, type, format_id, path); 2902 if (err) 2903 return err; 2904 2905 inode = d_inode(path->dentry); 2906 2907 inode_lock(inode); 2908 F2FS_I(inode)->i_flags |= F2FS_NOATIME_FL | F2FS_IMMUTABLE_FL; 2909 f2fs_set_inode_flags(inode); 2910 inode_unlock(inode); 2911 f2fs_mark_inode_dirty_sync(inode, false); 2912 2913 return 0; 2914 } 2915 2916 static int __f2fs_quota_off(struct super_block *sb, int type) 2917 { 2918 struct inode *inode = sb_dqopt(sb)->files[type]; 2919 int err; 2920 2921 if (!inode || !igrab(inode)) 2922 return dquot_quota_off(sb, type); 2923 2924 err = f2fs_quota_sync(sb, type); 2925 if (err) 2926 goto out_put; 2927 2928 err = dquot_quota_off(sb, type); 2929 if (err || f2fs_sb_has_quota_ino(F2FS_SB(sb))) 2930 goto out_put; 2931 2932 inode_lock(inode); 2933 F2FS_I(inode)->i_flags &= ~(F2FS_NOATIME_FL | F2FS_IMMUTABLE_FL); 2934 f2fs_set_inode_flags(inode); 2935 inode_unlock(inode); 2936 f2fs_mark_inode_dirty_sync(inode, false); 2937 out_put: 2938 iput(inode); 2939 return err; 2940 } 2941 2942 static int f2fs_quota_off(struct super_block *sb, int type) 2943 { 2944 struct f2fs_sb_info *sbi = F2FS_SB(sb); 2945 int err; 2946 2947 err = __f2fs_quota_off(sb, type); 2948 2949 /* 2950 * quotactl can shutdown journalled quota, result in inconsistence 2951 * between quota record and fs data by following updates, tag the 2952 * flag to let fsck be aware of it. 2953 */ 2954 if (is_journalled_quota(sbi)) 2955 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR); 2956 return err; 2957 } 2958 2959 void f2fs_quota_off_umount(struct super_block *sb) 2960 { 2961 int type; 2962 int err; 2963 2964 for (type = 0; type < MAXQUOTAS; type++) { 2965 err = __f2fs_quota_off(sb, type); 2966 if (err) { 2967 int ret = dquot_quota_off(sb, type); 2968 2969 f2fs_err(F2FS_SB(sb), "Fail to turn off disk quota (type: %d, err: %d, ret:%d), Please run fsck to fix it.", 2970 type, err, ret); 2971 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR); 2972 } 2973 } 2974 /* 2975 * In case of checkpoint=disable, we must flush quota blocks. 2976 * This can cause NULL exception for node_inode in end_io, since 2977 * put_super already dropped it. 2978 */ 2979 sync_filesystem(sb); 2980 } 2981 2982 static void f2fs_truncate_quota_inode_pages(struct super_block *sb) 2983 { 2984 struct quota_info *dqopt = sb_dqopt(sb); 2985 int type; 2986 2987 for (type = 0; type < MAXQUOTAS; type++) { 2988 if (!dqopt->files[type]) 2989 continue; 2990 f2fs_inode_synced(dqopt->files[type]); 2991 } 2992 } 2993 2994 static int f2fs_dquot_commit(struct dquot *dquot) 2995 { 2996 struct f2fs_sb_info *sbi = F2FS_SB(dquot->dq_sb); 2997 int ret; 2998 2999 f2fs_down_read_nested(&sbi->quota_sem, SINGLE_DEPTH_NESTING); 3000 ret = dquot_commit(dquot); 3001 if (ret < 0) 3002 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR); 3003 f2fs_up_read(&sbi->quota_sem); 3004 return ret; 3005 } 3006 3007 static int f2fs_dquot_acquire(struct dquot *dquot) 3008 { 3009 struct f2fs_sb_info *sbi = F2FS_SB(dquot->dq_sb); 3010 int ret; 3011 3012 f2fs_down_read(&sbi->quota_sem); 3013 ret = dquot_acquire(dquot); 3014 if (ret < 0) 3015 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR); 3016 f2fs_up_read(&sbi->quota_sem); 3017 return ret; 3018 } 3019 3020 static int f2fs_dquot_release(struct dquot *dquot) 3021 { 3022 struct f2fs_sb_info *sbi = F2FS_SB(dquot->dq_sb); 3023 int ret = dquot_release(dquot); 3024 3025 if (ret < 0) 3026 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR); 3027 return ret; 3028 } 3029 3030 static int f2fs_dquot_mark_dquot_dirty(struct dquot *dquot) 3031 { 3032 struct super_block *sb = dquot->dq_sb; 3033 struct f2fs_sb_info *sbi = F2FS_SB(sb); 3034 int ret = dquot_mark_dquot_dirty(dquot); 3035 3036 /* if we are using journalled quota */ 3037 if (is_journalled_quota(sbi)) 3038 set_sbi_flag(sbi, SBI_QUOTA_NEED_FLUSH); 3039 3040 return ret; 3041 } 3042 3043 static int f2fs_dquot_commit_info(struct super_block *sb, int type) 3044 { 3045 struct f2fs_sb_info *sbi = F2FS_SB(sb); 3046 int ret = dquot_commit_info(sb, type); 3047 3048 if (ret < 0) 3049 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR); 3050 return ret; 3051 } 3052 3053 static int f2fs_get_projid(struct inode *inode, kprojid_t *projid) 3054 { 3055 *projid = F2FS_I(inode)->i_projid; 3056 return 0; 3057 } 3058 3059 static const struct dquot_operations f2fs_quota_operations = { 3060 .get_reserved_space = f2fs_get_reserved_space, 3061 .write_dquot = f2fs_dquot_commit, 3062 .acquire_dquot = f2fs_dquot_acquire, 3063 .release_dquot = f2fs_dquot_release, 3064 .mark_dirty = f2fs_dquot_mark_dquot_dirty, 3065 .write_info = f2fs_dquot_commit_info, 3066 .alloc_dquot = dquot_alloc, 3067 .destroy_dquot = dquot_destroy, 3068 .get_projid = f2fs_get_projid, 3069 .get_next_id = dquot_get_next_id, 3070 }; 3071 3072 static const struct quotactl_ops f2fs_quotactl_ops = { 3073 .quota_on = f2fs_quota_on, 3074 .quota_off = f2fs_quota_off, 3075 .quota_sync = f2fs_quota_sync, 3076 .get_state = dquot_get_state, 3077 .set_info = dquot_set_dqinfo, 3078 .get_dqblk = dquot_get_dqblk, 3079 .set_dqblk = dquot_set_dqblk, 3080 .get_nextdqblk = dquot_get_next_dqblk, 3081 }; 3082 #else 3083 int f2fs_dquot_initialize(struct inode *inode) 3084 { 3085 return 0; 3086 } 3087 3088 int f2fs_quota_sync(struct super_block *sb, int type) 3089 { 3090 return 0; 3091 } 3092 3093 void f2fs_quota_off_umount(struct super_block *sb) 3094 { 3095 } 3096 #endif 3097 3098 static const struct super_operations f2fs_sops = { 3099 .alloc_inode = f2fs_alloc_inode, 3100 .free_inode = f2fs_free_inode, 3101 .drop_inode = f2fs_drop_inode, 3102 .write_inode = f2fs_write_inode, 3103 .dirty_inode = f2fs_dirty_inode, 3104 .show_options = f2fs_show_options, 3105 #ifdef CONFIG_QUOTA 3106 .quota_read = f2fs_quota_read, 3107 .quota_write = f2fs_quota_write, 3108 .get_dquots = f2fs_get_dquots, 3109 #endif 3110 .evict_inode = f2fs_evict_inode, 3111 .put_super = f2fs_put_super, 3112 .sync_fs = f2fs_sync_fs, 3113 .freeze_fs = f2fs_freeze, 3114 .unfreeze_fs = f2fs_unfreeze, 3115 .statfs = f2fs_statfs, 3116 .remount_fs = f2fs_remount, 3117 }; 3118 3119 #ifdef CONFIG_FS_ENCRYPTION 3120 static int f2fs_get_context(struct inode *inode, void *ctx, size_t len) 3121 { 3122 return f2fs_getxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION, 3123 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT, 3124 ctx, len, NULL); 3125 } 3126 3127 static int f2fs_set_context(struct inode *inode, const void *ctx, size_t len, 3128 void *fs_data) 3129 { 3130 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 3131 3132 /* 3133 * Encrypting the root directory is not allowed because fsck 3134 * expects lost+found directory to exist and remain unencrypted 3135 * if LOST_FOUND feature is enabled. 3136 * 3137 */ 3138 if (f2fs_sb_has_lost_found(sbi) && 3139 inode->i_ino == F2FS_ROOT_INO(sbi)) 3140 return -EPERM; 3141 3142 return f2fs_setxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION, 3143 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT, 3144 ctx, len, fs_data, XATTR_CREATE); 3145 } 3146 3147 static const union fscrypt_policy *f2fs_get_dummy_policy(struct super_block *sb) 3148 { 3149 return F2FS_OPTION(F2FS_SB(sb)).dummy_enc_policy.policy; 3150 } 3151 3152 static bool f2fs_has_stable_inodes(struct super_block *sb) 3153 { 3154 return true; 3155 } 3156 3157 static void f2fs_get_ino_and_lblk_bits(struct super_block *sb, 3158 int *ino_bits_ret, int *lblk_bits_ret) 3159 { 3160 *ino_bits_ret = 8 * sizeof(nid_t); 3161 *lblk_bits_ret = 8 * sizeof(block_t); 3162 } 3163 3164 static struct block_device **f2fs_get_devices(struct super_block *sb, 3165 unsigned int *num_devs) 3166 { 3167 struct f2fs_sb_info *sbi = F2FS_SB(sb); 3168 struct block_device **devs; 3169 int i; 3170 3171 if (!f2fs_is_multi_device(sbi)) 3172 return NULL; 3173 3174 devs = kmalloc_array(sbi->s_ndevs, sizeof(*devs), GFP_KERNEL); 3175 if (!devs) 3176 return ERR_PTR(-ENOMEM); 3177 3178 for (i = 0; i < sbi->s_ndevs; i++) 3179 devs[i] = FDEV(i).bdev; 3180 *num_devs = sbi->s_ndevs; 3181 return devs; 3182 } 3183 3184 static const struct fscrypt_operations f2fs_cryptops = { 3185 .key_prefix = "f2fs:", 3186 .get_context = f2fs_get_context, 3187 .set_context = f2fs_set_context, 3188 .get_dummy_policy = f2fs_get_dummy_policy, 3189 .empty_dir = f2fs_empty_dir, 3190 .has_stable_inodes = f2fs_has_stable_inodes, 3191 .get_ino_and_lblk_bits = f2fs_get_ino_and_lblk_bits, 3192 .get_devices = f2fs_get_devices, 3193 }; 3194 #endif 3195 3196 static struct inode *f2fs_nfs_get_inode(struct super_block *sb, 3197 u64 ino, u32 generation) 3198 { 3199 struct f2fs_sb_info *sbi = F2FS_SB(sb); 3200 struct inode *inode; 3201 3202 if (f2fs_check_nid_range(sbi, ino)) 3203 return ERR_PTR(-ESTALE); 3204 3205 /* 3206 * f2fs_iget isn't quite right if the inode is currently unallocated! 