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 88 static int __init f2fs_init_shrinker(void) 89 { 90 f2fs_shrinker_info = shrinker_alloc(0, "f2fs-shrinker"); 91 if (!f2fs_shrinker_info) 92 return -ENOMEM; 93 94 f2fs_shrinker_info->count_objects = f2fs_shrink_count; 95 f2fs_shrinker_info->scan_objects = f2fs_shrink_scan; 96 97 shrinker_register(f2fs_shrinker_info); 98 99 return 0; 100 } 101 102 static void f2fs_exit_shrinker(void) 103 { 104 shrinker_free(f2fs_shrinker_info); 105 } 106 107 enum { 108 Opt_gc_background, 109 Opt_disable_roll_forward, 110 Opt_norecovery, 111 Opt_discard, 112 Opt_nodiscard, 113 Opt_noheap, 114 Opt_heap, 115 Opt_user_xattr, 116 Opt_nouser_xattr, 117 Opt_acl, 118 Opt_noacl, 119 Opt_active_logs, 120 Opt_disable_ext_identify, 121 Opt_inline_xattr, 122 Opt_noinline_xattr, 123 Opt_inline_xattr_size, 124 Opt_inline_data, 125 Opt_inline_dentry, 126 Opt_noinline_dentry, 127 Opt_flush_merge, 128 Opt_noflush_merge, 129 Opt_barrier, 130 Opt_nobarrier, 131 Opt_fastboot, 132 Opt_extent_cache, 133 Opt_noextent_cache, 134 Opt_noinline_data, 135 Opt_data_flush, 136 Opt_reserve_root, 137 Opt_resgid, 138 Opt_resuid, 139 Opt_mode, 140 Opt_io_size_bits, 141 Opt_fault_injection, 142 Opt_fault_type, 143 Opt_lazytime, 144 Opt_nolazytime, 145 Opt_quota, 146 Opt_noquota, 147 Opt_usrquota, 148 Opt_grpquota, 149 Opt_prjquota, 150 Opt_usrjquota, 151 Opt_grpjquota, 152 Opt_prjjquota, 153 Opt_offusrjquota, 154 Opt_offgrpjquota, 155 Opt_offprjjquota, 156 Opt_jqfmt_vfsold, 157 Opt_jqfmt_vfsv0, 158 Opt_jqfmt_vfsv1, 159 Opt_alloc, 160 Opt_fsync, 161 Opt_test_dummy_encryption, 162 Opt_inlinecrypt, 163 Opt_checkpoint_disable, 164 Opt_checkpoint_disable_cap, 165 Opt_checkpoint_disable_cap_perc, 166 Opt_checkpoint_enable, 167 Opt_checkpoint_merge, 168 Opt_nocheckpoint_merge, 169 Opt_compress_algorithm, 170 Opt_compress_log_size, 171 Opt_compress_extension, 172 Opt_nocompress_extension, 173 Opt_compress_chksum, 174 Opt_compress_mode, 175 Opt_compress_cache, 176 Opt_atgc, 177 Opt_gc_merge, 178 Opt_nogc_merge, 179 Opt_discard_unit, 180 Opt_memory_mode, 181 Opt_age_extent_cache, 182 Opt_errors, 183 Opt_err, 184 }; 185 186 static match_table_t f2fs_tokens = { 187 {Opt_gc_background, "background_gc=%s"}, 188 {Opt_disable_roll_forward, "disable_roll_forward"}, 189 {Opt_norecovery, "norecovery"}, 190 {Opt_discard, "discard"}, 191 {Opt_nodiscard, "nodiscard"}, 192 {Opt_noheap, "no_heap"}, 193 {Opt_heap, "heap"}, 194 {Opt_user_xattr, "user_xattr"}, 195 {Opt_nouser_xattr, "nouser_xattr"}, 196 {Opt_acl, "acl"}, 197 {Opt_noacl, "noacl"}, 198 {Opt_active_logs, "active_logs=%u"}, 199 {Opt_disable_ext_identify, "disable_ext_identify"}, 200 {Opt_inline_xattr, "inline_xattr"}, 201 {Opt_noinline_xattr, "noinline_xattr"}, 202 {Opt_inline_xattr_size, "inline_xattr_size=%u"}, 203 {Opt_inline_data, "inline_data"}, 204 {Opt_inline_dentry, "inline_dentry"}, 205 {Opt_noinline_dentry, "noinline_dentry"}, 206 {Opt_flush_merge, "flush_merge"}, 207 {Opt_noflush_merge, "noflush_merge"}, 208 {Opt_barrier, "barrier"}, 209 {Opt_nobarrier, "nobarrier"}, 210 {Opt_fastboot, "fastboot"}, 211 {Opt_extent_cache, "extent_cache"}, 212 {Opt_noextent_cache, "noextent_cache"}, 213 {Opt_noinline_data, "noinline_data"}, 214 {Opt_data_flush, "data_flush"}, 215 {Opt_reserve_root, "reserve_root=%u"}, 216 {Opt_resgid, "resgid=%u"}, 217 {Opt_resuid, "resuid=%u"}, 218 {Opt_mode, "mode=%s"}, 219 {Opt_io_size_bits, "io_bits=%u"}, 220 {Opt_fault_injection, "fault_injection=%u"}, 221 {Opt_fault_type, "fault_type=%u"}, 222 {Opt_lazytime, "lazytime"}, 223 {Opt_nolazytime, "nolazytime"}, 224 {Opt_quota, "quota"}, 225 {Opt_noquota, "noquota"}, 226 {Opt_usrquota, "usrquota"}, 227 {Opt_grpquota, "grpquota"}, 228 {Opt_prjquota, "prjquota"}, 229 {Opt_usrjquota, "usrjquota=%s"}, 230 {Opt_grpjquota, "grpjquota=%s"}, 231 {Opt_prjjquota, "prjjquota=%s"}, 232 {Opt_offusrjquota, "usrjquota="}, 233 {Opt_offgrpjquota, "grpjquota="}, 234 {Opt_offprjjquota, "prjjquota="}, 235 {Opt_jqfmt_vfsold, "jqfmt=vfsold"}, 236 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"}, 237 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"}, 238 {Opt_alloc, "alloc_mode=%s"}, 239 {Opt_fsync, "fsync_mode=%s"}, 240 {Opt_test_dummy_encryption, "test_dummy_encryption=%s"}, 241 {Opt_test_dummy_encryption, "test_dummy_encryption"}, 242 {Opt_inlinecrypt, "inlinecrypt"}, 243 {Opt_checkpoint_disable, "checkpoint=disable"}, 244 {Opt_checkpoint_disable_cap, "checkpoint=disable:%u"}, 245 {Opt_checkpoint_disable_cap_perc, "checkpoint=disable:%u%%"}, 246 {Opt_checkpoint_enable, "checkpoint=enable"}, 247 {Opt_checkpoint_merge, "checkpoint_merge"}, 248 {Opt_nocheckpoint_merge, "nocheckpoint_merge"}, 249 {Opt_compress_algorithm, "compress_algorithm=%s"}, 250 {Opt_compress_log_size, "compress_log_size=%u"}, 251 {Opt_compress_extension, "compress_extension=%s"}, 252 {Opt_nocompress_extension, "nocompress_extension=%s"}, 253 {Opt_compress_chksum, "compress_chksum"}, 254 {Opt_compress_mode, "compress_mode=%s"}, 255 {Opt_compress_cache, "compress_cache"}, 256 {Opt_atgc, "atgc"}, 257 {Opt_gc_merge, "gc_merge"}, 258 {Opt_nogc_merge, "nogc_merge"}, 259 {Opt_discard_unit, "discard_unit=%s"}, 260 {Opt_memory_mode, "memory=%s"}, 261 {Opt_age_extent_cache, "age_extent_cache"}, 262 {Opt_errors, "errors=%s"}, 263 {Opt_err, NULL}, 264 }; 265 266 void f2fs_printk(struct f2fs_sb_info *sbi, const char *fmt, ...) 267 { 268 struct va_format vaf; 269 va_list args; 270 int level; 271 272 va_start(args, fmt); 273 274 level = printk_get_level(fmt); 275 vaf.fmt = printk_skip_level(fmt); 276 vaf.va = &args; 277 printk("%c%cF2FS-fs (%s): %pV\n", 278 KERN_SOH_ASCII, level, sbi->sb->s_id, &vaf); 279 280 va_end(args); 281 } 282 283 #if IS_ENABLED(CONFIG_UNICODE) 284 static const struct f2fs_sb_encodings { 285 __u16 magic; 286 char *name; 287 unsigned int version; 288 } f2fs_sb_encoding_map[] = { 289 {F2FS_ENC_UTF8_12_1, "utf8", UNICODE_AGE(12, 1, 0)}, 290 }; 291 292 static const struct f2fs_sb_encodings * 293 f2fs_sb_read_encoding(const struct f2fs_super_block *sb) 294 { 295 __u16 magic = le16_to_cpu(sb->s_encoding); 296 int i; 297 298 for (i = 0; i < ARRAY_SIZE(f2fs_sb_encoding_map); i++) 299 if (magic == f2fs_sb_encoding_map[i].magic) 300 return &f2fs_sb_encoding_map[i]; 301 302 return NULL; 303 } 304 305 struct kmem_cache *f2fs_cf_name_slab; 306 static int __init f2fs_create_casefold_cache(void) 307 { 308 f2fs_cf_name_slab = f2fs_kmem_cache_create("f2fs_casefolded_name", 309 F2FS_NAME_LEN); 310 return f2fs_cf_name_slab ? 0 : -ENOMEM; 311 } 312 313 static void f2fs_destroy_casefold_cache(void) 314 { 315 kmem_cache_destroy(f2fs_cf_name_slab); 316 } 317 #else 318 static int __init f2fs_create_casefold_cache(void) { return 0; } 319 static void f2fs_destroy_casefold_cache(void) { } 320 #endif 321 322 static inline void limit_reserve_root(struct f2fs_sb_info *sbi) 323 { 324 block_t limit = min((sbi->user_block_count >> 3), 325 sbi->user_block_count - sbi->reserved_blocks); 326 327 /* limit is 12.5% */ 328 if (test_opt(sbi, RESERVE_ROOT) && 329 F2FS_OPTION(sbi).root_reserved_blocks > limit) { 330 F2FS_OPTION(sbi).root_reserved_blocks = limit; 331 f2fs_info(sbi, "Reduce reserved blocks for root = %u", 332 F2FS_OPTION(sbi).root_reserved_blocks); 333 } 334 if (!test_opt(sbi, RESERVE_ROOT) && 335 (!uid_eq(F2FS_OPTION(sbi).s_resuid, 336 make_kuid(&init_user_ns, F2FS_DEF_RESUID)) || 337 !gid_eq(F2FS_OPTION(sbi).s_resgid, 338 make_kgid(&init_user_ns, F2FS_DEF_RESGID)))) 339 f2fs_info(sbi, "Ignore s_resuid=%u, s_resgid=%u w/o reserve_root", 340 from_kuid_munged(&init_user_ns, 341 F2FS_OPTION(sbi).s_resuid), 342 from_kgid_munged(&init_user_ns, 343 F2FS_OPTION(sbi).s_resgid)); 344 } 345 346 static inline int adjust_reserved_segment(struct f2fs_sb_info *sbi) 347 { 348 unsigned int sec_blks = sbi->blocks_per_seg * sbi->segs_per_sec; 349 unsigned int avg_vblocks; 350 unsigned int wanted_reserved_segments; 351 block_t avail_user_block_count; 352 353 if (!F2FS_IO_ALIGNED(sbi)) 354 return 0; 355 356 /* average valid block count in section in worst case */ 357 avg_vblocks = sec_blks / F2FS_IO_SIZE(sbi); 358 359 /* 360 * we need enough free space when migrating one section in worst case 361 */ 362 wanted_reserved_segments = (F2FS_IO_SIZE(sbi) / avg_vblocks) * 363 reserved_segments(sbi); 364 wanted_reserved_segments -= reserved_segments(sbi); 365 366 avail_user_block_count = sbi->user_block_count - 367 sbi->current_reserved_blocks - 368 F2FS_OPTION(sbi).root_reserved_blocks; 369 370 if (wanted_reserved_segments * sbi->blocks_per_seg > 371 avail_user_block_count) { 372 f2fs_err(sbi, "IO align feature can't grab additional reserved segment: %u, available segments: %u", 373 wanted_reserved_segments, 374 avail_user_block_count >> sbi->log_blocks_per_seg); 375 return -ENOSPC; 376 } 377 378 SM_I(sbi)->additional_reserved_segments = wanted_reserved_segments; 379 380 f2fs_info(sbi, "IO align feature needs additional reserved segment: %u", 381 wanted_reserved_segments); 382 383 return 0; 384 } 385 386 static inline void adjust_unusable_cap_perc(struct f2fs_sb_info *sbi) 387 { 388 if (!F2FS_OPTION(sbi).unusable_cap_perc) 389 return; 390 391 if (F2FS_OPTION(sbi).unusable_cap_perc == 100) 392 F2FS_OPTION(sbi).unusable_cap = sbi->user_block_count; 393 else 394 F2FS_OPTION(sbi).unusable_cap = (sbi->user_block_count / 100) * 395 F2FS_OPTION(sbi).unusable_cap_perc; 396 397 f2fs_info(sbi, "Adjust unusable cap for checkpoint=disable = %u / %u%%", 398 F2FS_OPTION(sbi).unusable_cap, 399 F2FS_OPTION(sbi).unusable_cap_perc); 400 } 401 402 static void init_once(void *foo) 403 { 404 struct f2fs_inode_info *fi = (struct f2fs_inode_info *) foo; 405 406 inode_init_once(&fi->vfs_inode); 407 } 408 409 #ifdef CONFIG_QUOTA 410 static const char * const quotatypes[] = INITQFNAMES; 411 #define QTYPE2NAME(t) (quotatypes[t]) 412 static int f2fs_set_qf_name(struct super_block *sb, int qtype, 413 substring_t *args) 414 { 415 struct f2fs_sb_info *sbi = F2FS_SB(sb); 416 char *qname; 417 int ret = -EINVAL; 418 419 if (sb_any_quota_loaded(sb) && !F2FS_OPTION(sbi).s_qf_names[qtype]) { 420 f2fs_err(sbi, "Cannot change journaled quota options when quota turned on"); 421 return -EINVAL; 422 } 423 if (f2fs_sb_has_quota_ino(sbi)) { 424 f2fs_info(sbi, "QUOTA feature is enabled, so ignore qf_name"); 425 return 0; 426 } 427 428 qname = match_strdup(args); 429 if (!qname) { 430 f2fs_err(sbi, "Not enough memory for storing quotafile name"); 431 return -ENOMEM; 432 } 433 if (F2FS_OPTION(sbi).s_qf_names[qtype]) { 434 if (strcmp(F2FS_OPTION(sbi).s_qf_names[qtype], qname) == 0) 435 ret = 0; 436 else 437 f2fs_err(sbi, "%s quota file already specified", 438 QTYPE2NAME(qtype)); 439 goto errout; 440 } 441 if (strchr(qname, '/')) { 442 f2fs_err(sbi, "quotafile must be on filesystem root"); 443 goto errout; 444 } 445 F2FS_OPTION(sbi).s_qf_names[qtype] = qname; 446 set_opt(sbi, QUOTA); 447 return 0; 448 errout: 449 kfree(qname); 450 return ret; 451 } 452 453 static int f2fs_clear_qf_name(struct super_block *sb, int qtype) 454 { 455 struct f2fs_sb_info *sbi = F2FS_SB(sb); 456 457 if (sb_any_quota_loaded(sb) && F2FS_OPTION(sbi).s_qf_names[qtype]) { 458 f2fs_err(sbi, "Cannot change journaled quota options when quota turned on"); 459 return -EINVAL; 460 } 461 kfree(F2FS_OPTION(sbi).s_qf_names[qtype]); 462 F2FS_OPTION(sbi).s_qf_names[qtype] = NULL; 463 return 0; 464 } 465 466 static int f2fs_check_quota_options(struct f2fs_sb_info *sbi) 467 { 468 /* 469 * We do the test below only for project quotas. 