3207 * However f2fs_iget currently does appropriate checks to handle stale 3208 * inodes so everything is OK. 3209 */ 3210 inode = f2fs_iget(sb, ino); 3211 if (IS_ERR(inode)) 3212 return ERR_CAST(inode); 3213 if (unlikely(generation && inode->i_generation != generation)) { 3214 /* we didn't find the right inode.. */ 3215 iput(inode); 3216 return ERR_PTR(-ESTALE); 3217 } 3218 return inode; 3219 } 3220 3221 static struct dentry *f2fs_fh_to_dentry(struct super_block *sb, struct fid *fid, 3222 int fh_len, int fh_type) 3223 { 3224 return generic_fh_to_dentry(sb, fid, fh_len, fh_type, 3225 f2fs_nfs_get_inode); 3226 } 3227 3228 static struct dentry *f2fs_fh_to_parent(struct super_block *sb, struct fid *fid, 3229 int fh_len, int fh_type) 3230 { 3231 return generic_fh_to_parent(sb, fid, fh_len, fh_type, 3232 f2fs_nfs_get_inode); 3233 } 3234 3235 static const struct export_operations f2fs_export_ops = { 3236 .fh_to_dentry = f2fs_fh_to_dentry, 3237 .fh_to_parent = f2fs_fh_to_parent, 3238 .get_parent = f2fs_get_parent, 3239 }; 3240 3241 loff_t max_file_blocks(struct inode *inode) 3242 { 3243 loff_t result = 0; 3244 loff_t leaf_count; 3245 3246 /* 3247 * note: previously, result is equal to (DEF_ADDRS_PER_INODE - 3248 * DEFAULT_INLINE_XATTR_ADDRS), but now f2fs try to reserve more 3249 * space in inode.i_addr, it will be more safe to reassign 3250 * result as zero. 3251 */ 3252 3253 if (inode && f2fs_compressed_file(inode)) 3254 leaf_count = ADDRS_PER_BLOCK(inode); 3255 else 3256 leaf_count = DEF_ADDRS_PER_BLOCK; 3257 3258 /* two direct node blocks */ 3259 result += (leaf_count * 2); 3260 3261 /* two indirect node blocks */ 3262 leaf_count *= NIDS_PER_BLOCK; 3263 result += (leaf_count * 2); 3264 3265 /* one double indirect node block */ 3266 leaf_count *= NIDS_PER_BLOCK; 3267 result += leaf_count; 3268 3269 return result; 3270 } 3271 3272 static int __f2fs_commit_super(struct buffer_head *bh, 3273 struct f2fs_super_block *super) 3274 { 3275 lock_buffer(bh); 3276 if (super) 3277 memcpy(bh->b_data + F2FS_SUPER_OFFSET, super, sizeof(*super)); 3278 set_buffer_dirty(bh); 3279 unlock_buffer(bh); 3280 3281 /* it's rare case, we can do fua all the time */ 3282 return __sync_dirty_buffer(bh, REQ_SYNC | REQ_PREFLUSH | REQ_FUA); 3283 } 3284 3285 static inline bool sanity_check_area_boundary(struct f2fs_sb_info *sbi, 3286 struct buffer_head *bh) 3287 { 3288 struct f2fs_super_block *raw_super = (struct f2fs_super_block *) 3289 (bh->b_data + F2FS_SUPER_OFFSET); 3290 struct super_block *sb = sbi->sb; 3291 u32 segment0_blkaddr = le32_to_cpu(raw_super->segment0_blkaddr); 3292 u32 cp_blkaddr = le32_to_cpu(raw_super->cp_blkaddr); 3293 u32 sit_blkaddr = le32_to_cpu(raw_super->sit_blkaddr); 3294 u32 nat_blkaddr = le32_to_cpu(raw_super->nat_blkaddr); 3295 u32 ssa_blkaddr = le32_to_cpu(raw_super->ssa_blkaddr); 3296 u32 main_blkaddr = le32_to_cpu(raw_super->main_blkaddr); 3297 u32 segment_count_ckpt = le32_to_cpu(raw_super->segment_count_ckpt); 3298 u32 segment_count_sit = le32_to_cpu(raw_super->segment_count_sit); 3299 u32 segment_count_nat = le32_to_cpu(raw_super->segment_count_nat); 3300 u32 segment_count_ssa = le32_to_cpu(raw_super->segment_count_ssa); 3301 u32 segment_count_main = le32_to_cpu(raw_super->segment_count_main); 3302 u32 segment_count = le32_to_cpu(raw_super->segment_count); 3303 u32 log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg); 3304 u64 main_end_blkaddr = main_blkaddr + 3305 (segment_count_main << log_blocks_per_seg); 3306 u64 seg_end_blkaddr = segment0_blkaddr + 3307 (segment_count << log_blocks_per_seg); 3308 3309 if (segment0_blkaddr != cp_blkaddr) { 3310 f2fs_info(sbi, "Mismatch start address, segment0(%u) cp_blkaddr(%u)", 3311 segment0_blkaddr, cp_blkaddr); 3312 return true; 3313 } 3314 3315 if (cp_blkaddr + (segment_count_ckpt << log_blocks_per_seg) != 3316 sit_blkaddr) { 3317 f2fs_info(sbi, "Wrong CP boundary, start(%u) end(%u) blocks(%u)", 3318 cp_blkaddr, sit_blkaddr, 3319 segment_count_ckpt << log_blocks_per_seg); 3320 return true; 3321 } 3322 3323 if (sit_blkaddr + (segment_count_sit << log_blocks_per_seg) != 3324 nat_blkaddr) { 3325 f2fs_info(sbi, "Wrong SIT boundary, start(%u) end(%u) blocks(%u)", 3326 sit_blkaddr, nat_blkaddr, 3327 segment_count_sit << log_blocks_per_seg); 3328 return true; 3329 } 3330 3331 if (nat_blkaddr + (segment_count_nat << log_blocks_per_seg) != 3332 ssa_blkaddr) { 3333 f2fs_info(sbi, "Wrong NAT boundary, start(%u) end(%u) blocks(%u)", 3334 nat_blkaddr, ssa_blkaddr, 3335 segment_count_nat << log_blocks_per_seg); 3336 return true; 3337 } 3338 3339 if (ssa_blkaddr + (segment_count_ssa << log_blocks_per_seg) != 3340 main_blkaddr) { 3341 f2fs_info(sbi, "Wrong SSA boundary, start(%u) end(%u) blocks(%u)", 3342 ssa_blkaddr, main_blkaddr, 3343 segment_count_ssa << log_blocks_per_seg); 3344 return true; 3345 } 3346 3347 if (main_end_blkaddr > seg_end_blkaddr) { 3348 f2fs_info(sbi, "Wrong MAIN_AREA boundary, start(%u) end(%llu) block(%u)", 3349 main_blkaddr, seg_end_blkaddr, 3350 segment_count_main << log_blocks_per_seg); 3351 return true; 3352 } else if (main_end_blkaddr < seg_end_blkaddr) { 3353 int err = 0; 3354 char *res; 3355 3356 /* fix in-memory information all the time */ 3357 raw_super->segment_count = cpu_to_le32((main_end_blkaddr - 3358 segment0_blkaddr) >> log_blocks_per_seg); 3359 3360 if (f2fs_readonly(sb) || f2fs_hw_is_readonly(sbi)) { 3361 set_sbi_flag(sbi, SBI_NEED_SB_WRITE); 3362 res = "internally"; 3363 } else { 3364 err = __f2fs_commit_super(bh, NULL); 3365 res = err ? "failed" : "done"; 3366 } 3367 f2fs_info(sbi, "Fix alignment : %s, start(%u) end(%llu) block(%u)", 3368 res, main_blkaddr, seg_end_blkaddr, 3369 segment_count_main << log_blocks_per_seg); 3370 if (err) 3371 return true; 3372 } 3373 return false; 3374 } 3375 3376 static int sanity_check_raw_super(struct f2fs_sb_info *sbi, 3377 struct buffer_head *bh) 3378 { 3379 block_t segment_count, segs_per_sec, secs_per_zone, segment_count_main; 3380 block_t total_sections, blocks_per_seg; 3381 struct f2fs_super_block *raw_super = (struct f2fs_super_block *) 3382 (bh->b_data + F2FS_SUPER_OFFSET); 3383 size_t crc_offset = 0; 3384 __u32 crc = 0; 3385 3386 if (le32_to_cpu(raw_super->magic) != F2FS_SUPER_MAGIC) { 3387 f2fs_info(sbi, "Magic Mismatch, valid(0x%x) - read(0x%x)", 3388 F2FS_SUPER_MAGIC, le32_to_cpu(raw_super->magic)); 3389 return -EINVAL; 3390 } 3391 3392 /* Check checksum_offset and crc in superblock */ 3393 if (__F2FS_HAS_FEATURE(raw_super, F2FS_FEATURE_SB_CHKSUM)) { 3394 crc_offset = le32_to_cpu(raw_super->checksum_offset); 3395 if (crc_offset != 3396 offsetof(struct f2fs_super_block, crc)) { 3397 f2fs_info(sbi, "Invalid SB checksum offset: %zu", 3398 crc_offset); 3399 return -EFSCORRUPTED; 3400 } 3401 crc = le32_to_cpu(raw_super->crc); 3402 if (!f2fs_crc_valid(sbi, crc, raw_super, crc_offset)) { 3403 f2fs_info(sbi, "Invalid SB checksum value: %u", crc); 3404 return -EFSCORRUPTED; 3405 } 3406 } 3407 3408 /* Currently, support only 4KB block size */ 3409 if (le32_to_cpu(raw_super->log_blocksize) != F2FS_BLKSIZE_BITS) { 3410 f2fs_info(sbi, "Invalid log_blocksize (%u), supports only %u", 3411 le32_to_cpu(raw_super->log_blocksize), 3412 F2FS_BLKSIZE_BITS); 3413 return -EFSCORRUPTED; 3414 } 3415 3416 /* check log blocks per segment */ 3417 if (le32_to_cpu(raw_super->log_blocks_per_seg) != 9) { 3418 f2fs_info(sbi, "Invalid log blocks per segment (%u)", 3419 le32_to_cpu(raw_super->log_blocks_per_seg)); 3420 return -EFSCORRUPTED; 3421 } 3422 3423 /* Currently, support 512/1024/2048/4096 bytes sector size */ 3424 if (le32_to_cpu(raw_super->log_sectorsize) > 3425 F2FS_MAX_LOG_SECTOR_SIZE || 3426 le32_to_cpu(raw_super->log_sectorsize) < 3427 F2FS_MIN_LOG_SECTOR_SIZE) { 3428 f2fs_info(sbi, "Invalid log sectorsize (%u)", 3429 le32_to_cpu(raw_super->log_sectorsize)); 3430 return -EFSCORRUPTED; 3431 } 3432 if (le32_to_cpu(raw_super->log_sectors_per_block) + 3433 le32_to_cpu(raw_super->log_sectorsize) != 3434 F2FS_MAX_LOG_SECTOR_SIZE) { 3435 f2fs_info(sbi, "Invalid log sectors per block(%u) log sectorsize(%u)", 3436 le32_to_cpu(raw_super->log_sectors_per_block), 3437 le32_to_cpu(raw_super->log_sectorsize)); 3438 return -EFSCORRUPTED; 3439 } 3440 3441 segment_count = le32_to_cpu(raw_super->segment_count); 3442 segment_count_main = le32_to_cpu(raw_super->segment_count_main); 3443 segs_per_sec = le32_to_cpu(raw_super->segs_per_sec); 3444 secs_per_zone = le32_to_cpu(raw_super->secs_per_zone); 3445 total_sections = le32_to_cpu(raw_super->section_count); 3446 3447 /* blocks_per_seg should be 512, given the above check */ 3448 blocks_per_seg = BIT(le32_to_cpu(raw_super->log_blocks_per_seg)); 3449 3450 if (segment_count > F2FS_MAX_SEGMENT || 3451 segment_count < F2FS_MIN_SEGMENTS) { 3452 f2fs_info(sbi, "Invalid segment count (%u)", segment_count); 3453 return -EFSCORRUPTED; 3454 } 3455 3456 if (total_sections > segment_count_main || total_sections < 1 || 3457 segs_per_sec > segment_count || !segs_per_sec) { 3458 f2fs_info(sbi, "Invalid segment/section count (%u, %u x %u)", 3459 segment_count, total_sections, segs_per_sec); 3460 return -EFSCORRUPTED; 3461 } 3462 3463 if (segment_count_main != total_sections * segs_per_sec) { 3464 f2fs_info(sbi, "Invalid segment/section count (%u != %u * %u)", 3465 segment_count_main, total_sections, segs_per_sec); 3466 return -EFSCORRUPTED; 3467 } 3468 3469 if ((segment_count / segs_per_sec) < total_sections) { 3470 f2fs_info(sbi, "Small segment_count (%u < %u * %u)", 3471 segment_count, segs_per_sec, total_sections); 3472 return -EFSCORRUPTED; 3473 } 3474 3475 if (segment_count > (le64_to_cpu(raw_super->block_count) >> 9)) { 3476 f2fs_info(sbi, "Wrong segment_count / block_count (%u > %llu)", 3477 segment_count, le64_to_cpu(raw_super->block_count)); 3478 return -EFSCORRUPTED; 3479 } 3480 3481 if (RDEV(0).path[0]) { 3482 block_t dev_seg_count = le32_to_cpu(RDEV(0).total_segments); 3483 int i = 1; 3484 3485 while (i < MAX_DEVICES && RDEV(i).path[0]) { 3486 dev_seg_count += le32_to_cpu(RDEV(i).total_segments); 3487 i++; 3488 } 3489 if (segment_count != dev_seg_count) { 3490 f2fs_info(sbi, "Segment count (%u) mismatch with total segments from devices (%u)", 3491 segment_count, dev_seg_count); 3492 return -EFSCORRUPTED; 3493 } 3494 } else { 3495 if (__F2FS_HAS_FEATURE(raw_super, F2FS_FEATURE_BLKZONED) && 3496 !bdev_is_zoned(sbi->sb->s_bdev)) { 3497 f2fs_info(sbi, "Zoned block device path is missing"); 3498 return -EFSCORRUPTED; 3499 } 3500 } 3501 3502 if (secs_per_zone > total_sections || !secs_per_zone) { 3503 f2fs_info(sbi, "Wrong secs_per_zone / total_sections (%u, %u)", 3504 secs_per_zone, total_sections); 3505 return -EFSCORRUPTED; 3506 } 3507 if (le32_to_cpu(raw_super->extension_count) > F2FS_MAX_EXTENSION || 3508 raw_super->hot_ext_count > F2FS_MAX_EXTENSION || 3509 (le32_to_cpu(raw_super->extension_count) + 3510 raw_super->hot_ext_count) > F2FS_MAX_EXTENSION) { 3511 f2fs_info(sbi, "Corrupted extension count (%u + %u > %u)", 3512 le32_to_cpu(raw_super->extension_count), 3513 raw_super->hot_ext_count, 3514 F2FS_MAX_EXTENSION); 3515 return -EFSCORRUPTED; 3516 } 3517 3518 if (le32_to_cpu(raw_super->cp_payload) >= 3519 (blocks_per_seg - F2FS_CP_PACKS - 3520 NR_CURSEG_PERSIST_TYPE)) { 3521 f2fs_info(sbi, "Insane cp_payload (%u >= %u)", 3522 le32_to_cpu(raw_super->cp_payload), 3523 blocks_per_seg - F2FS_CP_PACKS - 3524 NR_CURSEG_PERSIST_TYPE); 3525 return -EFSCORRUPTED; 3526 } 3527 3528 /* check reserved ino info */ 3529 if (le32_to_cpu(raw_super->node_ino) != 1 || 3530 le32_to_cpu(raw_super->meta_ino) != 2 || 3531 le32_to_cpu(raw_super->root_ino) != 3) { 3532 f2fs_info(sbi, "Invalid Fs Meta Ino: node(%u) meta(%u) root(%u)", 3533 le32_to_cpu(raw_super->node_ino), 3534 le32_to_cpu(raw_super->meta_ino), 3535 le32_to_cpu(raw_super->root_ino)); 3536 return -EFSCORRUPTED; 3537 } 3538 3539 /* check CP/SIT/NAT/SSA/MAIN_AREA area boundary */ 3540 if (sanity_check_area_boundary(sbi, bh)) 3541 return -EFSCORRUPTED; 3542 3543 return 0; 3544 } 3545 3546 int f2fs_sanity_check_ckpt(struct f2fs_sb_info *sbi) 3547 { 3548 unsigned int total, fsmeta; 3549 struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi); 3550 struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi); 3551 unsigned int ovp_segments, reserved_segments; 3552 unsigned int main_segs, blocks_per_seg; 3553 unsigned int sit_segs, nat_segs; 3554 unsigned int sit_bitmap_size, nat_bitmap_size; 3555 unsigned int log_blocks_per_seg; 3556 unsigned int segment_count_main; 3557 unsigned int cp_pack_start_sum, cp_payload; 3558 block_t user_block_count, valid_user_blocks; 3559 block_t avail_node_count, valid_node_count; 3560 unsigned int nat_blocks, nat_bits_bytes, nat_bits_blocks; 3561 int i, j; 3562 3563 total = le32_to_cpu(raw_super->segment_count); 3564 fsmeta = le32_to_cpu(raw_super->segment_count_ckpt); 3565 sit_segs = le32_to_cpu(raw_super->segment_count_sit); 3566 fsmeta += sit_segs; 3567 nat_segs = le32_to_cpu(raw_super->segment_count_nat); 3568 fsmeta += nat_segs; 3569 fsmeta += le32_to_cpu(ckpt->rsvd_segment_count); 3570 fsmeta += le32_to_cpu(raw_super->segment_count_ssa); 3571 3572 if (unlikely(fsmeta >= total)) 3573 return 1; 3574 3575 ovp_segments = le32_to_cpu(ckpt->overprov_segment_count); 3576 reserved_segments = le32_to_cpu(ckpt->rsvd_segment_count); 3577 3578 if (!f2fs_sb_has_readonly(sbi) && 3579 unlikely(fsmeta < F2FS_MIN_META_SEGMENTS || 3580 ovp_segments == 0 || reserved_segments == 0)) { 3581 f2fs_err(sbi, "Wrong layout: check mkfs.f2fs version"); 3582 return 1; 3583 } 3584 user_block_count = le64_to_cpu(ckpt->user_block_count); 3585 segment_count_main = le32_to_cpu(raw_super->segment_count_main) + 3586 (f2fs_sb_has_readonly(sbi) ? 1 : 0); 3587 log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg); 3588 if (!user_block_count || user_block_count >= 3589 segment_count_main << log_blocks_per_seg) { 3590 f2fs_err(sbi, "Wrong user_block_count: %u", 3591 user_block_count); 3592 return 1; 3593 } 3594 3595 valid_user_blocks = le64_to_cpu(ckpt->valid_block_count); 3596 if (valid_user_blocks > user_block_count) { 3597 f2fs_err(sbi, "Wrong valid_user_blocks: %u, user_block_count: %u", 3598 valid_user_blocks, user_block_count); 3599 return 1; 3600 } 3601 3602 valid_node_count = le32_to_cpu(ckpt->valid_node_count); 3603 avail_node_count = sbi->total_node_count - F2FS_RESERVED_NODE_NUM; 3604 if (valid_node_count > avail_node_count) { 3605 f2fs_err(sbi, "Wrong valid_node_count: %u, avail_node_count: %u", 3606 valid_node_count, avail_node_count); 3607 return 1; 3608 } 3609 3610 main_segs = le32_to_cpu(raw_super->segment_count_main); 3611 blocks_per_seg = sbi->blocks_per_seg; 3612 3613 for (i = 0; i < NR_CURSEG_NODE_TYPE; i++) { 3614 if (le32_to_cpu(ckpt->cur_node_segno[i]) >= main_segs || 3615 le16_to_cpu(ckpt->cur_node_blkoff[i]) >= blocks_per_seg) 3616 return 1; 3617 3618 if (f2fs_sb_has_readonly(sbi)) 3619 goto check_data; 3620 3621 for (j = i + 1; j < NR_CURSEG_NODE_TYPE; j++) { 3622 if (le32_to_cpu(ckpt->cur_node_segno[i]) == 3623 le32_to_cpu(ckpt->cur_node_segno[j])) { 3624 f2fs_err(sbi, "Node segment (%u, %u) has the same segno: %u", 3625 i, j, 3626 le32_to_cpu(ckpt->cur_node_segno[i])); 3627 return 1; 3628 } 3629 } 3630 } 3631 check_data: 3632 for (i = 0; i < NR_CURSEG_DATA_TYPE; i++) { 3633 if (le32_to_cpu(ckpt->cur_data_segno[i]) >= main_segs || 3634 le16_to_cpu(ckpt->cur_data_blkoff[i]) >= blocks_per_seg) 3635 return 1; 3636 3637 if (f2fs_sb_has_readonly(sbi)) 3638 goto skip_cross; 3639 3640 for (j = i + 1; j < NR_CURSEG_DATA_TYPE; j++) { 3641 if (le32_to_cpu(ckpt->cur_data_segno[i]) == 3642 le32_to_cpu(ckpt->cur_data_segno[j])) { 3643 f2fs_err(sbi, "Data segment (%u, %u) has the same segno: %u", 3644 i, j, 3645 le32_to_cpu(ckpt->cur_data_segno[i])); 3646 return 1; 3647 } 3648 } 3649 } 3650 for (i = 0; i < NR_CURSEG_NODE_TYPE; i++) { 3651 for (j = 0; j < NR_CURSEG_DATA_TYPE; j++) { 3652 if (le32_to_cpu(ckpt->cur_node_segno[i]) == 3653 le32_to_cpu(ckpt->cur_data_segno[j])) { 3654 f2fs_err(sbi, "Node segment (%u) and Data segment (%u) has the same segno: %u", 3655 i, j, 3656 le32_to_cpu(ckpt->cur_node_segno[i])); 3657 return 1; 3658 } 3659 } 3660 } 3661 skip_cross: 3662 sit_bitmap_size = le32_to_cpu(ckpt->sit_ver_bitmap_bytesize); 3663 nat_bitmap_size = le32_to_cpu(ckpt->nat_ver_bitmap_bytesize); 3664 3665 if (sit_bitmap_size != ((sit_segs / 2) << log_blocks_per_seg) / 8 || 3666 nat_bitmap_size != ((nat_segs / 2) << log_blocks_per_seg) / 8) { 3667 f2fs_err(sbi, "Wrong bitmap size: sit: %u, nat:%u", 3668 sit_bitmap_size, nat_bitmap_size); 3669 return 1; 3670 } 3671 3672 cp_pack_start_sum = __start_sum_addr(sbi); 3673 cp_payload = __cp_payload(sbi); 3674 if (cp_pack_start_sum < cp_payload + 1 || 3675 cp_pack_start_sum > blocks_per_seg - 1 - 3676 NR_CURSEG_PERSIST_TYPE) { 3677 f2fs_err(sbi, "Wrong cp_pack_start_sum: %u", 3678 cp_pack_start_sum); 3679 return 1; 3680 } 3681 3682 if (__is_set_ckpt_flags(ckpt, CP_LARGE_NAT_BITMAP_FLAG) && 3683 le32_to_cpu(ckpt->checksum_offset) != CP_MIN_CHKSUM_OFFSET) { 3684 f2fs_warn(sbi, "using deprecated layout of large_nat_bitmap, " 3685 "please run fsck v1.13.0 or higher to repair, chksum_offset: %u, " 3686 "fixed with patch: \"f2fs-tools: relocate chksum_offset for large_nat_bitmap feature\"", 3687 le32_to_cpu(ckpt->checksum_offset)); 3688 return 1; 3689 } 3690 3691 nat_blocks = nat_segs << log_blocks_per_seg; 3692 nat_bits_bytes = nat_blocks / BITS_PER_BYTE; 3693 nat_bits_blocks = F2FS_BLK_ALIGN((nat_bits_bytes << 1) + 8); 3694 if (__is_set_ckpt_flags(ckpt, CP_NAT_BITS_FLAG) && 3695 (cp_payload + F2FS_CP_PACKS + 3696 NR_CURSEG_PERSIST_TYPE + nat_bits_blocks >= blocks_per_seg)) { 3697 f2fs_warn(sbi, "Insane cp_payload: %u, nat_bits_blocks: %u)", 3698 cp_payload, nat_bits_blocks); 3699 return 1; 3700 } 3701 3702 if (unlikely(f2fs_cp_error(sbi))) { 3703 f2fs_err(sbi, "A bug case: need to run fsck"); 3704 return 1; 3705 } 3706 return 0; 3707 } 3708 3709 static void init_sb_info(struct f2fs_sb_info *sbi) 3710 { 3711 struct f2fs_super_block *raw_super = sbi->raw_super; 3712 int i; 3713 3714 sbi->log_sectors_per_block = 3715 le32_to_cpu(raw_super->log_sectors_per_block); 3716 sbi->log_blocksize = le32_to_cpu(raw_super->log_blocksize); 3717 sbi->blocksize = BIT(sbi->log_blocksize); 3718 sbi->log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg); 3719 sbi->blocks_per_seg = BIT(sbi->log_blocks_per_seg); 3720 sbi->segs_per_sec = le32_to_cpu(raw_super->segs_per_sec); 3721 sbi->secs_per_zone = le32_to_cpu(raw_super->secs_per_zone); 3722 sbi->total_sections = le32_to_cpu(raw_super->section_count); 3723 sbi->total_node_count = 3724 (le32_to_cpu(raw_super->segment_count_nat) / 2) 3725 * sbi->blocks_per_seg * NAT_ENTRY_PER_BLOCK; 3726 F2FS_ROOT_INO(sbi) = le32_to_cpu(raw_super->root_ino); 3727 F2FS_NODE_INO(sbi) = le32_to_cpu(raw_super->node_ino); 3728 F2FS_META_INO(sbi) = le32_to_cpu(raw_super->meta_ino); 3729 sbi->cur_victim_sec = NULL_SECNO; 3730 sbi->gc_mode = GC_NORMAL; 3731 sbi->next_victim_seg[BG_GC] = NULL_SEGNO; 3732 sbi->next_victim_seg[FG_GC] = NULL_SEGNO; 3733 sbi->max_victim_search = DEF_MAX_VICTIM_SEARCH; 3734 sbi->migration_granularity = sbi->segs_per_sec; 3735 sbi->seq_file_ra_mul = MIN_RA_MUL; 3736 sbi->max_fragment_chunk = DEF_FRAGMENT_SIZE; 3737 sbi->max_fragment_hole = DEF_FRAGMENT_SIZE; 3738 spin_lock_init(&sbi->gc_remaining_trials_lock); 3739 atomic64_set(&sbi->current_atomic_write, 0); 3740 3741 sbi->dir_level = DEF_DIR_LEVEL; 3742 sbi->interval_time[CP_TIME] = DEF_CP_INTERVAL; 3743 sbi->interval_time[REQ_TIME] = DEF_IDLE_INTERVAL; 3744 sbi->interval_time[DISCARD_TIME] = DEF_IDLE_INTERVAL; 3745 sbi->interval_time[GC_TIME] = DEF_IDLE_INTERVAL; 3746 sbi->interval_time[DISABLE_TIME] = DEF_DISABLE_INTERVAL; 3747 sbi->interval_time[UMOUNT_DISCARD_TIMEOUT] = 3748 DEF_UMOUNT_DISCARD_TIMEOUT; 3749 clear_sbi_flag(sbi, SBI_NEED_FSCK); 3750 3751 for (i = 0; i < NR_COUNT_TYPE; i++) 3752 atomic_set(&sbi->nr_pages[i], 0); 3753 3754 for (i = 0; i < META; i++) 3755 atomic_set(&sbi->wb_sync_req[i], 0); 3756 3757 INIT_LIST_HEAD(&sbi->s_list); 3758 mutex_init(&sbi->umount_mutex); 3759 init_f2fs_rwsem(&sbi->io_order_lock); 3760 spin_lock_init(&sbi->cp_lock); 3761 3762 sbi->dirty_device = 0; 3763 spin_lock_init(&sbi->dev_lock); 3764 3765 init_f2fs_rwsem(&sbi->sb_lock); 3766 init_f2fs_rwsem(&sbi->pin_sem); 3767 } 3768 3769 static int init_percpu_info(struct f2fs_sb_info *sbi) 3770 { 3771 int err; 3772 3773 err = percpu_counter_init(&sbi->alloc_valid_block_count, 0, GFP_KERNEL); 3774 if (err) 3775 return err; 3776 3777 err = percpu_counter_init(&sbi->rf_node_block_count, 0, GFP_KERNEL); 3778 if (err) 3779 goto err_valid_block; 3780 3781 err = percpu_counter_init(&sbi->total_valid_inode_count, 0, 3782 GFP_KERNEL); 3783 if (err) 3784 goto err_node_block; 3785 return 0; 3786 3787 err_node_block: 3788 percpu_counter_destroy(&sbi->rf_node_block_count); 3789 err_valid_block: 3790 percpu_counter_destroy(&sbi->alloc_valid_block_count); 3791 return err; 3792 } 3793 3794 #ifdef CONFIG_BLK_DEV_ZONED 3795 3796 struct f2fs_report_zones_args { 3797 struct f2fs_sb_info *sbi; 3798 struct f2fs_dev_info *dev; 3799 }; 3800 3801 static int f2fs_report_zone_cb(struct blk_zone *zone, unsigned int idx, 3802 void *data) 3803 { 3804 struct f2fs_report_zones_args *rz_args = data; 3805 block_t unusable_blocks = (zone->len - zone->capacity) >> 3806 F2FS_LOG_SECTORS_PER_BLOCK; 3807 3808 if (zone->type == BLK_ZONE_TYPE_CONVENTIONAL) 3809 return 0; 3810 3811 set_bit(idx, rz_args->dev->blkz_seq); 3812 if (!rz_args->sbi->unusable_blocks_per_sec) { 3813 rz_args->sbi->unusable_blocks_per_sec = unusable_blocks; 3814 return 0; 3815 } 3816 if (rz_args->sbi->unusable_blocks_per_sec != unusable_blocks) { 3817 f2fs_err(rz_args->sbi, "F2FS supports single zone capacity\n"); 3818 return -EINVAL; 3819 } 3820 return 0; 3821 } 3822 3823 static int init_blkz_info(struct f2fs_sb_info *sbi, int devi) 3824 { 3825 struct block_device *bdev = FDEV(devi).bdev; 3826 sector_t nr_sectors = bdev_nr_sectors(bdev); 3827 struct f2fs_report_zones_args rep_zone_arg; 3828 u64 zone_sectors; 3829 int ret; 3830 3831 if (!f2fs_sb_has_blkzoned(sbi)) 3832 return 0; 3833 3834 zone_sectors = bdev_zone_sectors(bdev); 3835 if (!is_power_of_2(zone_sectors)) { 3836 f2fs_err(sbi, "F2FS does not support non power of 2 zone sizes\n"); 3837 return -EINVAL; 3838 } 3839 3840 if (sbi->blocks_per_blkz && sbi->blocks_per_blkz != 3841 SECTOR_TO_BLOCK(zone_sectors)) 3842 return -EINVAL; 3843 sbi->blocks_per_blkz = SECTOR_TO_BLOCK(zone_sectors); 3844 FDEV(devi).nr_blkz = div_u64(SECTOR_TO_BLOCK(nr_sectors), 3845 sbi->blocks_per_blkz); 3846 if (nr_sectors & (zone_sectors - 1)) 3847 FDEV(devi).nr_blkz++; 3848 3849 FDEV(devi).blkz_seq = f2fs_kvzalloc(sbi, 3850 BITS_TO_LONGS(FDEV(devi).nr_blkz) 3851 * sizeof(unsigned long), 3852 GFP_KERNEL); 3853 if (!FDEV(devi).blkz_seq) 3854 return -ENOMEM; 3855 3856 rep_zone_arg.sbi = sbi; 3857 rep_zone_arg.dev = &FDEV(devi); 3858 3859 ret = blkdev_report_zones(bdev, 0, BLK_ALL_ZONES, f2fs_report_zone_cb, 3860 &rep_zone_arg); 3861 if (ret < 0) 3862 return ret; 3863 return 0; 3864 } 3865 #endif 3866 3867 /* 3868 * Read f2fs raw super block. 3869 * Because we have two copies of super block, so read both of them 3870 * to get the first valid one. If any one of them is broken, we pass 3871 * them recovery flag back to the caller. 