'usrquota' and 470 * 'grpquota' mount options are allowed even without quota feature 471 * to support legacy quotas in quota files. 472 */ 473 if (test_opt(sbi, PRJQUOTA) && !f2fs_sb_has_project_quota(sbi)) { 474 f2fs_err(sbi, "Project quota feature not enabled. Cannot enable project quota enforcement."); 475 return -1; 476 } 477 if (F2FS_OPTION(sbi).s_qf_names[USRQUOTA] || 478 F2FS_OPTION(sbi).s_qf_names[GRPQUOTA] || 479 F2FS_OPTION(sbi).s_qf_names[PRJQUOTA]) { 480 if (test_opt(sbi, USRQUOTA) && 481 F2FS_OPTION(sbi).s_qf_names[USRQUOTA]) 482 clear_opt(sbi, USRQUOTA); 483 484 if (test_opt(sbi, GRPQUOTA) && 485 F2FS_OPTION(sbi).s_qf_names[GRPQUOTA]) 486 clear_opt(sbi, GRPQUOTA); 487 488 if (test_opt(sbi, PRJQUOTA) && 489 F2FS_OPTION(sbi).s_qf_names[PRJQUOTA]) 490 clear_opt(sbi, PRJQUOTA); 491 492 if (test_opt(sbi, GRPQUOTA) || test_opt(sbi, USRQUOTA) || 493 test_opt(sbi, PRJQUOTA)) { 494 f2fs_err(sbi, "old and new quota format mixing"); 495 return -1; 496 } 497 498 if (!F2FS_OPTION(sbi).s_jquota_fmt) { 499 f2fs_err(sbi, "journaled quota format not specified"); 500 return -1; 501 } 502 } 503 504 if (f2fs_sb_has_quota_ino(sbi) && F2FS_OPTION(sbi).s_jquota_fmt) { 505 f2fs_info(sbi, "QUOTA feature is enabled, so ignore jquota_fmt"); 506 F2FS_OPTION(sbi).s_jquota_fmt = 0; 507 } 508 return 0; 509 } 510 #endif 511 512 static int f2fs_set_test_dummy_encryption(struct super_block *sb, 513 const char *opt, 514 const substring_t *arg, 515 bool is_remount) 516 { 517 struct f2fs_sb_info *sbi = F2FS_SB(sb); 518 struct fs_parameter param = { 519 .type = fs_value_is_string, 520 .string = arg->from ? arg->from : "", 521 }; 522 struct fscrypt_dummy_policy *policy = 523 &F2FS_OPTION(sbi).dummy_enc_policy; 524 int err; 525 526 if (!IS_ENABLED(CONFIG_FS_ENCRYPTION)) { 527 f2fs_warn(sbi, "test_dummy_encryption option not supported"); 528 return -EINVAL; 529 } 530 531 if (!f2fs_sb_has_encrypt(sbi)) { 532 f2fs_err(sbi, "Encrypt feature is off"); 533 return -EINVAL; 534 } 535 536 /* 537 * This mount option is just for testing, and it's not worthwhile to 538 * implement the extra complexity (e.g. RCU protection) that would be 539 * needed to allow it to be set or changed during remount. We do allow 540 * it to be specified during remount, but only if there is no change. 541 */ 542 if (is_remount && !fscrypt_is_dummy_policy_set(policy)) { 543 f2fs_warn(sbi, "Can't set test_dummy_encryption on remount"); 544 return -EINVAL; 545 } 546 547 err = fscrypt_parse_test_dummy_encryption(¶m, policy); 548 if (err) { 549 if (err == -EEXIST) 550 f2fs_warn(sbi, 551 "Can't change test_dummy_encryption on remount"); 552 else if (err == -EINVAL) 553 f2fs_warn(sbi, "Value of option \"%s\" is unrecognized", 554 opt); 555 else 556 f2fs_warn(sbi, "Error processing option \"%s\" [%d]", 557 opt, err); 558 return -EINVAL; 559 } 560 f2fs_warn(sbi, "Test dummy encryption mode enabled"); 561 return 0; 562 } 563 564 #ifdef CONFIG_F2FS_FS_COMPRESSION 565 /* 566 * 1. The same extension name cannot not appear in both compress and non-compress extension 567 * at the same time. 568 * 2. If the compress extension specifies all files, the types specified by the non-compress 569 * extension will be treated as special cases and will not be compressed. 570 * 3. Don't allow the non-compress extension specifies all files. 571 */ 572 static int f2fs_test_compress_extension(struct f2fs_sb_info *sbi) 573 { 574 unsigned char (*ext)[F2FS_EXTENSION_LEN]; 575 unsigned char (*noext)[F2FS_EXTENSION_LEN]; 576 int ext_cnt, noext_cnt, index = 0, no_index = 0; 577 578 ext = F2FS_OPTION(sbi).extensions; 579 ext_cnt = F2FS_OPTION(sbi).compress_ext_cnt; 580 noext = F2FS_OPTION(sbi).noextensions; 581 noext_cnt = F2FS_OPTION(sbi).nocompress_ext_cnt; 582 583 if (!noext_cnt) 584 return 0; 585 586 for (no_index = 0; no_index < noext_cnt; no_index++) { 587 if (!strcasecmp("*", noext[no_index])) { 588 f2fs_info(sbi, "Don't allow the nocompress extension specifies all files"); 589 return -EINVAL; 590 } 591 for (index = 0; index < ext_cnt; index++) { 592 if (!strcasecmp(ext[index], noext[no_index])) { 593 f2fs_info(sbi, "Don't allow the same extension %s appear in both compress and nocompress extension", 594 ext[index]); 595 return -EINVAL; 596 } 597 } 598 } 599 return 0; 600 } 601 602 #ifdef CONFIG_F2FS_FS_LZ4 603 static int f2fs_set_lz4hc_level(struct f2fs_sb_info *sbi, const char *str) 604 { 605 #ifdef CONFIG_F2FS_FS_LZ4HC 606 unsigned int level; 607 608 if (strlen(str) == 3) { 609 F2FS_OPTION(sbi).compress_level = 0; 610 return 0; 611 } 612 613 str += 3; 614 615 if (str[0] != ':') { 616 f2fs_info(sbi, "wrong format, e.g. <alg_name>:<compr_level>"); 617 return -EINVAL; 618 } 619 if (kstrtouint(str + 1, 10, &level)) 620 return -EINVAL; 621 622 if (!f2fs_is_compress_level_valid(COMPRESS_LZ4, level)) { 623 f2fs_info(sbi, "invalid lz4hc compress level: %d", level); 624 return -EINVAL; 625 } 626 627 F2FS_OPTION(sbi).compress_level = level; 628 return 0; 629 #else 630 if (strlen(str) == 3) { 631 F2FS_OPTION(sbi).compress_level = 0; 632 return 0; 633 } 634 f2fs_info(sbi, "kernel doesn't support lz4hc compression"); 635 return -EINVAL; 636 #endif 637 } 638 #endif 639 640 #ifdef CONFIG_F2FS_FS_ZSTD 641 static int f2fs_set_zstd_level(struct f2fs_sb_info *sbi, const char *str) 642 { 643 unsigned int level; 644 int len = 4; 645 646 if (strlen(str) == len) { 647 F2FS_OPTION(sbi).compress_level = F2FS_ZSTD_DEFAULT_CLEVEL; 648 return 0; 649 } 650 651 str += len; 652 653 if (str[0] != ':') { 654 f2fs_info(sbi, "wrong format, e.g. <alg_name>:<compr_level>"); 655 return -EINVAL; 656 } 657 if (kstrtouint(str + 1, 10, &level)) 658 return -EINVAL; 659 660 if (!f2fs_is_compress_level_valid(COMPRESS_ZSTD, level)) { 661 f2fs_info(sbi, "invalid zstd compress level: %d", level); 662 return -EINVAL; 663 } 664 665 F2FS_OPTION(sbi).compress_level = level; 666 return 0; 667 } 668 #endif 669 #endif 670 671 static int parse_options(struct super_block *sb, char *options, bool is_remount) 672 { 673 struct f2fs_sb_info *sbi = F2FS_SB(sb); 674 substring_t args[MAX_OPT_ARGS]; 675 #ifdef CONFIG_F2FS_FS_COMPRESSION 676 unsigned char (*ext)[F2FS_EXTENSION_LEN]; 677 unsigned char (*noext)[F2FS_EXTENSION_LEN]; 678 int ext_cnt, noext_cnt; 679 #endif 680 char *p, *name; 681 int arg = 0; 682 kuid_t uid; 683 kgid_t gid; 684 int ret; 685 686 if (!options) 687 goto default_check; 688 689 while ((p = strsep(&options, ",")) != NULL) { 690 int token; 691 692 if (!*p) 693 continue; 694 /* 695 * Initialize args struct so we know whether arg was 696 * found; some options take optional arguments. 697 */ 698 args[0].to = args[0].from = NULL; 699 token = match_token(p, f2fs_tokens, args); 700 701 switch (token) { 702 case Opt_gc_background: 703 name = match_strdup(&args[0]); 704 705 if (!name) 706 return -ENOMEM; 707 if (!strcmp(name, "on")) { 708 F2FS_OPTION(sbi).bggc_mode = BGGC_MODE_ON; 709 } else if (!strcmp(name, "off")) { 710 F2FS_OPTION(sbi).bggc_mode = BGGC_MODE_OFF; 711 } else if (!strcmp(name, "sync")) { 712 F2FS_OPTION(sbi).bggc_mode = BGGC_MODE_SYNC; 713 } else { 714 kfree(name); 715 return -EINVAL; 716 } 717 kfree(name); 718 break; 719 case Opt_disable_roll_forward: 720 set_opt(sbi, DISABLE_ROLL_FORWARD); 721 break; 722 case Opt_norecovery: 723 /* this option mounts f2fs with ro */ 724 set_opt(sbi, NORECOVERY); 725 if (!f2fs_readonly(sb)) 726 return -EINVAL; 727 break; 728 case Opt_discard: 729 if (!f2fs_hw_support_discard(sbi)) { 730 f2fs_warn(sbi, "device does not support discard"); 731 break; 732 } 733 set_opt(sbi, DISCARD); 734 break; 735 case Opt_nodiscard: 736 if (f2fs_hw_should_discard(sbi)) { 737 f2fs_warn(sbi, "discard is required for zoned block devices"); 738 return -EINVAL; 739 } 740 clear_opt(sbi, DISCARD); 741 break; 742 case Opt_noheap: 743 set_opt(sbi, NOHEAP); 744 break; 745 case Opt_heap: 746 clear_opt(sbi, NOHEAP); 747 break; 748 #ifdef CONFIG_F2FS_FS_XATTR 749 case Opt_user_xattr: 750 set_opt(sbi, XATTR_USER); 751 break; 752 case Opt_nouser_xattr: 753 clear_opt(sbi, XATTR_USER); 754 break; 755 case Opt_inline_xattr: 756 set_opt(sbi, INLINE_XATTR); 757 break; 758 case Opt_noinline_xattr: 759 clear_opt(sbi, INLINE_XATTR); 760 break; 761 case Opt_inline_xattr_size: 762 if (args->from && match_int(args, &arg)) 763 return -EINVAL; 764 set_opt(sbi, INLINE_XATTR_SIZE); 765 F2FS_OPTION(sbi).inline_xattr_size = arg; 766 break; 767 #else 768 case Opt_user_xattr: 769 f2fs_info(sbi, "user_xattr options not supported"); 770 break; 771 case Opt_nouser_xattr: 772 f2fs_info(sbi, "nouser_xattr options not supported"); 773 break; 774 case Opt_inline_xattr: 775 f2fs_info(sbi, "inline_xattr options not supported"); 776 break; 777 case Opt_noinline_xattr: 778 f2fs_info(sbi, "noinline_xattr options not supported"); 779 break; 780 #endif 781 #ifdef CONFIG_F2FS_FS_POSIX_ACL 782 case Opt_acl: 783 set_opt(sbi, POSIX_ACL); 784 break; 785 case Opt_noacl: 786 clear_opt(sbi, POSIX_ACL); 787 break; 788 #else 789 case Opt_acl: 790 f2fs_info(sbi, "acl options not supported"); 791 break; 792 case Opt_noacl: 793 f2fs_info(sbi, "noacl options not supported"); 794 break; 795 #endif 796 case Opt_active_logs: 797 if (args->from && match_int(args, &arg)) 798 return -EINVAL; 799 if (arg != 2 && arg != 4 && 800 arg != NR_CURSEG_PERSIST_TYPE) 801 return -EINVAL; 802 F2FS_OPTION(sbi).active_logs = arg; 803 break; 804 case Opt_disable_ext_identify: 805 set_opt(sbi, DISABLE_EXT_IDENTIFY); 806 break; 807 case Opt_inline_data: 808 set_opt(sbi, INLINE_DATA); 809 break; 810 case Opt_inline_dentry: 811 set_opt(sbi, INLINE_DENTRY); 812 break; 813 case Opt_noinline_dentry: 814 clear_opt(sbi, INLINE_DENTRY); 815 break; 816 case Opt_flush_merge: 817 set_opt(sbi, FLUSH_MERGE); 818 break; 819 case Opt_noflush_merge: 820 clear_opt(sbi, FLUSH_MERGE); 821 break; 822 case Opt_nobarrier: 823 set_opt(sbi, NOBARRIER); 824 break; 825 case Opt_barrier: 826 clear_opt(sbi, NOBARRIER); 827 break; 828 case Opt_fastboot: 829 set_opt(sbi, FASTBOOT); 830 break; 831 case Opt_extent_cache: 832 set_opt(sbi, READ_EXTENT_CACHE); 833 break; 834 case Opt_noextent_cache: 835 clear_opt(sbi, READ_EXTENT_CACHE); 836 break; 837 case Opt_noinline_data: 838 clear_opt(sbi, INLINE_DATA); 839 break; 840 case Opt_data_flush: 841 set_opt(sbi, DATA_FLUSH); 842 break; 843 case Opt_reserve_root: 844 if (args->from && match_int(args, &arg)) 845 return -EINVAL; 846 if (test_opt(sbi, RESERVE_ROOT)) { 847 f2fs_info(sbi, "Preserve previous reserve_root=%u", 848 F2FS_OPTION(sbi).root_reserved_blocks); 849 } else { 850 F2FS_OPTION(sbi).root_reserved_blocks = arg; 851 set_opt(sbi, RESERVE_ROOT); 852 } 853 break; 854 case Opt_resuid: 855 if (args->from && match_int(args, &arg)) 856 return -EINVAL; 857 uid = make_kuid(current_user_ns(), arg); 858 if (!uid_valid(uid)) { 859 f2fs_err(sbi, "Invalid uid value %d", arg); 860 return -EINVAL; 861 } 862 F2FS_OPTION(sbi).s_resuid = uid; 863 break; 864 case Opt_resgid: 865 if (args->from && match_int(args, &arg)) 866 return -EINVAL; 867 gid = make_kgid(current_user_ns(), arg); 868 if (!gid_valid(gid)) { 869 f2fs_err(sbi, "Invalid gid value %d", arg); 870 return -EINVAL; 871 } 872 F2FS_OPTION(sbi).s_resgid = gid; 873 break; 874 case Opt_mode: 875 name = match_strdup(&args[0]); 876 877 if (!name) 878 return -ENOMEM; 879 if (!strcmp(name, "adaptive")) { 880 F2FS_OPTION(sbi).fs_mode = FS_MODE_ADAPTIVE; 881 } else if (!strcmp(name, "lfs")) { 882 F2FS_OPTION(sbi).fs_mode = FS_MODE_LFS; 883 } else if (!strcmp(name, "fragment:segment")) { 884 F2FS_OPTION(sbi).fs_mode = FS_MODE_FRAGMENT_SEG; 885 } else if (!strcmp(name, "fragment:block")) { 886 F2FS_OPTION(sbi).fs_mode = FS_MODE_FRAGMENT_BLK; 887 } else { 888 kfree(name); 889 return -EINVAL; 890 } 891 kfree(name); 892 break; 893 case Opt_io_size_bits: 894 if (args->from && match_int(args, &arg)) 895 return -EINVAL; 896 if (arg <= 0 || arg > __ilog2_u32(BIO_MAX_VECS)) { 897 f2fs_warn(sbi, "Not support %ld, larger than %d", 898 BIT(arg), BIO_MAX_VECS); 899 return -EINVAL; 900 } 901 F2FS_OPTION(sbi).write_io_size_bits = arg; 902 break; 903 #ifdef CONFIG_F2FS_FAULT_INJECTION 904 case Opt_fault_injection: 905 if (args->from && match_int(args, &arg)) 906 return -EINVAL; 907 f2fs_build_fault_attr(sbi, arg, F2FS_ALL_FAULT_TYPE); 908 set_opt(sbi, FAULT_INJECTION); 909 break; 910 911 case Opt_fault_type: 912 if (args->from && match_int(args, &arg)) 913 return -EINVAL; 914 f2fs_build_fault_attr(sbi, 0, arg); 915 set_opt(sbi, FAULT_INJECTION); 916 break; 917 #else 918 case Opt_fault_injection: 919 f2fs_info(sbi, "fault_injection options not supported"); 920 break; 921 922 case Opt_fault_type: 923 f2fs_info(sbi, "fault_type options not supported"); 924 break; 925 #endif 926 case Opt_lazytime: 927 sb->s_flags |= SB_LAZYTIME; 928 break; 929 case Opt_nolazytime: 930 sb->s_flags &= ~SB_LAZYTIME; 931 break; 932 #ifdef CONFIG_QUOTA 933 case Opt_quota: 934 case Opt_usrquota: 935 set_opt(sbi, USRQUOTA); 936 break; 937 case Opt_grpquota: 938 set_opt(sbi, GRPQUOTA); 939 break; 940 case Opt_prjquota: 941 set_opt(sbi, PRJQUOTA); 942 break; 943 case Opt_usrjquota: 944 ret = f2fs_set_qf_name(sb, USRQUOTA, &args[0]); 945 if (ret) 946 return ret; 947 break; 948 case Opt_grpjquota: 949 ret = f2fs_set_qf_name(sb, GRPQUOTA, &args[0]); 950 if (ret) 951 return ret; 952 break; 953 case Opt_prjjquota: 954 ret = f2fs_set_qf_name(sb, PRJQUOTA, &args[0]); 955 if (ret) 956 return ret; 957 break; 958 case Opt_offusrjquota: 959 ret = f2fs_clear_qf_name(sb, USRQUOTA); 960 if (ret) 961 return ret; 962 break; 963 case Opt_offgrpjquota: 964 ret = f2fs_clear_qf_name(sb, GRPQUOTA); 965 if (ret) 966 return ret; 967 break; 968 case Opt_offprjjquota: 969 ret = f2fs_clear_qf_name(sb, PRJQUOTA); 970 if (ret) 971 return ret; 972 break; 973 case Opt_jqfmt_vfsold: 974 F2FS_OPTION(sbi).s_jquota_fmt = QFMT_VFS_OLD; 975 break; 976 case Opt_jqfmt_vfsv0: 977 F2FS_OPTION(sbi).s_jquota_fmt = QFMT_VFS_V0; 978 break; 979 case Opt_jqfmt_vfsv1: 980 F2FS_OPTION(sbi).s_jquota_fmt = QFMT_VFS_V1; 981 break; 982 case Opt_noquota: 983 clear_opt(sbi, QUOTA); 984 clear_opt(sbi, USRQUOTA); 985 clear_opt(sbi, GRPQUOTA); 986 clear_opt(sbi, PRJQUOTA); 987 break; 988 #else 989 case Opt_quota: 990 case Opt_usrquota: 991 case Opt_grpquota: 992 case Opt_prjquota: 993 case Opt_usrjquota: 994 case Opt_grpjquota: 995 case Opt_prjjquota: 996 case Opt_offusrjquota: 997 case Opt_offgrpjquota: 998 case Opt_offprjjquota: 999 case Opt_jqfmt_vfsold: 1000 case Opt_jqfmt_vfsv0: 1001 case Opt_jqfmt_vfsv1: 1002 case Opt_noquota: 1003 f2fs_info(sbi, "quota operations not supported"); 1004 break; 1005 #endif 1006 case Opt_alloc: 1007 name = match_strdup(&args[0]); 1008 if (!name) 1009 return -ENOMEM; 1010 1011 if (!strcmp(name, "default")) { 1012 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_DEFAULT; 1013 } else if (!strcmp(name, "reuse")) { 1014 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_REUSE; 1015 } else { 1016 kfree(name); 1017 return -EINVAL; 1018 } 1019 kfree(name); 1020 break; 1021 case Opt_fsync: 1022 name = match_strdup(&args[0]); 1023 if (!name) 1024 return -ENOMEM; 1025 if (!strcmp(name, "posix")) { 1026 F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_POSIX; 1027 } else if (!strcmp(name, "strict")) { 1028 F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_STRICT; 1029 } else if (!strcmp(name, "nobarrier")) { 1030 F2FS_OPTION(sbi).fsync_mode = 1031 FSYNC_MODE_NOBARRIER; 1032 } else { 1033 kfree(name); 1034 return -EINVAL; 1035 } 1036 kfree(name); 1037 break; 1038 case Opt_test_dummy_encryption: 1039 ret = f2fs_set_test_dummy_encryption(sb, p, &args[0], 1040 is_remount); 1041 if (ret) 1042 return ret; 1043 break; 1044 case Opt_inlinecrypt: 1045 #ifdef CONFIG_FS_ENCRYPTION_INLINE_CRYPT 1046 sb->s_flags |= SB_INLINECRYPT; 1047 #else 1048 f2fs_info(sbi, "inline encryption not supported"); 1049 #endif 1050 break; 1051 case Opt_checkpoint_disable_cap_perc: 1052 if (args->from && match_int(args, &arg)) 1053 return -EINVAL; 1054 if (arg < 0 || arg > 100) 1055 return -EINVAL; 1056 F2FS_OPTION(sbi).unusable_cap_perc = arg; 1057 set_opt(sbi, DISABLE_CHECKPOINT); 1058 break; 1059 case Opt_checkpoint_disable_cap: 1060 if (args->from && match_int(args, &arg)) 1061 return -EINVAL; 1062 F2FS_OPTION(sbi).unusable_cap = arg; 1063 set_opt(sbi, DISABLE_CHECKPOINT); 1064 break; 1065 case Opt_checkpoint_disable: 1066 set_opt(sbi, DISABLE_CHECKPOINT); 1067 break; 1068 case Opt_checkpoint_enable: 1069 clear_opt(sbi, DISABLE_CHECKPOINT); 1070 break; 1071 case Opt_checkpoint_merge: 1072 set_opt(sbi, MERGE_CHECKPOINT); 1073 break; 1074 case Opt_nocheckpoint_merge: 1075 clear_opt(sbi, MERGE_CHECKPOINT); 1076 break; 1077 #ifdef CONFIG_F2FS_FS_COMPRESSION 1078 case Opt_compress_algorithm: 1079 if (!f2fs_sb_has_compression(sbi)) { 1080 f2fs_info(sbi, "Image doesn't support compression"); 1081 break; 1082 } 1083 name = match_strdup(&args[0]); 1084 if (!name) 1085 return -ENOMEM; 1086 if (!strcmp(name, "lzo")) { 1087 #ifdef CONFIG_F2FS_FS_LZO 1088 F2FS_OPTION(sbi).compress_level = 0; 1089 F2FS_OPTION(sbi).compress_algorithm = 1090 COMPRESS_LZO; 1091 #else 1092 f2fs_info(sbi, "kernel doesn't support lzo compression"); 1093 #endif 1094 } else if (!strncmp(name, "lz4", 3)) { 1095 #ifdef CONFIG_F2FS_FS_LZ4 1096 ret = f2fs_set_lz4hc_level(sbi, name); 1097 if (ret) { 1098 kfree(name); 1099 return -EINVAL; 1100 } 1101 F2FS_OPTION(sbi).compress_algorithm = 1102 COMPRESS_LZ4; 1103 #else 1104 f2fs_info(sbi, "kernel doesn't support lz4 compression"); 1105 #endif 1106 } else if (!strncmp(name, "zstd", 4)) { 1107 #ifdef CONFIG_F2FS_FS_ZSTD 1108 ret = f2fs_set_zstd_level(sbi, name); 1109 if (ret) { 1110 kfree(name); 1111 return -EINVAL; 1112 } 1113 F2FS_OPTION(sbi).compress_algorithm = 1114 COMPRESS_ZSTD; 1115 #else 1116 f2fs_info(sbi, "kernel doesn't support zstd compression"); 1117 #endif 1118 } else if (!strcmp(name, "lzo-rle")) { 1119 #ifdef CONFIG_F2FS_FS_LZORLE 1120 F2FS_OPTION(sbi).compress_level = 0; 1121 F2FS_OPTION(sbi).compress_algorithm = 1122 COMPRESS_LZORLE; 1123 #else 1124 f2fs_info(sbi, "kernel doesn't support lzorle compression"); 1125 #endif 1126 } else { 1127 kfree(name); 1128 return -EINVAL; 1129 } 1130 kfree(name); 1131 break; 1132 case Opt_compress_log_size: 1133 if (!f2fs_sb_has_compression(sbi)) { 1134 f2fs_info(sbi, "Image doesn't support compression"); 1135 break; 1136 } 1137 if (args->from && match_int(args, &arg)) 1138 return -EINVAL; 1139 if (arg < MIN_COMPRESS_LOG_SIZE || 1140 arg > MAX_COMPRESS_LOG_SIZE) { 1141 f2fs_err(sbi, 1142 "Compress cluster log size is out of range"); 1143 return -EINVAL; 1144 } 1145 F2FS_OPTION(sbi).compress_log_size = arg; 1146 break; 1147 case Opt_compress_extension: 1148 if (!f2fs_sb_has_compression(sbi)) { 1149 f2fs_info(sbi, "Image doesn't