3872 */ 3873 static int read_raw_super_block(struct f2fs_sb_info *sbi, 3874 struct f2fs_super_block **raw_super, 3875 int *valid_super_block, int *recovery) 3876 { 3877 struct super_block *sb = sbi->sb; 3878 int block; 3879 struct buffer_head *bh; 3880 struct f2fs_super_block *super; 3881 int err = 0; 3882 3883 super = kzalloc(sizeof(struct f2fs_super_block), GFP_KERNEL); 3884 if (!super) 3885 return -ENOMEM; 3886 3887 for (block = 0; block < 2; block++) { 3888 bh = sb_bread(sb, block); 3889 if (!bh) { 3890 f2fs_err(sbi, "Unable to read %dth superblock", 3891 block + 1); 3892 err = -EIO; 3893 *recovery = 1; 3894 continue; 3895 } 3896 3897 /* sanity checking of raw super */ 3898 err = sanity_check_raw_super(sbi, bh); 3899 if (err) { 3900 f2fs_err(sbi, "Can't find valid F2FS filesystem in %dth superblock", 3901 block + 1); 3902 brelse(bh); 3903 *recovery = 1; 3904 continue; 3905 } 3906 3907 if (!*raw_super) { 3908 memcpy(super, bh->b_data + F2FS_SUPER_OFFSET, 3909 sizeof(*super)); 3910 *valid_super_block = block; 3911 *raw_super = super; 3912 } 3913 brelse(bh); 3914 } 3915 3916 /* No valid superblock */ 3917 if (!*raw_super) 3918 kfree(super); 3919 else 3920 err = 0; 3921 3922 return err; 3923 } 3924 3925 int f2fs_commit_super(struct f2fs_sb_info *sbi, bool recover) 3926 { 3927 struct buffer_head *bh; 3928 __u32 crc = 0; 3929 int err; 3930 3931 if ((recover && f2fs_readonly(sbi->sb)) || 3932 f2fs_hw_is_readonly(sbi)) { 3933 set_sbi_flag(sbi, SBI_NEED_SB_WRITE); 3934 return -EROFS; 3935 } 3936 3937 /* we should update superblock crc here */ 3938 if (!recover && f2fs_sb_has_sb_chksum(sbi)) { 3939 crc = f2fs_crc32(sbi, F2FS_RAW_SUPER(sbi), 3940 offsetof(struct f2fs_super_block, crc)); 3941 F2FS_RAW_SUPER(sbi)->crc = cpu_to_le32(crc); 3942 } 3943 3944 /* write back-up superblock first */ 3945 bh = sb_bread(sbi->sb, sbi->valid_super_block ? 0 : 1); 3946 if (!bh) 3947 return -EIO; 3948 err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi)); 3949 brelse(bh); 3950 3951 /* if we are in recovery path, skip writing valid superblock */ 3952 if (recover || err) 3953 return err; 3954 3955 /* write current valid superblock */ 3956 bh = sb_bread(sbi->sb, sbi->valid_super_block); 3957 if (!bh) 3958 return -EIO; 3959 err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi)); 3960 brelse(bh); 3961 return err; 3962 } 3963 3964 static void save_stop_reason(struct f2fs_sb_info *sbi, unsigned char reason) 3965 { 3966 unsigned long flags; 3967 3968 spin_lock_irqsave(&sbi->error_lock, flags); 3969 if (sbi->stop_reason[reason] < GENMASK(BITS_PER_BYTE - 1, 0)) 3970 sbi->stop_reason[reason]++; 3971 spin_unlock_irqrestore(&sbi->error_lock, flags); 3972 } 3973 3974 static void f2fs_record_stop_reason(struct f2fs_sb_info *sbi) 3975 { 3976 struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi); 3977 unsigned long flags; 3978 int err; 3979 3980 f2fs_down_write(&sbi->sb_lock); 3981 3982 spin_lock_irqsave(&sbi->error_lock, flags); 3983 memcpy(raw_super->s_stop_reason, sbi->stop_reason, MAX_STOP_REASON); 3984 spin_unlock_irqrestore(&sbi->error_lock, flags); 3985 3986 err = f2fs_commit_super(sbi, false); 3987 3988 f2fs_up_write(&sbi->sb_lock); 3989 if (err) 3990 f2fs_err(sbi, "f2fs_commit_super fails to record err:%d", err); 3991 } 3992 3993 void f2fs_save_errors(struct f2fs_sb_info *sbi, unsigned char flag) 3994 { 3995 unsigned long flags; 3996 3997 spin_lock_irqsave(&sbi->error_lock, flags); 3998 if (!test_bit(flag, (unsigned long *)sbi->errors)) { 3999 set_bit(flag, (unsigned long *)sbi->errors); 4000 sbi->error_dirty = true; 4001 } 4002 spin_unlock_irqrestore(&sbi->error_lock, flags); 4003 } 4004 4005 static bool f2fs_update_errors(struct f2fs_sb_info *sbi) 4006 { 4007 unsigned long flags; 4008 bool need_update = false; 4009 4010 spin_lock_irqsave(&sbi->error_lock, flags); 4011 if (sbi->error_dirty) { 4012 memcpy(F2FS_RAW_SUPER(sbi)->s_errors, sbi->errors, 4013 MAX_F2FS_ERRORS); 4014 sbi->error_dirty = false; 4015 need_update = true; 4016 } 4017 spin_unlock_irqrestore(&sbi->error_lock, flags); 4018 4019 return need_update; 4020 } 4021 4022 void f2fs_handle_error(struct f2fs_sb_info *sbi, unsigned char error) 4023 { 4024 int err; 4025 4026 f2fs_save_errors(sbi, error); 4027 4028 f2fs_down_write(&sbi->sb_lock); 4029 4030 if (!f2fs_update_errors(sbi)) 4031 goto out_unlock; 4032 4033 err = f2fs_commit_super(sbi, false); 4034 if (err) 4035 f2fs_err(sbi, "f2fs_commit_super fails to record errors:%u, err:%d", 4036 error, err); 4037 out_unlock: 4038 f2fs_up_write(&sbi->sb_lock); 4039 } 4040 4041 static bool system_going_down(void) 4042 { 4043 return system_state == SYSTEM_HALT || system_state == SYSTEM_POWER_OFF 4044 || system_state == SYSTEM_RESTART; 4045 } 4046 4047 void f2fs_handle_critical_error(struct f2fs_sb_info *sbi, unsigned char reason, 4048 bool irq_context) 4049 { 4050 struct super_block *sb = sbi->sb; 4051 bool shutdown = reason == STOP_CP_REASON_SHUTDOWN; 4052 bool continue_fs = !shutdown && 4053 F2FS_OPTION(sbi).errors == MOUNT_ERRORS_CONTINUE; 4054 4055 set_ckpt_flags(sbi, CP_ERROR_FLAG); 4056 4057 if (!f2fs_hw_is_readonly(sbi)) { 4058 save_stop_reason(sbi, reason); 4059 4060 if (irq_context && !shutdown) 4061 schedule_work(&sbi->s_error_work); 4062 else 4063 f2fs_record_stop_reason(sbi); 4064 } 4065 4066 /* 4067 * We force ERRORS_RO behavior when system is rebooting. Otherwise we 4068 * could panic during 'reboot -f' as the underlying device got already 4069 * disabled. 4070 */ 4071 if (F2FS_OPTION(sbi).errors == MOUNT_ERRORS_PANIC && 4072 !shutdown && !system_going_down() && 4073 !is_sbi_flag_set(sbi, SBI_IS_SHUTDOWN)) 4074 panic("F2FS-fs (device %s): panic forced after error\n", 4075 sb->s_id); 4076 4077 if (shutdown) 4078 set_sbi_flag(sbi, SBI_IS_SHUTDOWN); 4079 4080 /* continue filesystem operators if errors=continue */ 4081 if (continue_fs || f2fs_readonly(sb)) 4082 return; 4083 4084 f2fs_warn(sbi, "Remounting filesystem read-only"); 4085 /* 4086 * Make sure updated value of ->s_mount_flags will be visible before 4087 * ->s_flags update 4088 */ 4089 smp_wmb(); 4090 sb->s_flags |= SB_RDONLY; 4091 } 4092 4093 static void f2fs_record_error_work(struct work_struct *work) 4094 { 4095 struct f2fs_sb_info *sbi = container_of(work, 4096 struct f2fs_sb_info, s_error_work); 4097 4098 f2fs_record_stop_reason(sbi); 4099 } 4100 4101 static int f2fs_scan_devices(struct f2fs_sb_info *sbi) 4102 { 4103 struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi); 4104 unsigned int max_devices = MAX_DEVICES; 4105 unsigned int logical_blksize; 4106 int i; 4107 4108 /* Initialize single device information */ 4109 if (!RDEV(0).path[0]) { 4110 if (!bdev_is_zoned(sbi->sb->s_bdev)) 4111 return 0; 4112 max_devices = 1; 4113 } 4114 4115 /* 4116 * Initialize multiple devices information, or single 4117 * zoned block device information. 4118 */ 4119 sbi->devs = f2fs_kzalloc(sbi, 4120 array_size(max_devices, 4121 sizeof(struct f2fs_dev_info)), 4122 GFP_KERNEL); 4123 if (!sbi->devs) 4124 return -ENOMEM; 4125 4126 logical_blksize = bdev_logical_block_size(sbi->sb->s_bdev); 4127 sbi->aligned_blksize = true; 4128 4129 for (i = 0; i < max_devices; i++) { 4130 4131 if (i > 0 && !RDEV(i).path[0]) 4132 break; 4133 4134 if (max_devices == 1) { 4135 /* Single zoned block device mount */ 4136 FDEV(0).bdev = 4137 blkdev_get_by_dev(sbi->sb->s_bdev->bd_dev, 4138 sbi->sb->s_mode, sbi->sb->s_type); 4139 } else { 4140 /* Multi-device mount */ 4141 memcpy(FDEV(i).path, RDEV(i).path, MAX_PATH_LEN); 4142 FDEV(i).total_segments = 4143 le32_to_cpu(RDEV(i).total_segments); 4144 if (i == 0) { 4145 FDEV(i).start_blk = 0; 4146 FDEV(i).end_blk = FDEV(i).start_blk + 4147 (FDEV(i).total_segments << 4148 sbi->log_blocks_per_seg) - 1 + 4149 le32_to_cpu(raw_super->segment0_blkaddr); 4150 } else { 4151 FDEV(i).start_blk = FDEV(i - 1).end_blk + 1; 4152 FDEV(i).end_blk = FDEV(i).start_blk + 4153 (FDEV(i).total_segments << 4154 sbi->log_blocks_per_seg) - 1; 4155 } 4156 FDEV(i).bdev = blkdev_get_by_path(FDEV(i).path, 4157 sbi->sb->s_mode, sbi->sb->s_type); 4158 } 4159 if (IS_ERR(FDEV(i).bdev)) 4160 return PTR_ERR(FDEV(i).bdev); 4161 4162 /* to release errored devices */ 4163 sbi->s_ndevs = i + 1; 4164 4165 if (logical_blksize != bdev_logical_block_size(FDEV(i).bdev)) 4166 sbi->aligned_blksize = false; 4167 4168 #ifdef CONFIG_BLK_DEV_ZONED 4169 if (bdev_zoned_model(FDEV(i).bdev) == BLK_ZONED_HM && 4170 !f2fs_sb_has_blkzoned(sbi)) { 4171 f2fs_err(sbi, "Zoned block device feature not enabled"); 4172 return -EINVAL; 4173 } 4174 if (bdev_zoned_model(FDEV(i).bdev) != BLK_ZONED_NONE) { 4175 if (init_blkz_info(sbi, i)) { 4176 f2fs_err(sbi, "Failed to initialize F2FS blkzone information"); 4177 return -EINVAL; 4178 } 4179 if (max_devices == 1) 4180 break; 4181 f2fs_info(sbi, "Mount Device [%2d]: %20s, %8u, %8x - %8x (zone: %s)", 4182 i, FDEV(i).path, 4183 FDEV(i).total_segments, 4184 FDEV(i).start_blk, FDEV(i).end_blk, 4185 bdev_zoned_model(FDEV(i).bdev) == BLK_ZONED_HA ? 