support compression"); 1150 break; 1151 } 1152 name = match_strdup(&args[0]); 1153 if (!name) 1154 return -ENOMEM; 1155 1156 ext = F2FS_OPTION(sbi).extensions; 1157 ext_cnt = F2FS_OPTION(sbi).compress_ext_cnt; 1158 1159 if (strlen(name) >= F2FS_EXTENSION_LEN || 1160 ext_cnt >= COMPRESS_EXT_NUM) { 1161 f2fs_err(sbi, 1162 "invalid extension length/number"); 1163 kfree(name); 1164 return -EINVAL; 1165 } 1166 1167 strcpy(ext[ext_cnt], name); 1168 F2FS_OPTION(sbi).compress_ext_cnt++; 1169 kfree(name); 1170 break; 1171 case Opt_nocompress_extension: 1172 if (!f2fs_sb_has_compression(sbi)) { 1173 f2fs_info(sbi, "Image doesn't support compression"); 1174 break; 1175 } 1176 name = match_strdup(&args[0]); 1177 if (!name) 1178 return -ENOMEM; 1179 1180 noext = F2FS_OPTION(sbi).noextensions; 1181 noext_cnt = F2FS_OPTION(sbi).nocompress_ext_cnt; 1182 1183 if (strlen(name) >= F2FS_EXTENSION_LEN || 1184 noext_cnt >= COMPRESS_EXT_NUM) { 1185 f2fs_err(sbi, 1186 "invalid extension length/number"); 1187 kfree(name); 1188 return -EINVAL; 1189 } 1190 1191 strcpy(noext[noext_cnt], name); 1192 F2FS_OPTION(sbi).nocompress_ext_cnt++; 1193 kfree(name); 1194 break; 1195 case Opt_compress_chksum: 1196 if (!f2fs_sb_has_compression(sbi)) { 1197 f2fs_info(sbi, "Image doesn't support compression"); 1198 break; 1199 } 1200 F2FS_OPTION(sbi).compress_chksum = true; 1201 break; 1202 case Opt_compress_mode: 1203 if (!f2fs_sb_has_compression(sbi)) { 1204 f2fs_info(sbi, "Image doesn't support compression"); 1205 break; 1206 } 1207 name = match_strdup(&args[0]); 1208 if (!name) 1209 return -ENOMEM; 1210 if (!strcmp(name, "fs")) { 1211 F2FS_OPTION(sbi).compress_mode = COMPR_MODE_FS; 1212 } else if (!strcmp(name, "user")) { 1213 F2FS_OPTION(sbi).compress_mode = COMPR_MODE_USER; 1214 } else { 1215 kfree(name); 1216 return -EINVAL; 1217 } 1218 kfree(name); 1219 break; 1220 case Opt_compress_cache: 1221 if (!f2fs_sb_has_compression(sbi)) { 1222 f2fs_info(sbi, "Image doesn't support compression"); 1223 break; 1224 } 1225 set_opt(sbi, COMPRESS_CACHE); 1226 break; 1227 #else 1228 case Opt_compress_algorithm: 1229 case Opt_compress_log_size: 1230 case Opt_compress_extension: 1231 case Opt_nocompress_extension: 1232 case Opt_compress_chksum: 1233 case Opt_compress_mode: 1234 case Opt_compress_cache: 1235 f2fs_info(sbi, "compression options not supported"); 1236 break; 1237 #endif 1238 case Opt_atgc: 1239 set_opt(sbi, ATGC); 1240 break; 1241 case Opt_gc_merge: 1242 set_opt(sbi, GC_MERGE); 1243 break; 1244 case Opt_nogc_merge: 1245 clear_opt(sbi, GC_MERGE); 1246 break; 1247 case Opt_discard_unit: 1248 name = match_strdup(&args[0]); 1249 if (!name) 1250 return -ENOMEM; 1251 if (!strcmp(name, "block")) { 1252 F2FS_OPTION(sbi).discard_unit = 1253 DISCARD_UNIT_BLOCK; 1254 } else if (!strcmp(name, "segment")) { 1255 F2FS_OPTION(sbi).discard_unit = 1256 DISCARD_UNIT_SEGMENT; 1257 } else if (!strcmp(name, "section")) { 1258 F2FS_OPTION(sbi).discard_unit = 1259 DISCARD_UNIT_SECTION; 1260 } else { 1261 kfree(name); 1262 return -EINVAL; 1263 } 1264 kfree(name); 1265 break; 1266 case Opt_memory_mode: 1267 name = match_strdup(&args[0]); 1268 if (!name) 1269 return -ENOMEM; 1270 if (!strcmp(name, "normal")) { 1271 F2FS_OPTION(sbi).memory_mode = 1272 MEMORY_MODE_NORMAL; 1273 } else if (!strcmp(name, "low")) { 1274 F2FS_OPTION(sbi).memory_mode = 1275 MEMORY_MODE_LOW; 1276 } else { 1277 kfree(name); 1278 return -EINVAL; 1279 } 1280 kfree(name); 1281 break; 1282 case Opt_age_extent_cache: 1283 set_opt(sbi, AGE_EXTENT_CACHE); 1284 break; 1285 case Opt_errors: 1286 name = match_strdup(&args[0]); 1287 if (!name) 1288 return -ENOMEM; 1289 if (!strcmp(name, "remount-ro")) { 1290 F2FS_OPTION(sbi).errors = 1291 MOUNT_ERRORS_READONLY; 1292 } else if (!strcmp(name, "continue")) { 1293 F2FS_OPTION(sbi).errors = 1294 MOUNT_ERRORS_CONTINUE; 1295 } else if (!strcmp(name, "panic")) { 1296 F2FS_OPTION(sbi).errors = 1297 MOUNT_ERRORS_PANIC; 1298 } else { 1299 kfree(name); 1300 return -EINVAL; 1301 } 1302 kfree(name); 1303 break; 1304 default: 1305 f2fs_err(sbi, "Unrecognized mount option \"%s\" or missing value", 1306 p); 1307 return -EINVAL; 1308 } 1309 } 1310 default_check: 1311 #ifdef CONFIG_QUOTA 1312 if (f2fs_check_quota_options(sbi)) 1313 return -EINVAL; 1314 #else 1315 if (f2fs_sb_has_quota_ino(sbi) && !f2fs_readonly(sbi->sb)) { 1316 f2fs_info(sbi, "Filesystem with quota feature cannot be mounted RDWR without CONFIG_QUOTA"); 1317 return -EINVAL; 1318 } 1319 if (f2fs_sb_has_project_quota(sbi) && !f2fs_readonly(sbi->sb)) { 1320 f2fs_err(sbi, "Filesystem with project quota feature cannot be mounted RDWR without CONFIG_QUOTA"); 1321 return -EINVAL; 1322 } 1323 #endif 1324 #if !IS_ENABLED(CONFIG_UNICODE) 1325 if (f2fs_sb_has_casefold(sbi)) { 1326 f2fs_err(sbi, 1327 "Filesystem with casefold feature cannot be mounted without CONFIG_UNICODE"); 1328 return -EINVAL; 1329 } 1330 #endif 1331 /* 1332 * The BLKZONED feature indicates that the drive was formatted with 1333 * zone alignment optimization. This is optional for host-aware 1334 * devices, but mandatory for host-managed zoned block devices. 1335 */ 1336 if (f2fs_sb_has_blkzoned(sbi)) { 1337 #ifdef CONFIG_BLK_DEV_ZONED 1338 if (F2FS_OPTION(sbi).discard_unit != 1339 DISCARD_UNIT_SECTION) { 1340 f2fs_info(sbi, "Zoned block device doesn't need small discard, set discard_unit=section by default"); 1341 F2FS_OPTION(sbi).discard_unit = 1342 DISCARD_UNIT_SECTION; 1343 } 1344 1345 if (F2FS_OPTION(sbi).fs_mode != FS_MODE_LFS) { 1346 f2fs_info(sbi, "Only lfs mode is allowed with zoned block device feature"); 1347 return -EINVAL; 1348 } 1349 #else 1350 f2fs_err(sbi, "Zoned block device support is not enabled"); 1351 return -EINVAL; 1352 #endif 1353 } 1354 1355 #ifdef CONFIG_F2FS_FS_COMPRESSION 1356 if (f2fs_test_compress_extension(sbi)) { 1357 f2fs_err(sbi, "invalid compress or nocompress extension"); 1358 return -EINVAL; 1359 } 1360 #endif 1361 1362 if (F2FS_IO_SIZE_BITS(sbi) && !f2fs_lfs_mode(sbi)) { 1363 f2fs_err(sbi, "Should set mode=lfs with %luKB-sized IO", 1364 F2FS_IO_SIZE_KB(sbi)); 1365 return -EINVAL; 1366 } 1367 1368 if (test_opt(sbi, INLINE_XATTR_SIZE)) { 1369 int min_size, max_size; 1370 1371 if (!f2fs_sb_has_extra_attr(sbi) || 1372 !f2fs_sb_has_flexible_inline_xattr(sbi)) { 1373 f2fs_err(sbi, "extra_attr or flexible_inline_xattr feature is off"); 1374 return -EINVAL; 1375 } 1376 if (!test_opt(sbi, INLINE_XATTR)) { 1377 f2fs_err(sbi, "inline_xattr_size option should be set with inline_xattr option"); 1378 return -EINVAL; 1379 } 1380 1381 min_size = MIN_INLINE_XATTR_SIZE; 1382 max_size = MAX_INLINE_XATTR_SIZE; 1383 1384 if (F2FS_OPTION(sbi).inline_xattr_size < min_size || 1385 F2FS_OPTION(sbi).inline_xattr_size > max_size) { 1386 f2fs_err(sbi, "inline xattr size is out of range: %d ~ %d", 1387 min_size, max_size); 1388 return -EINVAL; 1389 } 1390 } 1391 1392 if (test_opt(sbi, DISABLE_CHECKPOINT) && f2fs_lfs_mode(sbi)) { 1393 f2fs_err(sbi, "LFS is not compatible with checkpoint=disable"); 1394 return -EINVAL; 1395 } 1396 1397 if (test_opt(sbi, ATGC) && f2fs_lfs_mode(sbi)) { 1398 f2fs_err(sbi, "LFS is not compatible with ATGC"); 1399 return -EINVAL; 1400 } 1401 1402 if (f2fs_is_readonly(sbi) && test_opt(sbi, FLUSH_MERGE)) { 1403 f2fs_err(sbi, "FLUSH_MERGE not compatible with readonly mode"); 1404 return -EINVAL; 1405 } 1406 1407 if (f2fs_sb_has_readonly(sbi) && !f2fs_readonly(sbi->sb)) { 1408 f2fs_err(sbi, "Allow to mount readonly mode only"); 1409 return -EROFS; 1410 } 1411 return 0; 1412 } 1413 1414 static struct inode *f2fs_alloc_inode(struct super_block *sb) 1415 { 1416 struct f2fs_inode_info *fi; 1417 1418 if (time_to_inject(F2FS_SB(sb), FAULT_SLAB_ALLOC)) 1419 return NULL; 1420 1421 fi = alloc_inode_sb(sb, f2fs_inode_cachep, GFP_F2FS_ZERO); 1422 if (!fi) 1423 return NULL; 1424 1425 init_once((void *) fi); 1426 1427 /* Initialize f2fs-specific inode info */ 1428 atomic_set(&fi->dirty_pages, 0); 1429 atomic_set(&fi->i_compr_blocks, 0); 1430 init_f2fs_rwsem(&fi->i_sem); 1431 spin_lock_init(&fi->i_size_lock); 1432 INIT_LIST_HEAD(&fi->dirty_list); 1433 INIT_LIST_HEAD(&fi->gdirty_list); 1434 init_f2fs_rwsem(&fi->i_gc_rwsem[READ]); 1435 init_f2fs_rwsem(&fi->i_gc_rwsem[WRITE]); 1436 init_f2fs_rwsem(&fi->i_xattr_sem); 1437 1438 /* Will be used by directory only */ 1439 fi->i_dir_level = F2FS_SB(sb)->dir_level; 1440 1441 return &fi->vfs_inode; 1442 } 1443 1444 static int f2fs_drop_inode(struct inode *inode) 1445 { 1446 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1447 int ret; 1448 1449 /* 1450 * during filesystem shutdown, if checkpoint is disabled, 1451 * drop useless meta/node dirty pages. 1452 */ 1453 if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED))) { 1454 if (inode->i_ino == F2FS_NODE_INO(sbi) || 1455 inode->i_ino == F2FS_META_INO(sbi)) { 1456 trace_f2fs_drop_inode(inode, 1); 1457 return 1; 1458 } 1459 } 1460 1461 /* 1462 * This is to avoid a deadlock condition like below. 