4186 "Host-aware" : "Host-managed"); 4187 continue; 4188 } 4189 #endif 4190 f2fs_info(sbi, "Mount Device [%2d]: %20s, %8u, %8x - %8x", 4191 i, FDEV(i).path, 4192 FDEV(i).total_segments, 4193 FDEV(i).start_blk, FDEV(i).end_blk); 4194 } 4195 f2fs_info(sbi, 4196 "IO Block Size: %8ld KB", F2FS_IO_SIZE_KB(sbi)); 4197 return 0; 4198 } 4199 4200 static int f2fs_setup_casefold(struct f2fs_sb_info *sbi) 4201 { 4202 #if IS_ENABLED(CONFIG_UNICODE) 4203 if (f2fs_sb_has_casefold(sbi) && !sbi->sb->s_encoding) { 4204 const struct f2fs_sb_encodings *encoding_info; 4205 struct unicode_map *encoding; 4206 __u16 encoding_flags; 4207 4208 encoding_info = f2fs_sb_read_encoding(sbi->raw_super); 4209 if (!encoding_info) { 4210 f2fs_err(sbi, 4211 "Encoding requested by superblock is unknown"); 4212 return -EINVAL; 4213 } 4214 4215 encoding_flags = le16_to_cpu(sbi->raw_super->s_encoding_flags); 4216 encoding = utf8_load(encoding_info->version); 4217 if (IS_ERR(encoding)) { 4218 f2fs_err(sbi, 4219 "can't mount with superblock charset: %s-%u.%u.%u " 4220 "not supported by the kernel. flags: 0x%x.", 4221 encoding_info->name, 4222 unicode_major(encoding_info->version), 4223 unicode_minor(encoding_info->version), 4224 unicode_rev(encoding_info->version), 4225 encoding_flags); 4226 return PTR_ERR(encoding); 4227 } 4228 f2fs_info(sbi, "Using encoding defined by superblock: " 4229 "%s-%u.%u.%u with flags 0x%hx", encoding_info->name, 4230 unicode_major(encoding_info->version), 4231 unicode_minor(encoding_info->version), 4232 unicode_rev(encoding_info->version), 4233 encoding_flags); 4234 4235 sbi->sb->s_encoding = encoding; 4236 sbi->sb->s_encoding_flags = encoding_flags; 4237 } 4238 #else 4239 if (f2fs_sb_has_casefold(sbi)) { 4240 f2fs_err(sbi, "Filesystem with casefold feature cannot be mounted without CONFIG_UNICODE"); 4241 return -EINVAL; 4242 } 4243 #endif 4244 return 0; 4245 } 4246 4247 static void f2fs_tuning_parameters(struct f2fs_sb_info *sbi) 4248 { 4249 /* adjust parameters according to the volume size */ 4250 if (MAIN_SEGS(sbi) <= SMALL_VOLUME_SEGMENTS) { 4251 if (f2fs_block_unit_discard(sbi)) 4252 SM_I(sbi)->dcc_info->discard_granularity = 4253 MIN_DISCARD_GRANULARITY; 4254 if (!f2fs_lfs_mode(sbi)) 4255 SM_I(sbi)->ipu_policy = BIT(F2FS_IPU_FORCE) | 4256 BIT(F2FS_IPU_HONOR_OPU_WRITE); 4257 } 4258 4259 sbi->readdir_ra = true; 4260 } 4261 4262 static int f2fs_fill_super(struct super_block *sb, void *data, int silent) 4263 { 4264 struct f2fs_sb_info *sbi; 4265 struct f2fs_super_block *raw_super; 4266 struct inode *root; 4267 int err; 4268 bool skip_recovery = false, need_fsck = false; 4269 char *options = NULL; 4270 int recovery, i, valid_super_block; 4271 struct curseg_info *seg_i; 4272 int retry_cnt = 1; 4273 #ifdef CONFIG_QUOTA 4274 bool quota_enabled = false; 4275 #endif 4276 4277 try_onemore: 4278 err = -EINVAL; 4279 raw_super = NULL; 4280 valid_super_block = -1; 4281 recovery = 0; 4282 4283 /* allocate memory for f2fs-specific super block info */ 4284 sbi = kzalloc(sizeof(struct f2fs_sb_info), GFP_KERNEL); 4285 if (!sbi) 4286 return -ENOMEM; 4287 4288 sbi->sb = sb; 4289 4290 /* initialize locks within allocated memory */ 4291 init_f2fs_rwsem(&sbi->gc_lock); 4292 mutex_init(&sbi->writepages); 4293 init_f2fs_rwsem(&sbi->cp_global_sem); 4294 init_f2fs_rwsem(&sbi->node_write); 4295 init_f2fs_rwsem(&sbi->node_change); 4296 spin_lock_init(&sbi->stat_lock); 4297 init_f2fs_rwsem(&sbi->cp_rwsem); 4298 init_f2fs_rwsem(&sbi->quota_sem); 4299 init_waitqueue_head(&sbi->cp_wait); 4300 spin_lock_init(&sbi->error_lock); 4301 4302 for (i = 0; i < NR_INODE_TYPE; i++) { 4303 INIT_LIST_HEAD(&sbi->inode_list[i]); 4304 spin_lock_init(&sbi->inode_lock[i]); 4305 } 4306 mutex_init(&sbi->flush_lock); 4307 4308 /* Load the checksum driver */ 4309 sbi->s_chksum_driver = crypto_alloc_shash("crc32", 0, 0); 4310 if (IS_ERR(sbi->s_chksum_driver)) { 4311 f2fs_err(sbi, "Cannot load crc32 driver."); 4312 err = PTR_ERR(sbi->s_chksum_driver); 4313 sbi->s_chksum_driver = NULL; 4314 goto free_sbi; 4315 } 4316 4317 /* set a block size */ 4318 if (unlikely(!sb_set_blocksize(sb, F2FS_BLKSIZE))) { 4319 f2fs_err(sbi, "unable to set blocksize"); 4320 goto free_sbi; 4321 } 4322 4323 err = read_raw_super_block(sbi, &raw_super, &valid_super_block, 4324 &recovery); 4325 if (err) 4326 goto free_sbi; 4327 4328 sb->s_fs_info = sbi; 4329 sbi->raw_super = raw_super; 4330 4331 INIT_WORK(&sbi->s_error_work, f2fs_record_error_work); 4332 memcpy(sbi->errors, raw_super->s_errors, MAX_F2FS_ERRORS); 4333 memcpy(sbi->stop_reason, raw_super->s_stop_reason, MAX_STOP_REASON); 4334 4335 /* precompute checksum seed for metadata */ 4336 if (f2fs_sb_has_inode_chksum(sbi)) 4337 sbi->s_chksum_seed = f2fs_chksum(sbi, ~0, raw_super->uuid, 4338 sizeof(raw_super->uuid)); 4339 4340 default_options(sbi); 4341 /* parse mount options */ 4342 options = kstrdup((const char *)data, GFP_KERNEL); 4343 if (data && !options) { 4344 err = -ENOMEM; 4345 goto free_sb_buf; 4346 } 4347 4348 err = parse_options(sb, options, false); 4349 if (err) 4350 goto free_options; 4351 4352 sb->s_maxbytes = max_file_blocks(NULL) << 4353 le32_to_cpu(raw_super->log_blocksize); 4354 sb->s_max_links = F2FS_LINK_MAX; 4355 4356 err = f2fs_setup_casefold(sbi); 4357 if (err) 4358 goto free_options; 4359 4360 #ifdef CONFIG_QUOTA 4361 sb->dq_op = &f2fs_quota_operations; 4362 sb->s_qcop = &f2fs_quotactl_ops; 4363 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ; 4364 4365 if (f2fs_sb_has_quota_ino(sbi)) { 4366 for (i = 0; i < MAXQUOTAS; i++) { 4367 if (f2fs_qf_ino(sbi->sb, i)) 4368 sbi->nquota_files++; 4369 } 4370 } 4371 #endif 4372 4373 sb->s_op = &f2fs_sops; 4374 #ifdef CONFIG_FS_ENCRYPTION 4375 sb->s_cop = &f2fs_cryptops; 4376 #endif 4377 #ifdef CONFIG_FS_VERITY 4378 sb->s_vop = &f2fs_verityops; 4379 #endif 4380 sb->s_xattr = f2fs_xattr_handlers; 4381 sb->s_export_op = &f2fs_export_ops; 4382 sb->s_magic = F2FS_SUPER_MAGIC; 4383 sb->s_time_gran = 1; 4384 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) | 4385 (test_opt(sbi, POSIX_ACL) ? SB_POSIXACL : 0); 4386 memcpy(&sb->s_uuid, raw_super->uuid, sizeof(raw_super->uuid)); 4387 sb->s_iflags |= SB_I_CGROUPWB; 4388 4389 /* init f2fs-specific super block info */ 4390 sbi->valid_super_block = valid_super_block; 4391 4392 /* disallow all the data/node/meta page writes */ 4393 set_sbi_flag(sbi, SBI_POR_DOING); 4394 4395 err = f2fs_init_write_merge_io(sbi); 4396 if (err) 4397 goto free_bio_info; 4398 4399 init_sb_info(sbi); 4400 4401 err = f2fs_init_iostat(sbi); 4402 if (err) 4403 goto free_bio_info; 4404 4405 err = init_percpu_info(sbi); 4406 if (err) 4407 goto free_iostat; 4408 4409 if (F2FS_IO_ALIGNED(sbi)) { 4410 sbi->write_io_dummy = 4411 mempool_create_page_pool(2 * (F2FS_IO_SIZE(sbi) - 1), 0); 4412 if (!sbi->write_io_dummy) { 4413 err = -ENOMEM; 4414 goto free_percpu; 4415 } 4416 } 4417 4418 /* init per sbi slab cache */ 4419 err = f2fs_init_xattr_caches(sbi); 4420 if (err) 4421 goto free_io_dummy; 4422 err = f2fs_init_page_array_cache(sbi); 4423 if (err) 4424 goto free_xattr_cache; 4425 4426 /* get an inode for meta space */ 4427 sbi->meta_inode = f2fs_iget(sb, F2FS_META_INO(sbi)); 4428 if (IS_ERR(sbi->meta_inode)) { 4429 f2fs_err(sbi, "Failed to read F2FS meta data inode"); 4430 err = PTR_ERR(sbi->meta_inode); 4431 goto free_page_array_cache; 4432 } 4433 4434 err = f2fs_get_valid_checkpoint(sbi); 4435 if (err) { 4436 f2fs_err(sbi, "Failed to get valid F2FS checkpoint"); 4437 goto free_meta_inode; 4438 } 4439 4440 if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_QUOTA_NEED_FSCK_FLAG)) 4441 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR); 4442 if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_DISABLED_QUICK_FLAG)) { 4443 set_sbi_flag(sbi, SBI_CP_DISABLED_QUICK); 4444 sbi->interval_time[DISABLE_TIME] = DEF_DISABLE_QUICK_INTERVAL; 4445 } 4446 4447 if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_FSCK_FLAG)) 4448 set_sbi_flag(sbi, SBI_NEED_FSCK); 4449 4450 /* Initialize device list */ 4451 err = f2fs_scan_devices(sbi); 4452 if (err) { 4453 f2fs_err(sbi, "Failed to find devices"); 4454 goto free_devices; 4455 } 4456 4457 err = f2fs_init_post_read_wq(sbi); 4458 if (err) { 4459 f2fs_err(sbi, "Failed to initialize post read workqueue"); 4460 goto free_devices; 4461 } 4462 4463 sbi->total_valid_node_count = 4464 le32_to_cpu(sbi->ckpt->valid_node_count); 4465 percpu_counter_set(&sbi->total_valid_inode_count, 4466 le32_to_cpu(sbi->ckpt->valid_inode_count)); 4467 sbi->user_block_count = le64_to_cpu(sbi->ckpt->user_block_count); 4468 sbi->total_valid_block_count = 4469 le64_to_cpu(sbi->ckpt->valid_block_count); 4470 sbi->last_valid_block_count = sbi->total_valid_block_count; 4471 sbi->reserved_blocks = 0; 4472 sbi->current_reserved_blocks = 0; 4473 limit_reserve_root(sbi); 4474 adjust_unusable_cap_perc(sbi); 4475 4476 f2fs_init_extent_cache_info(sbi); 4477 4478 f2fs_init_ino_entry_info(sbi); 4479 4480 f2fs_init_fsync_node_info(sbi); 4481 4482 /* setup checkpoint request control and start checkpoint issue thread */ 4483 f2fs_init_ckpt_req_control(sbi); 4484 if (!f2fs_readonly(sb) && !test_opt(sbi, DISABLE_CHECKPOINT) && 4485 test_opt(sbi, MERGE_CHECKPOINT)) { 4486 err = f2fs_start_ckpt_thread(sbi); 4487 if (err) { 4488 f2fs_err(sbi, 4489 "Failed to start F2FS issue_checkpoint_thread (%d)", 4490 err); 4491 goto stop_ckpt_thread; 4492 } 4493 } 4494 4495 /* setup f2fs internal modules */ 4496 err = f2fs_build_segment_manager(sbi); 4497 if (err) { 4498 f2fs_err(sbi, "Failed to initialize F2FS segment manager (%d)", 4499 err); 4500 goto free_sm; 4501 } 4502 err = f2fs_build_node_manager(sbi); 4503 if (err) { 4504 f2fs_err(sbi, "Failed to initialize F2FS node manager (%d)", 4505 err); 4506 goto free_nm; 4507 } 4508 4509 err = adjust_reserved_segment(sbi); 4510 if (err) 4511 goto free_nm; 4512 4513 /* For write statistics */ 4514 sbi->sectors_written_start = f2fs_get_sectors_written(sbi); 4515 4516 /* Read accumulated write IO statistics if exists */ 4517 seg_i = CURSEG_I(sbi, CURSEG_HOT_NODE); 4518 if (__exist_node_summaries(sbi)) 4519 sbi->kbytes_written = 4520 le64_to_cpu(seg_i->journal->info.kbytes_written); 4521 4522 f2fs_build_gc_manager(sbi); 4523 4524 err = f2fs_build_stats(sbi); 4525 if (err) 4526 goto free_nm; 4527 4528 /* get an inode for node space */ 4529 sbi->node_inode = f2fs_iget(sb, F2FS_NODE_INO(sbi)); 4530 if (IS_ERR(sbi->node_inode)) { 4531 f2fs_err(sbi, "Failed to read node inode"); 4532 err = PTR_ERR(sbi->node_inode); 4533 goto free_stats; 4534 } 4535 4536 /* read root inode and dentry */ 4537 root = f2fs_iget(sb, F2FS_ROOT_INO(sbi)); 4538 if (IS_ERR(root)) { 4539 f2fs_err(sbi, "Failed to read root inode"); 4540 err = PTR_ERR(root); 4541 goto free_node_inode; 4542 } 4543 if (!S_ISDIR(root->i_mode) || !root->i_blocks || 4544 !root->i_size || !root->i_nlink) { 4545 iput(root); 4546 err = -EINVAL; 4547 goto free_node_inode; 4548 } 4549 4550 sb->s_root = d_make_root(root); /* allocate root dentry */ 4551 if (!sb->s_root) { 4552 err = -ENOMEM; 4553 goto free_node_inode; 4554 } 4555 4556 err = f2fs_init_compress_inode(sbi); 4557 if (err) 4558 goto free_root_inode; 4559 4560 err = f2fs_register_sysfs(sbi); 4561 if (err) 4562 goto free_compress_inode; 4563 4564 #ifdef CONFIG_QUOTA 4565 /* Enable quota usage during mount */ 4566 if (f2fs_sb_has_quota_ino(sbi) && !f2fs_readonly(sb)) { 4567 err = f2fs_enable_quotas(sb); 4568 if (err) 4569 f2fs_err(sbi, "Cannot turn on quotas: error %d", err); 4570 } 4571 4572 quota_enabled = f2fs_recover_quota_begin(sbi); 4573 #endif 4574 /* if there are any orphan inodes, free them */ 4575 err = f2fs_recover_orphan_inodes(sbi); 4576 if (err) 4577 goto free_meta; 4578 4579 if (unlikely(is_set_ckpt_flags(sbi, CP_DISABLED_FLAG))) 4580 goto reset_checkpoint; 4581 4582 /* recover fsynced data */ 4583 if (!test_opt(sbi, DISABLE_ROLL_FORWARD) && 4584 !test_opt(sbi, NORECOVERY)) { 4585 /* 4586 * mount should be failed, when device has readonly mode, and 4587 * previous checkpoint was not done by clean system shutdown. 4588 */ 4589 if (f2fs_hw_is_readonly(sbi)) { 4590 if (!is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG)) { 4591 err = f2fs_recover_fsync_data(sbi, true); 4592 if (err > 0) { 4593 err = -EROFS; 4594 f2fs_err(sbi, "Need to recover fsync data, but " 4595 "write access unavailable, please try " 4596 "mount w/ disable_roll_forward or norecovery"); 4597 } 4598 if (err < 0) 4599 goto free_meta; 4600 } 4601 f2fs_info(sbi, "write access unavailable, skipping recovery"); 4602 goto reset_checkpoint; 4603 } 4604 4605 if (need_fsck) 4606 set_sbi_flag(sbi, SBI_NEED_FSCK); 4607 4608 if (skip_recovery) 4609 goto reset_checkpoint; 4610 4611 err = f2fs_recover_fsync_data(sbi, false); 4612 if (err < 0) { 4613 if (err != -ENOMEM) 4614 skip_recovery = true; 4615 need_fsck = true; 4616 f2fs_err(sbi, "Cannot recover all fsync data errno=%d", 4617 err); 4618 goto free_meta; 4619 } 4620 } else { 4621 err = f2fs_recover_fsync_data(sbi, true); 4622 4623 if (!f2fs_readonly(sb) && err > 0) { 4624 err = -EINVAL; 4625 f2fs_err(sbi, "Need to recover fsync data"); 4626 goto free_meta; 4627 } 4628 } 4629 4630 #ifdef CONFIG_QUOTA 4631 f2fs_recover_quota_end(sbi, quota_enabled); 4632 #endif 4633 4634 /* 4635 * If the f2fs is not readonly and fsync data recovery succeeds, 4636 * check zoned block devices' write pointer consistency. 4637 */ 4638 if (!err && !f2fs_readonly(sb) && f2fs_sb_has_blkzoned(sbi)) { 4639 err = f2fs_check_write_pointer(sbi); 4640 if (err) 4641 goto free_meta; 4642 } 4643 4644 reset_checkpoint: 4645 f2fs_init_inmem_curseg(sbi); 4646 4647 /* f2fs_recover_fsync_data() cleared this already */ 4648 clear_sbi_flag(sbi, SBI_POR_DOING); 4649 4650 if (test_opt(sbi, DISABLE_CHECKPOINT)) { 4651 err = f2fs_disable_checkpoint(sbi); 4652 if (err) 4653 goto sync_free_meta; 4654 } else if (is_set_ckpt_flags(sbi, CP_DISABLED_FLAG)) { 4655 f2fs_enable_checkpoint(sbi); 4656 } 4657 4658 /* 4659 * If filesystem is not mounted as read-only then 4660 * do start the gc_thread. 4661 */ 4662 if ((F2FS_OPTION(sbi).bggc_mode != BGGC_MODE_OFF || 4663 test_opt(sbi, GC_MERGE)) && !f2fs_readonly(sb)) { 4664 /* After POR, we can run background GC thread.*/ 4665 err = f2fs_start_gc_thread(sbi); 4666 if (err) 4667 goto sync_free_meta; 4668 } 4669 kvfree(options); 4670 4671 /* recover broken superblock */ 4672 if (recovery) { 4673 err = f2fs_commit_super(sbi, true); 4674 f2fs_info(sbi, "Try to recover %dth superblock, ret: %d", 4675 sbi->valid_super_block ? 1 : 2, err); 4676 } 4677 4678 f2fs_join_shrinker(sbi); 4679 4680 f2fs_tuning_parameters(sbi); 4681 4682 f2fs_notice(sbi, "Mounted with checkpoint version = %llx", 4683 cur_cp_version(F2FS_CKPT(sbi))); 4684 f2fs_update_time(sbi, CP_TIME); 4685 f2fs_update_time(sbi, REQ_TIME); 4686 clear_sbi_flag(sbi, SBI_CP_DISABLED_QUICK); 4687 return 0; 4688 4689 sync_free_meta: 4690 /* safe to flush all the data */ 4691 sync_filesystem(sbi->sb); 4692 retry_cnt = 0; 4693 4694 free_meta: 4695 #ifdef CONFIG_QUOTA 4696 f2fs_truncate_quota_inode_pages(sb); 4697 if (f2fs_sb_has_quota_ino(sbi) && !f2fs_readonly(sb)) 4698 f2fs_quota_off_umount(sbi->sb); 4699 #endif 4700 /* 4701 * Some dirty meta pages can be produced by f2fs_recover_orphan_inodes() 4702 * failed by EIO. Then, iput(node_inode) can trigger balance_fs_bg() 4703 * followed by f2fs_write_checkpoint() through f2fs_write_node_pages(), which 4704 * falls into an infinite loop in f2fs_sync_meta_pages(). 