1463 * writeback_single_inode(inode) 1464 * - f2fs_write_data_page 1465 * - f2fs_gc -> iput -> evict 1466 * - inode_wait_for_writeback(inode) 1467 */ 1468 if ((!inode_unhashed(inode) && inode->i_state & I_SYNC)) { 1469 if (!inode->i_nlink && !is_bad_inode(inode)) { 1470 /* to avoid evict_inode call simultaneously */ 1471 atomic_inc(&inode->i_count); 1472 spin_unlock(&inode->i_lock); 1473 1474 /* should remain fi->extent_tree for writepage */ 1475 f2fs_destroy_extent_node(inode); 1476 1477 sb_start_intwrite(inode->i_sb); 1478 f2fs_i_size_write(inode, 0); 1479 1480 f2fs_submit_merged_write_cond(F2FS_I_SB(inode), 1481 inode, NULL, 0, DATA); 1482 truncate_inode_pages_final(inode->i_mapping); 1483 1484 if (F2FS_HAS_BLOCKS(inode)) 1485 f2fs_truncate(inode); 1486 1487 sb_end_intwrite(inode->i_sb); 1488 1489 spin_lock(&inode->i_lock); 1490 atomic_dec(&inode->i_count); 1491 } 1492 trace_f2fs_drop_inode(inode, 0); 1493 return 0; 1494 } 1495 ret = generic_drop_inode(inode); 1496 if (!ret) 1497 ret = fscrypt_drop_inode(inode); 1498 trace_f2fs_drop_inode(inode, ret); 1499 return ret; 1500 } 1501 1502 int f2fs_inode_dirtied(struct inode *inode, bool sync) 1503 { 1504 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1505 int ret = 0; 1506 1507 spin_lock(&sbi->inode_lock[DIRTY_META]); 1508 if (is_inode_flag_set(inode, FI_DIRTY_INODE)) { 1509 ret = 1; 1510 } else { 1511 set_inode_flag(inode, FI_DIRTY_INODE); 1512 stat_inc_dirty_inode(sbi, DIRTY_META); 1513 } 1514 if (sync && list_empty(&F2FS_I(inode)->gdirty_list)) { 1515 list_add_tail(&F2FS_I(inode)->gdirty_list, 1516 &sbi->inode_list[DIRTY_META]); 1517 inc_page_count(sbi, F2FS_DIRTY_IMETA); 1518 } 1519 spin_unlock(&sbi->inode_lock[DIRTY_META]); 1520 return ret; 1521 } 1522 1523 void f2fs_inode_synced(struct inode *inode) 1524 { 1525 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1526 1527 spin_lock(&sbi->inode_lock[DIRTY_META]); 1528 if (!is_inode_flag_set(inode, FI_DIRTY_INODE)) { 1529 spin_unlock(&sbi->inode_lock[DIRTY_META]); 1530 return; 1531 } 1532 if (!list_empty(&F2FS_I(inode)->gdirty_list)) { 1533 list_del_init(&F2FS_I(inode)->gdirty_list); 1534 dec_page_count(sbi, F2FS_DIRTY_IMETA); 1535 } 1536 clear_inode_flag(inode, FI_DIRTY_INODE); 1537 clear_inode_flag(inode, FI_AUTO_RECOVER); 1538 stat_dec_dirty_inode(F2FS_I_SB(inode), DIRTY_META); 1539 spin_unlock(&sbi->inode_lock[DIRTY_META]); 1540 } 1541 1542 /* 1543 * f2fs_dirty_inode() is called from __mark_inode_dirty() 1544 * 1545 * We should call set_dirty_inode to write the dirty inode through write_inode. 1546 */ 1547 static void f2fs_dirty_inode(struct inode *inode, int flags) 1548 { 1549 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1550 1551 if (inode->i_ino == F2FS_NODE_INO(sbi) || 1552 inode->i_ino == F2FS_META_INO(sbi)) 1553 return; 1554 1555 if (is_inode_flag_set(inode, FI_AUTO_RECOVER)) 1556 clear_inode_flag(inode, FI_AUTO_RECOVER); 1557 1558 f2fs_inode_dirtied(inode, false); 1559 } 1560 1561 static void f2fs_free_inode(struct inode *inode) 1562 { 1563 fscrypt_free_inode(inode); 1564 kmem_cache_free(f2fs_inode_cachep, F2FS_I(inode)); 1565 } 1566 1567 static void destroy_percpu_info(struct f2fs_sb_info *sbi) 1568 { 1569 percpu_counter_destroy(&sbi->total_valid_inode_count); 1570 percpu_counter_destroy(&sbi->rf_node_block_count); 1571 percpu_counter_destroy(&sbi->alloc_valid_block_count); 1572 } 1573 1574 static void destroy_device_list(struct f2fs_sb_info *sbi) 1575 { 1576 int i; 1577 1578 for (i = 0; i < sbi->s_ndevs; i++) { 1579 if (i > 0) 1580 bdev_release(FDEV(i).bdev_handle); 1581 #ifdef CONFIG_BLK_DEV_ZONED 1582 kvfree(FDEV(i).blkz_seq); 1583 #endif 1584 } 1585 kvfree(sbi->devs); 1586 } 1587 1588 static void f2fs_put_super(struct super_block *sb) 1589 { 1590 struct f2fs_sb_info *sbi = F2FS_SB(sb); 1591 int i; 1592 int err = 0; 1593 bool done; 1594 1595 /* unregister procfs/sysfs entries in advance to avoid race case */ 1596 f2fs_unregister_sysfs(sbi); 1597 1598 f2fs_quota_off_umount(sb); 1599 1600 /* prevent remaining shrinker jobs */ 1601 mutex_lock(&sbi->umount_mutex); 1602 1603 /* 1604 * flush all issued checkpoints and stop checkpoint issue thread. 1605 * after then, all checkpoints should be done by each process context. 1606 */ 1607 f2fs_stop_ckpt_thread(sbi); 1608 1609 /* 1610 * We don't need to do checkpoint when superblock is clean. 1611 * But, the previous checkpoint was not done by umount, it needs to do 1612 * clean checkpoint again. 1613 */ 1614 if ((is_sbi_flag_set(sbi, SBI_IS_DIRTY) || 1615 !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG))) { 1616 struct cp_control cpc = { 1617 .reason = CP_UMOUNT, 1618 }; 1619 stat_inc_cp_call_count(sbi, TOTAL_CALL); 1620 err = f2fs_write_checkpoint(sbi, &cpc); 1621 } 1622 1623 /* be sure to wait for any on-going discard commands */ 1624 done = f2fs_issue_discard_timeout(sbi); 1625 if (f2fs_realtime_discard_enable(sbi) && !sbi->discard_blks && done) { 1626 struct cp_control cpc = { 1627 .reason = CP_UMOUNT | CP_TRIMMED, 1628 }; 1629 stat_inc_cp_call_count(sbi, TOTAL_CALL); 1630 err = f2fs_write_checkpoint(sbi, &cpc); 1631 } 1632 1633 /* 1634 * normally superblock is clean, so we need to release this. 1635 * In addition, EIO will skip do checkpoint, we need this as well. 1636 */ 1637 f2fs_release_ino_entry(sbi, true); 1638 1639 f2fs_leave_shrinker(sbi); 1640 mutex_unlock(&sbi->umount_mutex); 1641 1642 /* our cp_error case, we can wait for any writeback page */ 1643 f2fs_flush_merged_writes(sbi); 1644 1645 f2fs_wait_on_all_pages(sbi, F2FS_WB_CP_DATA); 1646 1647 if (err) { 1648 truncate_inode_pages_final(NODE_MAPPING(sbi)); 1649 truncate_inode_pages_final(META_MAPPING(sbi)); 1650 } 1651 1652 for (i = 0; i < NR_COUNT_TYPE; i++) { 1653 if (!get_pages(sbi, i)) 1654 continue; 1655 f2fs_err(sbi, "detect filesystem reference count leak during " 1656 "umount, type: %d, count: %lld", i, get_pages(sbi, i)); 1657 f2fs_bug_on(sbi, 1); 1658 } 1659 1660 f2fs_bug_on(sbi, sbi->fsync_node_num); 1661 1662 f2fs_destroy_compress_inode(sbi); 1663 1664 iput(sbi->node_inode); 1665 sbi->node_inode = NULL; 1666 1667 iput(sbi->meta_inode); 1668 sbi->meta_inode = NULL; 1669 1670 /* 1671 * iput() can update stat information, if f2fs_write_checkpoint() 1672 * above failed with error. 1673 */ 1674 f2fs_destroy_stats(sbi); 1675 1676 /* destroy f2fs internal modules */ 1677 f2fs_destroy_node_manager(sbi); 1678 f2fs_destroy_segment_manager(sbi); 1679 1680 /* flush s_error_work before sbi destroy */ 1681 flush_work(&sbi->s_error_work); 1682 1683 f2fs_destroy_post_read_wq(sbi); 1684 1685 kvfree(sbi->ckpt); 1686 1687 sb->s_fs_info = NULL; 1688 if (sbi->s_chksum_driver) 1689 crypto_free_shash(sbi->s_chksum_driver); 1690 kfree(sbi->raw_super); 1691 1692 destroy_device_list(sbi); 1693 f2fs_destroy_page_array_cache(sbi); 1694 f2fs_destroy_xattr_caches(sbi); 1695 mempool_destroy(sbi->write_io_dummy); 1696 #ifdef CONFIG_QUOTA 1697 for (i = 0; i < MAXQUOTAS; i++) 1698 kfree(F2FS_OPTION(sbi).s_qf_names[i]); 1699 #endif 1700 fscrypt_free_dummy_policy(&F2FS_OPTION(sbi).dummy_enc_policy); 1701 destroy_percpu_info(sbi); 1702 f2fs_destroy_iostat(sbi); 1703 for (i = 0; i < NR_PAGE_TYPE; i++) 1704 kvfree(sbi->write_io[i]); 1705 #if IS_ENABLED(CONFIG_UNICODE) 1706 utf8_unload(sb->s_encoding); 1707 #endif 1708 kfree(sbi); 1709 } 1710 1711 int f2fs_sync_fs(struct super_block *sb, int sync) 1712 { 1713 struct f2fs_sb_info *sbi = F2FS_SB(sb); 1714 int err = 0; 1715 1716 if (unlikely(f2fs_cp_error(sbi))) 1717 return 0; 1718 if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED))) 1719 return 0; 1720 1721 trace_f2fs_sync_fs(sb, sync); 1722 1723 if (unlikely(is_sbi_flag_set(sbi, SBI_POR_DOING))) 1724 return -EAGAIN; 1725 1726 if (sync) { 1727 stat_inc_cp_call_count(sbi, TOTAL_CALL); 1728 err = f2fs_issue_checkpoint(sbi); 1729 } 1730 1731 return err; 1732 } 1733 1734 static int f2fs_freeze(struct super_block *sb) 1735 { 1736 if (f2fs_readonly(sb)) 1737 return 0; 1738 1739 /* IO error happened before */ 1740 if (unlikely(f2fs_cp_error(F2FS_SB(sb)))) 1741 return -EIO; 1742 1743 /* must be clean, since sync_filesystem() was already called */ 1744 if (is_sbi_flag_set(F2FS_SB(sb), SBI_IS_DIRTY)) 1745 return -EINVAL; 1746 1747 /* Let's flush checkpoints and stop the thread. */ 1748 f2fs_flush_ckpt_thread(F2FS_SB(sb)); 1749 1750 /* to avoid deadlock on f2fs_evict_inode->SB_FREEZE_FS */ 1751 set_sbi_flag(F2FS_SB(sb), SBI_IS_FREEZING); 1752 return 0; 1753 } 1754 1755 static int f2fs_unfreeze(struct super_block *sb) 1756 { 1757 clear_sbi_flag(F2FS_SB(sb), SBI_IS_FREEZING); 1758 return 0; 1759 } 1760 1761 #ifdef CONFIG_QUOTA 1762 static int f2fs_statfs_project(struct super_block *sb, 1763 kprojid_t projid, struct kstatfs *buf) 1764 { 1765 struct kqid qid; 1766 struct dquot *dquot; 1767 u64 limit; 1768 u64 curblock; 1769 1770 qid = make_kqid_projid(projid); 1771 dquot = dqget(sb, qid); 1772 if (IS_ERR(dquot)) 1773 return PTR_ERR(dquot); 1774 spin_lock(&dquot->dq_dqb_lock); 1775 1776 limit = min_not_zero(dquot->dq_dqb.dqb_bsoftlimit, 1777 dquot->dq_dqb.dqb_bhardlimit); 1778 if (limit) 1779 limit >>= sb->s_blocksize_bits; 1780 1781 if (limit && buf->f_blocks > limit) { 1782 curblock = (dquot->dq_dqb.dqb_curspace + 1783 dquot->dq_dqb.dqb_rsvspace) >> sb->s_blocksize_bits; 1784 buf->f_blocks = limit; 1785 buf->f_bfree = buf->f_bavail = 1786 (buf->f_blocks > curblock) ? 1787 (buf->f_blocks - curblock) : 0; 1788 } 1789 1790 limit = min_not_zero(dquot->dq_dqb.dqb_isoftlimit, 1791 dquot->dq_dqb.dqb_ihardlimit); 1792 1793 if (limit && buf->f_files > limit) { 1794 buf->f_files = limit; 1795 buf->f_ffree = 1796 (buf->f_files > dquot->dq_dqb.dqb_curinodes) ? 1797 (buf->f_files - dquot->dq_dqb.dqb_curinodes) : 0; 1798 } 1799 1800 spin_unlock(&dquot->dq_dqb_lock); 1801 dqput(dquot); 1802 return 0; 1803 } 1804 #endif 1805 1806 static int f2fs_statfs(struct dentry *dentry, struct kstatfs *buf) 1807 { 1808 struct super_block *sb = dentry->d_sb; 1809 struct f2fs_sb_info *sbi = F2FS_SB(sb); 1810 u64 id = huge_encode_dev(sb->s_bdev->bd_dev); 1811 block_t total_count, user_block_count, start_count; 1812 u64 avail_node_count; 1813 unsigned int total_valid_node_count; 1814 1815 total_count = le64_to_cpu(sbi->raw_super->block_count); 1816 start_count = le32_to_cpu(sbi->raw_super->segment0_blkaddr); 1817 buf->f_type = F2FS_SUPER_MAGIC; 1818 buf->f_bsize = sbi->blocksize; 1819 1820 buf->f_blocks = total_count - start_count; 1821 1822 spin_lock(&sbi->stat_lock); 1823 1824 user_block_count = sbi->user_block_count; 1825 total_valid_node_count = valid_node_count(sbi); 1826 avail_node_count = sbi->total_node_count - F2FS_RESERVED_NODE_NUM; 1827 buf->f_bfree = user_block_count - valid_user_blocks(sbi) - 1828 sbi->current_reserved_blocks; 1829 1830 if (unlikely(buf->f_bfree <= sbi->unusable_block_count)) 1831 buf->f_bfree = 0; 1832 else 1833 buf->f_bfree -= sbi->unusable_block_count; 1834 spin_unlock(&sbi->stat_lock); 1835 1836 if (buf->f_bfree > F2FS_OPTION(sbi).root_reserved_blocks) 1837 buf->f_bavail = buf->f_bfree - 1838 F2FS_OPTION(sbi).root_reserved_blocks; 1839 else 1840 buf->f_bavail = 0; 1841 1842 if (avail_node_count > user_block_count) { 1843 buf->f_files = user_block_count; 1844 buf->f_ffree = buf->f_bavail; 1845 } else { 1846 buf->f_files = avail_node_count; 1847 buf->f_ffree = min(avail_node_count - total_valid_node_count, 1848 buf->f_bavail); 1849 } 1850 1851 buf->f_namelen = F2FS_NAME_LEN; 1852 buf->f_fsid = u64_to_fsid(id); 1853 1854 #ifdef CONFIG_QUOTA 1855 if (is_inode_flag_set(dentry->d_inode, FI_PROJ_INHERIT) && 1856 sb_has_quota_limits_enabled(sb, PRJQUOTA)) { 1857 f2fs_statfs_project(sb, F2FS_I(dentry->d_inode)->i_projid, buf); 1858 } 1859 #endif 1860 return 0; 1861 } 1862 1863 static inline void f2fs_show_quota_options(struct seq_file *seq, 1864 struct super_block *sb) 1865 { 1866 #ifdef CONFIG_QUOTA 1867 struct f2fs_sb_info *sbi = F2FS_SB(sb); 1868 1869 if (F2FS_OPTION(sbi).s_jquota_fmt) { 1870 char *fmtname = ""; 1871 1872 switch (F2FS_OPTION(sbi).s_jquota_fmt) { 1873 case QFMT_VFS_OLD: 1874 fmtname = "vfsold"; 1875 break; 1876 case QFMT_VFS_V0: 1877 fmtname = "vfsv0"; 1878 break; 1879 case QFMT_VFS_V1: 1880 fmtname = "vfsv1"; 1881 break; 1882 } 1883 seq_printf(seq, ",jqfmt=%s", fmtname); 1884 } 1885 1886 if (F2FS_OPTION(sbi).s_qf_names[USRQUOTA]) 1887 seq_show_option(seq, "usrjquota", 1888 F2FS_OPTION(sbi).s_qf_names[USRQUOTA]); 1889 1890 if (F2FS_OPTION(sbi).s_qf_names[GRPQUOTA]) 1891 seq_show_option(seq, "grpjquota", 1892 F2FS_OPTION(sbi).s_qf_names[GRPQUOTA]); 1893 1894 if (F2FS_OPTION(sbi).s_qf_names[PRJQUOTA]) 1895 seq_show_option(seq, "prjjquota", 1896 F2FS_OPTION(sbi).s_qf_names[PRJQUOTA]); 1897 #endif 1898 } 1899 1900 #ifdef CONFIG_F2FS_FS_COMPRESSION 1901 static inline void f2fs_show_compress_options(struct seq_file *seq, 1902 struct super_block *sb) 1903 { 1904 struct f2fs_sb_info *sbi = F2FS_SB(sb); 1905 char *algtype = ""; 1906 int i; 1907 1908 if (!f2fs_sb_has_compression(sbi)) 1909 return; 1910 1911 switch (F2FS_OPTION(sbi).compress_algorithm) { 1912 case COMPRESS_LZO: 1913 algtype = "lzo"; 1914 break; 1915 case COMPRESS_LZ4: 1916 algtype = "lz4"; 1917 break; 1918 case COMPRESS_ZSTD: 1919 algtype = "zstd"; 1920 break; 1921 case COMPRESS_LZORLE: 1922 algtype = "lzo-rle"; 1923 break; 1924 } 1925 seq_printf(seq, ",compress_algorithm=%s", algtype); 1926 1927 if (F2FS_OPTION(sbi).compress_level) 1928 seq_printf(seq, ":%d", F2FS_OPTION(sbi).compress_level); 1929 1930 seq_printf(seq, ",compress_log_size=%u", 1931 F2FS_OPTION(sbi).compress_log_size); 1932 1933 for (i = 0; i < F2FS_OPTION(sbi).compress_ext_cnt; i++) { 1934 seq_printf(seq, ",compress_extension=%s", 1935 F2FS_OPTION(sbi).extensions[i]); 1936 } 1937 1938 for (i = 0; i < F2FS_OPTION(sbi).nocompress_ext_cnt; i++) { 1939 seq_printf(seq, ",nocompress_extension=%s", 1940 F2FS_OPTION(sbi).noextensions[i]); 1941 } 1942 1943 if (F2FS_OPTION(sbi).compress_chksum) 1944 seq_puts(seq, ",compress_chksum"); 1945 1946 if (F2FS_OPTION(sbi).compress_mode == COMPR_MODE_FS) 1947 seq_printf(seq, ",compress_mode=%s", "fs"); 1948 else if (F2FS_OPTION(sbi).compress_mode == COMPR_MODE_USER) 1949 seq_printf(seq, ",compress_mode=%s", "user"); 1950 1951 if (test_opt(sbi, COMPRESS_CACHE)) 1952 seq_puts(seq, ",compress_cache"); 1953 } 1954 #endif 1955 1956 static int f2fs_show_options(struct seq_file *seq, struct dentry *root) 1957 { 1958 struct f2fs_sb_info *sbi = F2FS_SB(root->d_sb); 1959 1960 if (F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_SYNC) 1961 seq_printf(seq, ",background_gc=%s", "sync"); 1962 else if (F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_ON) 1963 seq_printf(seq, ",background_gc=%s", "on"); 1964 else if (F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_OFF) 1965 seq_printf(seq, ",background_gc=%s", "off"); 1966 1967 if (test_opt(sbi, GC_MERGE)) 1968 seq_puts(seq, ",gc_merge"); 1969 else 1970 seq_puts(seq, ",nogc_merge"); 1971 1972 if (test_opt(sbi, DISABLE_ROLL_FORWARD)) 1973 seq_puts(seq, ",disable_roll_forward"); 1974 if (test_opt(sbi, NORECOVERY)) 1975 seq_puts(seq, ",norecovery"); 1976 if (test_opt(sbi, DISCARD)) { 1977 seq_puts(seq, ",discard"); 1978 if (F2FS_OPTION(sbi).discard_unit == DISCARD_UNIT_BLOCK) 1979 seq_printf(seq, ",discard_unit=%s", "block"); 1980 else if (F2FS_OPTION(sbi).discard_unit == DISCARD_UNIT_SEGMENT) 1981 seq_printf(seq, ",discard_unit=%s", "segment"); 1982 else if (F2FS_OPTION(sbi).discard_unit == DISCARD_UNIT_SECTION) 1983 seq_printf(seq, ",discard_unit=%s", "section"); 1984 } else { 1985 seq_puts(seq, ",nodiscard"); 1986 } 1987 if (test_opt(sbi, NOHEAP)) 1988 seq_puts(seq, ",no_heap"); 1989 else 1990 seq_puts(seq, ",heap"); 1991 #ifdef CONFIG_F2FS_FS_XATTR 1992 if (test_opt(sbi, XATTR_USER)) 1993 seq_puts(seq, ",user_xattr"); 1994 else 1995 seq_puts(seq, ",nouser_xattr"); 1996 if (test_opt(sbi, INLINE_XATTR)) 1997 seq_puts(seq, ",inline_xattr"); 1998 else 1999 seq_puts(seq, ",noinline_xattr"); 2000 if (test_opt(sbi, INLINE_XATTR_SIZE)) 2001 seq_printf(seq, ",inline_xattr_size=%u", 2002 F2FS_OPTION(sbi).inline_xattr_size); 2003 #endif 2004 #ifdef CONFIG_F2FS_FS_POSIX_ACL 2005 if (test_opt(sbi, POSIX_ACL)) 2006 seq_puts(seq, ",acl"); 2007 else 2008 seq_puts(seq, ",noacl"); 2009 #endif 2010 if (test_opt(sbi, DISABLE_EXT_IDENTIFY)) 2011 seq_puts(seq, ",disable_ext_identify"); 2012 if (test_opt(sbi, INLINE_DATA)) 2013 seq_puts(seq, ",inline_data"); 2014 else 2015 seq_puts(seq, ",noinline_data"); 2016 if (test_opt(sbi, INLINE_DENTRY)) 2017 seq_puts(seq, ",inline_dentry"); 2018 else 2019 seq_puts(seq, ",noinline_dentry"); 2020 if (test_opt(sbi, FLUSH_MERGE)) 2021 seq_puts(seq, ",flush_merge"); 2022 else 2023 seq_puts(seq, ",noflush_merge"); 2024 if (test_opt(sbi, NOBARRIER)) 2025 seq_puts(seq, ",nobarrier"); 2026 else 2027 seq_puts(seq, ",barrier"); 2028 if (test_opt(sbi, FASTBOOT)) 2029 seq_puts(seq, ",fastboot"); 2030 if (test_opt(sbi, READ_EXTENT_CACHE)) 2031 seq_puts(seq, ",extent_cache"); 2032 else 2033 seq_puts(seq, ",noextent_cache"); 2034 if (test_opt(sbi, AGE_EXTENT_CACHE)) 2035 seq_puts(seq, ",age_extent_cache"); 2036 if (test_opt(sbi, DATA_FLUSH)) 2037 seq_puts(seq, ",data_flush"); 2038 2039 seq_puts(seq, ",mode="); 2040 if (F2FS_OPTION(sbi).fs_mode == FS_MODE_ADAPTIVE) 2041 seq_puts(seq, "adaptive"); 2042 else if (F2FS_OPTION(sbi).fs_mode == FS_MODE_LFS) 2043 seq_puts(seq, "lfs"); 2044 else if (F2FS_OPTION(sbi).fs_mode == FS_MODE_FRAGMENT_SEG) 2045 seq_puts(seq, "fragment:segment"); 2046 else if (F2FS_OPTION(sbi).fs_mode == FS_MODE_FRAGMENT_BLK) 2047 seq_puts(seq, "fragment:block"); 2048 seq_printf(seq, ",active_logs=%u", F2FS_OPTION(sbi).active_logs); 2049 if (test_opt(sbi, RESERVE_ROOT)) 2050 seq_printf(seq, ",reserve_root=%u,resuid=%u,resgid=%u", 2051 F2FS_OPTION(sbi).root_reserved_blocks, 2052 from_kuid_munged(&init_user_ns, 2053 F2FS_OPTION(sbi).s_resuid), 2054 from_kgid_munged(&init_user_ns, 2055 F2FS_OPTION(sbi).s_resgid)); 2056 if (F2FS_IO_SIZE_BITS(sbi)) 2057 seq_printf(seq, ",io_bits=%u", 2058 F2FS_OPTION(sbi).write_io_size_bits); 2059 #ifdef CONFIG_F2FS_FAULT_INJECTION 2060 if (test_opt(sbi, FAULT_INJECTION)) { 2061 seq_printf(seq, ",fault_injection=%u", 2062 F2FS_OPTION(sbi).fault_info.inject_rate); 2063 seq_printf(seq, ",fault_type=%u", 2064 F2FS_OPTION(sbi).fault_info.inject_type); 2065 } 2066 #endif 2067 #ifdef CONFIG_QUOTA 2068 if (test_opt(sbi, QUOTA)) 2069 seq_puts(seq, ",quota"); 2070 if (test_opt(sbi, USRQUOTA)) 2071 seq_puts(seq, ",usrquota"); 2072 if (test_opt(sbi, GRPQUOTA)) 2073 seq_puts(seq, ",grpquota"); 2074 if (test_opt(sbi, PRJQUOTA)) 2075 seq_puts(seq, ",prjquota"); 2076 #endif 2077 f2fs_show_quota_options(seq, sbi->sb); 2078 2079 fscrypt_show_test_dummy_encryption(seq, ',', sbi->sb); 2080 2081 if (sbi->sb->s_flags & SB_INLINECRYPT) 2082 seq_puts(seq, ",inlinecrypt"); 2083 2084 if (F2FS_OPTION(sbi).alloc_mode == ALLOC_MODE_DEFAULT) 2085 seq_printf(seq, ",alloc_mode=%s", "default"); 2086 else if (F2FS_OPTION(sbi).alloc_mode == ALLOC_MODE_REUSE) 2087 seq_printf(seq, ",alloc_mode=%s", "reuse"); 2088 2089 if (test_opt(sbi, DISABLE_CHECKPOINT)) 2090 seq_printf(seq, ",checkpoint=disable:%u", 2091 F2FS_OPTION(sbi).unusable_cap); 2092 if (test_opt(sbi, MERGE_CHECKPOINT)) 2093 seq_puts(seq, ",checkpoint_merge"); 2094 else 2095 seq_puts(seq, ",nocheckpoint_merge"); 2096 if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_POSIX) 2097 seq_printf(seq, ",fsync_mode=%s", "posix"); 2098 else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_STRICT) 2099 seq_printf(seq, ",fsync_mode=%s", "strict"); 2100 else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_NOBARRIER) 2101 seq_printf(seq, ",fsync_mode=%s", "nobarrier"); 2102 2103 #ifdef CONFIG_F2FS_FS_COMPRESSION 2104 f2fs_show_compress_options(seq, sbi->sb); 2105 #endif 2106 2107 if (test_opt(sbi, ATGC)) 2108 seq_puts(seq, ",atgc"); 2109 2110 if (F2FS_OPTION(sbi).memory_mode == MEMORY_MODE_NORMAL) 2111 seq_printf(seq, ",memory=%s", "normal"); 2112 else if (F2FS_OPTION(sbi).memory_mode == MEMORY_MODE_LOW) 2113 seq_printf(seq, ",memory=%s", "low"); 2114 2115 if (F2FS_OPTION(sbi).errors == MOUNT_ERRORS_READONLY) 2116 seq_printf(seq, ",errors=%s", "remount-ro"); 2117 else if (F2FS_OPTION(sbi).errors == MOUNT_ERRORS_CONTINUE) 2118 seq_printf(seq, ",errors=%s", "continue"); 2119 else if (F2FS_OPTION(sbi).errors == MOUNT_ERRORS_PANIC) 2120 seq_printf(seq, ",errors=%s", "panic"); 2121 2122 return 0; 2123 } 2124 2125 static void default_options(struct f2fs_sb_info *sbi, bool remount) 2126 { 2127 /* init some FS parameters */ 2128 if (!remount) { 2129 set_opt(sbi, READ_EXTENT_CACHE); 2130 clear_opt(sbi, DISABLE_CHECKPOINT); 2131 2132 if (f2fs_hw_support_discard(sbi) || f2fs_hw_should_discard(sbi)) 2133 set_opt(sbi, DISCARD); 2134 2135 if (f2fs_sb_has_blkzoned(sbi)) 2136 F2FS_OPTION(sbi).discard_unit = DISCARD_UNIT_SECTION; 2137 else 2138 F2FS_OPTION(sbi).discard_unit = DISCARD_UNIT_BLOCK; 2139 } 2140 2141 if (f2fs_sb_has_readonly(sbi)) 2142 F2FS_OPTION(sbi).active_logs = NR_CURSEG_RO_TYPE; 2143 else 2144 F2FS_OPTION(sbi).active_logs = NR_CURSEG_PERSIST_TYPE; 2145 2146 F2FS_OPTION(sbi).inline_xattr_size = DEFAULT_INLINE_XATTR_ADDRS; 2147 if (le32_to_cpu(F2FS_RAW_SUPER(sbi)->segment_count_main) <= 2148 SMALL_VOLUME_SEGMENTS) 2149 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_REUSE; 2150 