4705 */ 4706 truncate_inode_pages_final(META_MAPPING(sbi)); 4707 /* evict some inodes being cached by GC */ 4708 evict_inodes(sb); 4709 f2fs_unregister_sysfs(sbi); 4710 free_compress_inode: 4711 f2fs_destroy_compress_inode(sbi); 4712 free_root_inode: 4713 dput(sb->s_root); 4714 sb->s_root = NULL; 4715 free_node_inode: 4716 f2fs_release_ino_entry(sbi, true); 4717 truncate_inode_pages_final(NODE_MAPPING(sbi)); 4718 iput(sbi->node_inode); 4719 sbi->node_inode = NULL; 4720 free_stats: 4721 f2fs_destroy_stats(sbi); 4722 free_nm: 4723 /* stop discard thread before destroying node manager */ 4724 f2fs_stop_discard_thread(sbi); 4725 f2fs_destroy_node_manager(sbi); 4726 free_sm: 4727 f2fs_destroy_segment_manager(sbi); 4728 stop_ckpt_thread: 4729 f2fs_stop_ckpt_thread(sbi); 4730 /* flush s_error_work before sbi destroy */ 4731 flush_work(&sbi->s_error_work); 4732 f2fs_destroy_post_read_wq(sbi); 4733 free_devices: 4734 destroy_device_list(sbi); 4735 kvfree(sbi->ckpt); 4736 free_meta_inode: 4737 make_bad_inode(sbi->meta_inode); 4738 iput(sbi->meta_inode); 4739 sbi->meta_inode = NULL; 4740 free_page_array_cache: 4741 f2fs_destroy_page_array_cache(sbi); 4742 free_xattr_cache: 4743 f2fs_destroy_xattr_caches(sbi); 4744 free_io_dummy: 4745 mempool_destroy(sbi->write_io_dummy); 4746 free_percpu: 4747 destroy_percpu_info(sbi); 4748 free_iostat: 4749 f2fs_destroy_iostat(sbi); 4750 free_bio_info: 4751 for (i = 0; i < NR_PAGE_TYPE; i++) 4752 kvfree(sbi->write_io[i]); 4753 4754 #if IS_ENABLED(CONFIG_UNICODE) 4755 utf8_unload(sb->s_encoding); 4756 sb->s_encoding = NULL; 4757 #endif 4758 free_options: 4759 #ifdef CONFIG_QUOTA 4760 for (i = 0; i < MAXQUOTAS; i++) 4761 kfree(F2FS_OPTION(sbi).s_qf_names[i]); 4762 #endif 4763 fscrypt_free_dummy_policy(&F2FS_OPTION(sbi).dummy_enc_policy); 4764 kvfree(options); 4765 free_sb_buf: 4766 kfree(raw_super); 4767 free_sbi: 4768 if (sbi->s_chksum_driver) 4769 crypto_free_shash(sbi->s_chksum_driver); 4770 kfree(sbi); 4771 4772 /* give only one another chance */ 4773 if (retry_cnt > 0 && skip_recovery) { 4774 retry_cnt--; 4775 shrink_dcache_sb(sb); 4776 goto try_onemore; 4777 } 4778 return err; 4779 } 4780 4781 static struct dentry *f2fs_mount(struct file_system_type *fs_type, int flags, 4782 const char *dev_name, void *data) 4783 { 4784 return mount_bdev(fs_type, flags, dev_name, data, f2fs_fill_super); 4785 } 4786 4787 static void kill_f2fs_super(struct super_block *sb) 4788 { 4789 if (sb->s_root) { 4790 struct f2fs_sb_info *sbi = F2FS_SB(sb); 4791 4792 set_sbi_flag(sbi, SBI_IS_CLOSE); 4793 f2fs_stop_gc_thread(sbi); 4794 f2fs_stop_discard_thread(sbi); 4795 4796 #ifdef CONFIG_F2FS_FS_COMPRESSION 4797 /* 4798 * latter evict_inode() can bypass checking and invalidating 4799 * compress inode cache. 4800 */ 4801 if (test_opt(sbi, COMPRESS_CACHE)) 4802 truncate_inode_pages_final(COMPRESS_MAPPING(sbi)); 4803 #endif 4804 4805 if (is_sbi_flag_set(sbi, SBI_IS_DIRTY) || 4806 !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG)) { 4807 struct cp_control cpc = { 4808 .reason = CP_UMOUNT, 4809 }; 4810 f2fs_write_checkpoint(sbi, &cpc); 4811 } 4812 4813 if (is_sbi_flag_set(sbi, SBI_IS_RECOVERED) && f2fs_readonly(sb)) 4814 sb->s_flags &= ~SB_RDONLY; 4815 } 4816 kill_block_super(sb); 4817 } 4818 4819 static struct file_system_type f2fs_fs_type = { 4820 .owner = THIS_MODULE, 4821 .name = "f2fs", 4822 .mount = f2fs_mount, 4823 .kill_sb = kill_f2fs_super, 4824 .fs_flags = FS_REQUIRES_DEV | FS_ALLOW_IDMAP, 4825 }; 4826 MODULE_ALIAS_FS("f2fs"); 4827 4828 static int __init init_inodecache(void) 4829 { 4830 f2fs_inode_cachep = kmem_cache_create("f2fs_inode_cache", 4831 sizeof(struct f2fs_inode_info), 0, 4832 SLAB_RECLAIM_ACCOUNT|SLAB_ACCOUNT, NULL); 4833 return f2fs_inode_cachep ? 0 : -ENOMEM; 4834 } 4835 4836 static void destroy_inodecache(void) 4837 { 4838 /* 4839 * Make sure all delayed rcu free inodes are flushed before we 4840 * destroy cache. 4841 */ 4842 rcu_barrier(); 4843 kmem_cache_destroy(f2fs_inode_cachep); 4844 } 4845 4846 static int __init init_f2fs_fs(void) 4847 { 4848 int err; 4849 4850 if (PAGE_SIZE != F2FS_BLKSIZE) { 4851 printk("F2FS not supported on PAGE_SIZE(%lu) != %d\n", 4852 PAGE_SIZE, F2FS_BLKSIZE); 4853 return -EINVAL; 4854 } 4855 4856 err = init_inodecache(); 4857 if (err) 4858 goto fail; 4859 err = f2fs_create_node_manager_caches(); 4860 if (err) 4861 goto free_inodecache; 4862 err = f2fs_create_segment_manager_caches(); 4863 if (err) 4864 goto free_node_manager_caches; 4865 err = f2fs_create_checkpoint_caches(); 4866 if (err) 4867 goto free_segment_manager_caches; 4868 err = f2fs_create_recovery_cache(); 4869 if (err) 4870 goto free_checkpoint_caches; 4871 err = f2fs_create_extent_cache(); 4872 if (err) 4873 goto free_recovery_cache; 4874 err = f2fs_create_garbage_collection_cache(); 4875 if (err) 4876 goto free_extent_cache; 4877 err = f2fs_init_sysfs(); 4878 if (err) 4879 goto free_garbage_collection_cache; 4880 err = register_shrinker(&f2fs_shrinker_info, "f2fs-shrinker"); 4881 if (err) 4882 goto free_sysfs; 4883 err = register_filesystem(&f2fs_fs_type); 4884 if (err) 4885 goto free_shrinker; 4886 f2fs_create_root_stats(); 4887 err = f2fs_init_post_read_processing(); 4888 if (err) 4889 goto free_root_stats; 4890 err = f2fs_init_iostat_processing(); 4891 if (err) 4892 goto free_post_read; 4893 err = f2fs_init_bio_entry_cache(); 4894 if (err) 4895 goto free_iostat; 4896 err = f2fs_init_bioset(); 4897 if (err) 4898 goto free_bio_entry_cache; 4899 err = f2fs_init_compress_mempool(); 4900 if (err) 4901 goto free_bioset; 4902 err = f2fs_init_compress_cache(); 4903 if (err) 4904 goto free_compress_mempool; 4905 err = f2fs_create_casefold_cache(); 4906 if (err) 4907 goto free_compress_cache; 4908 return 0; 4909 free_compress_cache: 4910 f2fs_destroy_compress_cache(); 4911 free_compress_mempool: 4912 f2fs_destroy_compress_mempool(); 4913 free_bioset: 4914 f2fs_destroy_bioset(); 4915 free_bio_entry_cache: 4916 f2fs_destroy_bio_entry_cache(); 4917 free_iostat: 4918 f2fs_destroy_iostat_processing(); 4919 free_post_read: 4920 f2fs_destroy_post_read_processing(); 4921 free_root_stats: 4922 f2fs_destroy_root_stats(); 4923 unregister_filesystem(&f2fs_fs_type); 4924 free_shrinker: 4925 unregister_shrinker(&f2fs_shrinker_info); 4926 free_sysfs: 4927 f2fs_exit_sysfs(); 4928 free_garbage_collection_cache: 4929 f2fs_destroy_garbage_collection_cache(); 4930 free_extent_cache: 4931 f2fs_destroy_extent_cache(); 4932 free_recovery_cache: 4933 f2fs_destroy_recovery_cache(); 4934 free_checkpoint_caches: 4935 f2fs_destroy_checkpoint_caches(); 4936 free_segment_manager_caches: 4937 f2fs_destroy_segment_manager_caches(); 4938 free_node_manager_caches: 4939 f2fs_destroy_node_manager_caches(); 4940 free_inodecache: 4941 destroy_inodecache(); 4942 fail: 4943 return err; 4944 } 4945 4946 static void __exit exit_f2fs_fs(void) 4947 { 4948 f2fs_destroy_casefold_cache(); 4949 f2fs_destroy_compress_cache(); 4950 f2fs_destroy_compress_mempool(); 4951 f2fs_destroy_bioset(); 4952 f2fs_destroy_bio_entry_cache(); 4953 f2fs_destroy_iostat_processing(); 4954 f2fs_destroy_post_read_processing(); 4955 f2fs_destroy_root_stats(); 4956 unregister_filesystem(&f2fs_fs_type); 4957 unregister_shrinker(&f2fs_shrinker_info); 4958 f2fs_exit_sysfs(); 4959 f2fs_destroy_garbage_collection_cache(); 4960 f2fs_destroy_extent_cache(); 4961 f2fs_destroy_recovery_cache(); 4962 f2fs_destroy_checkpoint_caches(); 4963 f2fs_destroy_segment_manager_caches(); 4964 f2fs_destroy_node_manager_caches(); 4965 destroy_inodecache(); 4966 } 4967 4968 module_init(init_f2fs_fs) 4969 module_exit(exit_f2fs_fs) 4970 4971 MODULE_AUTHOR("Samsung Electronics's Praesto Team"); 4972 MODULE_DESCRIPTION("Flash Friendly File System"); 4973 MODULE_LICENSE("GPL"); 4974 MODULE_SOFTDEP("pre: crc32"); 4975 4976