else 2151 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_DEFAULT; 2152 F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_POSIX; 2153 F2FS_OPTION(sbi).s_resuid = make_kuid(&init_user_ns, F2FS_DEF_RESUID); 2154 F2FS_OPTION(sbi).s_resgid = make_kgid(&init_user_ns, F2FS_DEF_RESGID); 2155 if (f2fs_sb_has_compression(sbi)) { 2156 F2FS_OPTION(sbi).compress_algorithm = COMPRESS_LZ4; 2157 F2FS_OPTION(sbi).compress_log_size = MIN_COMPRESS_LOG_SIZE; 2158 F2FS_OPTION(sbi).compress_ext_cnt = 0; 2159 F2FS_OPTION(sbi).compress_mode = COMPR_MODE_FS; 2160 } 2161 F2FS_OPTION(sbi).bggc_mode = BGGC_MODE_ON; 2162 F2FS_OPTION(sbi).memory_mode = MEMORY_MODE_NORMAL; 2163 F2FS_OPTION(sbi).errors = MOUNT_ERRORS_CONTINUE; 2164 2165 sbi->sb->s_flags &= ~SB_INLINECRYPT; 2166 2167 set_opt(sbi, INLINE_XATTR); 2168 set_opt(sbi, INLINE_DATA); 2169 set_opt(sbi, INLINE_DENTRY); 2170 set_opt(sbi, NOHEAP); 2171 set_opt(sbi, MERGE_CHECKPOINT); 2172 F2FS_OPTION(sbi).unusable_cap = 0; 2173 sbi->sb->s_flags |= SB_LAZYTIME; 2174 if (!f2fs_is_readonly(sbi)) 2175 set_opt(sbi, FLUSH_MERGE); 2176 if (f2fs_sb_has_blkzoned(sbi)) 2177 F2FS_OPTION(sbi).fs_mode = FS_MODE_LFS; 2178 else 2179 F2FS_OPTION(sbi).fs_mode = FS_MODE_ADAPTIVE; 2180 2181 #ifdef CONFIG_F2FS_FS_XATTR 2182 set_opt(sbi, XATTR_USER); 2183 #endif 2184 #ifdef CONFIG_F2FS_FS_POSIX_ACL 2185 set_opt(sbi, POSIX_ACL); 2186 #endif 2187 2188 f2fs_build_fault_attr(sbi, 0, 0); 2189 } 2190 2191 #ifdef CONFIG_QUOTA 2192 static int f2fs_enable_quotas(struct super_block *sb); 2193 #endif 2194 2195 static int f2fs_disable_checkpoint(struct f2fs_sb_info *sbi) 2196 { 2197 unsigned int s_flags = sbi->sb->s_flags; 2198 struct cp_control cpc; 2199 unsigned int gc_mode = sbi->gc_mode; 2200 int err = 0; 2201 int ret; 2202 block_t unusable; 2203 2204 if (s_flags & SB_RDONLY) { 2205 f2fs_err(sbi, "checkpoint=disable on readonly fs"); 2206 return -EINVAL; 2207 } 2208 sbi->sb->s_flags |= SB_ACTIVE; 2209 2210 /* check if we need more GC first */ 2211 unusable = f2fs_get_unusable_blocks(sbi); 2212 if (!f2fs_disable_cp_again(sbi, unusable)) 2213 goto skip_gc; 2214 2215 f2fs_update_time(sbi, DISABLE_TIME); 2216 2217 sbi->gc_mode = GC_URGENT_HIGH; 2218 2219 while (!f2fs_time_over(sbi, DISABLE_TIME)) { 2220 struct f2fs_gc_control gc_control = { 2221 .victim_segno = NULL_SEGNO, 2222 .init_gc_type = FG_GC, 2223 .should_migrate_blocks = false, 2224 .err_gc_skipped = true, 2225 .nr_free_secs = 1 }; 2226 2227 f2fs_down_write(&sbi->gc_lock); 2228 stat_inc_gc_call_count(sbi, FOREGROUND); 2229 err = f2fs_gc(sbi, &gc_control); 2230 if (err == -ENODATA) { 2231 err = 0; 2232 break; 2233 } 2234 if (err && err != -EAGAIN) 2235 break; 2236 } 2237 2238 ret = sync_filesystem(sbi->sb); 2239 if (ret || err) { 2240 err = ret ? ret : err; 2241 goto restore_flag; 2242 } 2243 2244 unusable = f2fs_get_unusable_blocks(sbi); 2245 if (f2fs_disable_cp_again(sbi, unusable)) { 2246 err = -EAGAIN; 2247 goto restore_flag; 2248 } 2249 2250 skip_gc: 2251 f2fs_down_write(&sbi->gc_lock); 2252 cpc.reason = CP_PAUSE; 2253 set_sbi_flag(sbi, SBI_CP_DISABLED); 2254 stat_inc_cp_call_count(sbi, TOTAL_CALL); 2255 err = f2fs_write_checkpoint(sbi, &cpc); 2256 if (err) 2257 goto out_unlock; 2258 2259 spin_lock(&sbi->stat_lock); 2260 sbi->unusable_block_count = unusable; 2261 spin_unlock(&sbi->stat_lock); 2262 2263 out_unlock: 2264 f2fs_up_write(&sbi->gc_lock); 2265 restore_flag: 2266 sbi->gc_mode = gc_mode; 2267 sbi->sb->s_flags = s_flags; /* Restore SB_RDONLY status */ 2268 return err; 2269 } 2270 2271 static void f2fs_enable_checkpoint(struct f2fs_sb_info *sbi) 2272 { 2273 int retry = DEFAULT_RETRY_IO_COUNT; 2274 2275 /* we should flush all the data to keep data consistency */ 2276 do { 2277 sync_inodes_sb(sbi->sb); 2278 f2fs_io_schedule_timeout(DEFAULT_IO_TIMEOUT); 2279 } while (get_pages(sbi, F2FS_DIRTY_DATA) && retry--); 2280 2281 if (unlikely(retry < 0)) 2282 f2fs_warn(sbi, "checkpoint=enable has some unwritten data."); 2283 2284 f2fs_down_write(&sbi->gc_lock); 2285 f2fs_dirty_to_prefree(sbi); 2286 2287 clear_sbi_flag(sbi, SBI_CP_DISABLED); 2288 set_sbi_flag(sbi, SBI_IS_DIRTY); 2289 f2fs_up_write(&sbi->gc_lock); 2290 2291 f2fs_sync_fs(sbi->sb, 1); 2292 2293 /* Let's ensure there's no pending checkpoint anymore */ 2294 f2fs_flush_ckpt_thread(sbi); 2295 } 2296 2297 static int f2fs_remount(struct super_block *sb, int *flags, char *data) 2298 { 2299 struct f2fs_sb_info *sbi = F2FS_SB(sb); 2300 struct f2fs_mount_info org_mount_opt; 2301 unsigned long old_sb_flags; 2302 int err; 2303 bool need_restart_gc = false, need_stop_gc = false; 2304 bool need_restart_ckpt = false, need_stop_ckpt = false; 2305 bool need_restart_flush = false, need_stop_flush = false; 2306 bool need_restart_discard = false, need_stop_discard = false; 2307 bool no_read_extent_cache = !test_opt(sbi, READ_EXTENT_CACHE); 2308 bool no_age_extent_cache = !test_opt(sbi, AGE_EXTENT_CACHE); 2309 bool enable_checkpoint = !test_opt(sbi, DISABLE_CHECKPOINT); 2310 bool no_io_align = !F2FS_IO_ALIGNED(sbi); 2311 bool no_atgc = !test_opt(sbi, ATGC); 2312 bool no_discard = !test_opt(sbi, DISCARD); 2313 bool no_compress_cache = !test_opt(sbi, COMPRESS_CACHE); 2314 bool block_unit_discard = f2fs_block_unit_discard(sbi); 2315 #ifdef CONFIG_QUOTA 2316 int i, j; 2317 #endif 2318 2319 /* 2320 * Save the old mount options in case we 2321 * need to restore them. 2322 */ 2323 org_mount_opt = sbi->mount_opt; 2324 old_sb_flags = sb->s_flags; 2325 2326 #ifdef CONFIG_QUOTA 2327 org_mount_opt.s_jquota_fmt = F2FS_OPTION(sbi).s_jquota_fmt; 2328 for (i = 0; i < MAXQUOTAS; i++) { 2329 if (F2FS_OPTION(sbi).s_qf_names[i]) { 2330 org_mount_opt.s_qf_names[i] = 2331 kstrdup(F2FS_OPTION(sbi).s_qf_names[i], 2332 GFP_KERNEL); 2333 if (!org_mount_opt.s_qf_names[i]) { 2334 for (j = 0; j < i; j++) 2335 kfree(org_mount_opt.s_qf_names[j]); 2336 return -ENOMEM; 2337 } 2338 } else { 2339 org_mount_opt.s_qf_names[i] = NULL; 2340 } 2341 } 2342 #endif 2343 2344 /* recover superblocks we couldn't write due to previous RO mount */ 2345 if (!(*flags & SB_RDONLY) && is_sbi_flag_set(sbi, SBI_NEED_SB_WRITE)) { 2346 err = f2fs_commit_super(sbi, false); 2347 f2fs_info(sbi, "Try to recover all the superblocks, ret: %d", 2348 err); 2349 if (!err) 2350 clear_sbi_flag(sbi, SBI_NEED_SB_WRITE); 2351 } 2352 2353 default_options(sbi, true); 2354 2355 /* parse mount options */ 2356 err = parse_options(sb, data, true); 2357 if (err) 2358 goto restore_opts; 2359 2360 /* flush outstanding errors before changing fs state */ 2361 flush_work(&sbi->s_error_work); 2362 2363 /* 2364 * Previous and new state of filesystem is RO, 2365 * so skip checking GC and FLUSH_MERGE conditions. 2366 */ 2367 if (f2fs_readonly(sb) && (*flags & SB_RDONLY)) 2368 goto skip; 2369 2370 if (f2fs_dev_is_readonly(sbi) && !(*flags & SB_RDONLY)) { 2371 err = -EROFS; 2372 goto restore_opts; 2373 } 2374 2375 #ifdef CONFIG_QUOTA 2376 if (!f2fs_readonly(sb) && (*flags & SB_RDONLY)) { 2377 err = dquot_suspend(sb, -1); 2378 if (err < 0) 2379 goto restore_opts; 2380 } else if (f2fs_readonly(sb) && !(*flags & SB_RDONLY)) { 2381 /* dquot_resume needs RW */ 2382 sb->s_flags &= ~SB_RDONLY; 2383 if (sb_any_quota_suspended(sb)) { 2384 dquot_resume(sb, -1); 2385 } else if (f2fs_sb_has_quota_ino(sbi)) { 2386 err = f2fs_enable_quotas(sb); 2387 if (err) 2388 goto restore_opts; 2389 } 2390 } 2391 #endif 2392 if (f2fs_lfs_mode(sbi) && !IS_F2FS_IPU_DISABLE(sbi)) { 2393 err = -EINVAL; 2394 f2fs_warn(sbi, "LFS is not compatible with IPU"); 2395 goto restore_opts; 2396 } 2397 2398 /* disallow enable atgc dynamically */ 2399 if (no_atgc == !!test_opt(sbi, ATGC)) { 2400 err = -EINVAL; 2401 f2fs_warn(sbi, "switch atgc option is not allowed"); 2402 goto restore_opts; 2403 } 2404 2405 /* disallow enable/disable extent_cache dynamically */ 2406 if (no_read_extent_cache == !!test_opt(sbi, READ_EXTENT_CACHE)) { 2407 err = -EINVAL; 2408 f2fs_warn(sbi, "switch extent_cache option is not allowed"); 2409 goto restore_opts; 2410 } 2411 /* disallow enable/disable age extent_cache dynamically */ 2412 if (no_age_extent_cache == !!test_opt(sbi, AGE_EXTENT_CACHE)) { 2413 err = -EINVAL; 2414 f2fs_warn(sbi, "switch age_extent_cache option is not allowed"); 2415 goto restore_opts; 2416 } 2417 2418 if (no_io_align == !!F2FS_IO_ALIGNED(sbi)) { 2419 err = -EINVAL; 2420 f2fs_warn(sbi, "switch io_bits option is not allowed"); 2421 goto restore_opts; 2422 } 2423 2424 if (no_compress_cache == !!test_opt(sbi, COMPRESS_CACHE)) { 2425 err = -EINVAL; 2426 f2fs_warn(sbi, "switch compress_cache option is not allowed"); 2427 goto restore_opts; 2428 } 2429 2430 if (block_unit_discard != f2fs_block_unit_discard(sbi)) { 2431 err = -EINVAL; 2432 f2fs_warn(sbi, "switch discard_unit option is not allowed"); 2433 goto restore_opts; 2434 } 2435 2436 if ((*flags & SB_RDONLY) && test_opt(sbi, DISABLE_CHECKPOINT)) { 2437 err = -EINVAL; 2438 f2fs_warn(sbi, "disabling checkpoint not compatible with read-only"); 2439 goto restore_opts; 2440 } 2441 2442 /* 2443 * We stop the GC thread if FS is mounted as RO 2444 * or if background_gc = off is passed in mount 2445 * option. Also sync the filesystem. 2446 */ 2447 if ((*flags & SB_RDONLY) || 2448 (F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_OFF && 2449 !test_opt(sbi, GC_MERGE))) { 2450 if (sbi->gc_thread) { 2451 f2fs_stop_gc_thread(sbi); 2452 need_restart_gc = true; 2453 } 2454 } else if (!sbi->gc_thread) { 2455 err = f2fs_start_gc_thread(sbi); 2456 if (err) 2457 goto restore_opts; 2458 need_stop_gc = true; 2459 } 2460 2461 if (*flags & SB_RDONLY) { 2462 sync_inodes_sb(sb); 2463 2464 set_sbi_flag(sbi, SBI_IS_DIRTY); 2465 set_sbi_flag(sbi, SBI_IS_CLOSE); 2466 f2fs_sync_fs(sb, 1); 2467 clear_sbi_flag(sbi, SBI_IS_CLOSE); 2468 } 2469 2470 if ((*flags & SB_RDONLY) || test_opt(sbi, DISABLE_CHECKPOINT) || 2471 !test_opt(sbi, MERGE_CHECKPOINT)) { 2472 f2fs_stop_ckpt_thread(sbi); 2473 need_restart_ckpt = true; 2474 } else { 2475 /* Flush if the prevous checkpoint, if exists. */ 2476 f2fs_flush_ckpt_thread(sbi); 2477 2478 err = f2fs_start_ckpt_thread(sbi); 2479 if (err) { 2480 f2fs_err(sbi, 2481 "Failed to start F2FS issue_checkpoint_thread (%d)", 2482 err); 2483 goto restore_gc; 2484 } 2485 need_stop_ckpt = true; 2486 } 2487 2488 /* 2489 * We stop issue flush thread if FS is mounted as RO 2490 * or if flush_merge is not passed in mount option. 2491 */ 2492 if ((*flags & SB_RDONLY) || !test_opt(sbi, FLUSH_MERGE)) { 2493 clear_opt(sbi, FLUSH_MERGE); 2494 f2fs_destroy_flush_cmd_control(sbi, false); 2495 need_restart_flush = true; 2496 } else { 2497 err = f2fs_create_flush_cmd_control(sbi); 2498 if (err) 2499 goto restore_ckpt; 2500 need_stop_flush = true; 2501 } 2502 2503 if (no_discard == !!test_opt(sbi, DISCARD)) { 2504 if (test_opt(sbi, DISCARD)) { 2505 err = f2fs_start_discard_thread(sbi); 2506 if (err) 2507 goto restore_flush; 2508 need_stop_discard = true; 2509 } else { 2510 f2fs_stop_discard_thread(sbi); 2511 f2fs_issue_discard_timeout(sbi); 2512 need_restart_discard = true; 2513 } 2514 } 2515 2516 if (enable_checkpoint == !!test_opt(sbi, DISABLE_CHECKPOINT)) { 2517 if (test_opt(sbi, DISABLE_CHECKPOINT)) { 2518 err = f2fs_disable_checkpoint(sbi); 2519 if (err) 2520 goto restore_discard; 2521 } else { 2522 f2fs_enable_checkpoint(sbi); 2523 } 2524 } 2525 2526 skip: 2527 #ifdef CONFIG_QUOTA 2528 /* Release old quota file names */ 2529 for (i = 0; i < MAXQUOTAS; i++) 2530 kfree(org_mount_opt.s_qf_names[i]); 2531 #endif 2532 /* Update the POSIXACL Flag */ 2533 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) | 2534 (test_opt(sbi, POSIX_ACL) ? SB_POSIXACL : 0); 2535 2536 limit_reserve_root(sbi); 2537 adjust_unusable_cap_perc(sbi); 2538 *flags = (*flags & ~SB_LAZYTIME) | (sb->s_flags & SB_LAZYTIME); 2539 return 0; 2540 restore_discard: 2541 if (need_restart_discard) { 2542 if (f2fs_start_discard_thread(sbi)) 2543 f2fs_warn(sbi, "discard has been stopped"); 2544 } else if (need_stop_discard) { 2545 f2fs_stop_discard_thread(sbi); 2546 } 2547 restore_flush: 2548 if (need_restart_flush) { 2549 if (f2fs_create_flush_cmd_control(sbi)) 2550 f2fs_warn(sbi, "background flush thread has stopped"); 2551 } else if (need_stop_flush) { 2552 clear_opt(sbi, FLUSH_MERGE); 2553 f2fs_destroy_flush_cmd_control(sbi, false); 2554 } 2555 restore_ckpt: 2556 if (need_restart_ckpt) { 2557 if (f2fs_start_ckpt_thread(sbi)) 2558 f2fs_warn(sbi, "background ckpt thread has stopped"); 2559 } else if (need_stop_ckpt) { 2560 f2fs_stop_ckpt_thread(sbi); 2561 } 2562 restore_gc: 2563 if (need_restart_gc) { 2564 if (f2fs_start_gc_thread(sbi)) 2565 f2fs_warn(sbi, "background gc thread has stopped"); 2566 } else if (need_stop_gc) { 2567 f2fs_stop_gc_thread(sbi); 2568 } 2569 restore_opts: 2570 #ifdef CONFIG_QUOTA 2571 F2FS_OPTION(sbi).s_jquota_fmt = org_mount_opt.s_jquota_fmt; 2572 for (i = 0; i < MAXQUOTAS; i++) { 2573 kfree(F2FS_OPTION(sbi).s_qf_names[i]); 2574 F2FS_OPTION(sbi).s_qf_names[i] = org_mount_opt.s_qf_names[i]; 2575 } 2576 #endif 2577 sbi->mount_opt = org_mount_opt; 2578 sb->s_flags = old_sb_flags; 2579 return err; 2580 } 2581 2582 #ifdef CONFIG_QUOTA 2583 static bool f2fs_need_recovery(struct f2fs_sb_info *sbi) 2584 { 2585 /* need to recovery orphan */ 2586 if (is_set_ckpt_flags(sbi, CP_ORPHAN_PRESENT_FLAG)) 2587 return true; 2588 /* need to recovery data */ 2589 if (test_opt(sbi, DISABLE_ROLL_FORWARD)) 2590 return false; 2591 if (test_opt(sbi, NORECOVERY)) 2592 return false; 2593 return !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG); 2594 } 2595 2596 static bool f2fs_recover_quota_begin(struct f2fs_sb_info *sbi) 2597 { 2598 bool readonly = f2fs_readonly(sbi->sb); 2599 2600 if (!f2fs_need_recovery(sbi)) 2601 return false; 2602 2603 /* it doesn't need to check f2fs_sb_has_readonly() */ 2604 if (f2fs_hw_is_readonly(sbi)) 2605 return false; 2606 2607 if (readonly) { 2608 sbi->sb->s_flags &= ~SB_RDONLY; 2609 set_sbi_flag(sbi, SBI_IS_WRITABLE); 2610 } 2611 2612 /* 2613 * Turn on quotas which were not enabled for read-only mounts if 2614 * filesystem has quota feature, so that they are updated correctly. 2615 */ 2616 return f2fs_enable_quota_files(sbi, readonly); 2617 } 2618 2619 static void f2fs_recover_quota_end(struct f2fs_sb_info *sbi, 2620 bool quota_enabled) 2621 { 2622 if (quota_enabled) 2623 f2fs_quota_off_umount(sbi->sb); 2624 2625 if (is_sbi_flag_set(sbi, SBI_IS_WRITABLE)) { 2626 clear_sbi_flag(sbi, SBI_IS_WRITABLE); 2627 sbi->sb->s_flags |= SB_RDONLY; 2628 } 2629 } 2630 2631 /* Read data from quotafile */ 2632 static ssize_t f2fs_quota_read(struct super_block *sb, int type, char *data, 2633 size_t len, loff_t off) 2634 { 2635 struct inode *inode = sb_dqopt(sb)->files[type]; 2636 struct address_space *mapping = inode->i_mapping; 2637 block_t blkidx = F2FS_BYTES_TO_BLK(off); 2638 int offset = off & (sb->s_blocksize - 1); 2639 int tocopy; 2640 size_t toread; 2641 loff_t i_size = i_size_read(inode); 2642 struct page *page; 2643 2644 if (off > i_size) 2645 return 0; 2646 2647 if (off + len > i_size) 2648 len = i_size - off; 2649 toread = len; 2650 while (toread > 0) { 2651 tocopy = min_t(unsigned long, sb->s_blocksize - offset, toread); 2652 repeat: 2653 page = read_cache_page_gfp(mapping, blkidx, GFP_NOFS); 2654 if (IS_ERR(page)) { 2655 if (PTR_ERR(page) == -ENOMEM) { 2656 memalloc_retry_wait(GFP_NOFS); 2657 goto repeat; 2658 } 2659 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR); 2660 return PTR_ERR(page); 2661 } 2662 2663 lock_page(page); 2664 2665 if (unlikely(page->mapping != mapping)) { 2666 f2fs_put_page(page, 1); 2667 goto repeat; 2668 } 2669 if (unlikely(!PageUptodate(page))) { 2670 f2fs_put_page(page, 1); 2671 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR); 2672 return -EIO; 2673 } 2674 2675 memcpy_from_page(data, page, offset, tocopy); 2676 f2fs_put_page(page, 1); 2677 2678 offset = 0; 2679 toread -= tocopy; 2680 data += tocopy; 2681 blkidx++; 2682 } 2683 return len; 2684 } 2685 2686 /* Write to quotafile */ 2687 static ssize_t f2fs_quota_write(struct super_block *sb, int type, 2688 const char *data, size_t len, loff_t off) 2689 { 2690 struct inode *inode = sb_dqopt(sb)->files[type]; 2691 struct address_space *mapping = inode->i_mapping; 2692 const struct address_space_operations *a_ops = mapping->a_ops; 2693 int offset = off & (sb->s_blocksize - 1); 2694 size_t towrite = len; 2695 struct page *page; 2696 void *fsdata = NULL; 2697 int err = 0; 2698 int tocopy; 2699 2700 while (towrite > 0) { 2701 tocopy = min_t(unsigned long, sb->s_blocksize - offset, 2702 towrite); 2703 retry: 2704 err = a_ops->write_begin(NULL, mapping, off, tocopy, 2705 &page, &fsdata); 2706 if (unlikely(err)) { 2707 if (err == -ENOMEM) { 2708 f2fs_io_schedule_timeout(DEFAULT_IO_TIMEOUT); 2709 goto retry; 2710 } 2711 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR); 2712 break; 2713 } 2714 2715 memcpy_to_page(page, offset, data, tocopy); 2716 2717 a_ops->write_end(NULL, mapping, off, tocopy, tocopy, 2718 page, fsdata); 2719 offset = 0; 2720 towrite -= tocopy; 2721 off += tocopy; 2722 data += tocopy; 2723 cond_resched(); 2724 } 2725 2726 if (len == towrite) 2727 return err; 2728 inode_set_mtime_to_ts(inode, inode_set_ctime_current(inode)); 2729 f2fs_mark_inode_dirty_sync(inode, false); 2730 return len - towrite; 2731 } 2732 2733 int f2fs_dquot_initialize(struct inode *inode) 2734 { 2735 if (time_to_inject(F2FS_I_SB(inode), FAULT_DQUOT_INIT)) 2736 return -ESRCH; 2737 2738 return dquot_initialize(inode); 2739 } 2740 2741 static struct dquot **f2fs_get_dquots(struct inode *inode) 2742 { 2743 return F2FS_I(inode)->i_dquot; 2744 } 2745 2746 static qsize_t *f2fs_get_reserved_space(struct inode *inode) 2747 { 2748 return &F2FS_I(inode)->i_reserved_quota; 2749 } 2750 2751 static int f2fs_quota_on_mount(struct f2fs_sb_info *sbi, int type) 2752 { 2753 if (is_set_ckpt_flags(sbi, CP_QUOTA_NEED_FSCK_FLAG)) { 2754 f2fs_err(sbi, "quota sysfile may be corrupted, skip loading it"); 2755 return 0; 2756 } 2757 2758 return dquot_quota_on_mount(sbi->sb, F2FS_OPTION(sbi).s_qf_names[type], 2759 F2FS_OPTION(sbi).s_jquota_fmt, type); 2760 } 2761 2762 int f2fs_enable_quota_files(struct f2fs_sb_info *sbi, bool rdonly) 2763 { 2764 int enabled = 0; 2765 int i, err; 2766 2767 if (f2fs_sb_has_quota_ino(sbi) && rdonly) { 2768 err = f2fs_enable_quotas(sbi->sb); 2769 if (err) { 2770 f2fs_err(sbi, "Cannot turn on quota_ino: %d", err); 2771 return 0; 2772 } 2773 return 1; 2774 } 2775 2776 for (i = 0; i < MAXQUOTAS; i++) { 2777 if (F2FS_OPTION(sbi).s_qf_names[i]) { 2778 err = f2fs_quota_on_mount(sbi, i); 2779 if (!err) { 2780 enabled = 1; 2781 continue; 2782 } 2783 f2fs_err(sbi, "Cannot turn on quotas: %d on %d", 2784 err, i); 2785 } 2786 } 2787 return enabled; 2788 } 2789 2790 static int f2fs_quota_enable(struct super_block *sb, int type, int format_id, 2791 unsigned int flags) 2792 { 2793 struct inode *qf_inode; 2794 unsigned long qf_inum; 2795 unsigned long qf_flag = F2FS_QUOTA_DEFAULT_FL; 2796 int err; 2797 2798 BUG_ON(!f2fs_sb_has_quota_ino(F2FS_SB(sb))); 2799 2800 qf_inum = f2fs_qf_ino(sb, type); 2801 if (!qf_inum) 2802 return -EPERM; 2803 2804 qf_inode = f2fs_iget(sb, qf_inum); 2805 if (IS_ERR(qf_inode)) { 2806 f2fs_err(F2FS_SB(sb), "Bad quota inode %u:%lu", type, qf_inum); 2807 return PTR_ERR(qf_inode); 2808 } 2809 2810 /* Don't account quota for quota files to avoid recursion */ 2811 inode_lock(qf_inode); 2812 qf_inode->i_flags |= S_NOQUOTA; 2813 2814 if ((F2FS_I(qf_inode)->i_flags & qf_flag) != qf_flag) { 2815 F2FS_I(qf_inode)->i_flags |= qf_flag; 2816 f2fs_set_inode_flags(qf_inode); 2817 } 2818 inode_unlock(qf_inode); 2819 2820 err = dquot_load_quota_inode(qf_inode, type, format_id, flags); 2821 iput(qf_inode); 2822 return err; 2823 } 2824 2825 static int f2fs_enable_quotas(struct super_block *sb) 2826 { 2827 struct f2fs_sb_info *sbi = F2FS_SB(sb); 2828 int type, err = 0; 2829 unsigned long qf_inum; 2830 bool quota_mopt[MAXQUOTAS] = { 2831 test_opt(sbi, USRQUOTA), 2832 test_opt(sbi, GRPQUOTA), 2833 test_opt(sbi, PRJQUOTA), 2834 }; 2835 2836 if (is_set_ckpt_flags(F2FS_SB(sb), CP_QUOTA_NEED_FSCK_FLAG)) { 2837 f2fs_err(sbi, "quota file may be corrupted, skip loading it"); 2838 return 0; 2839 } 2840 2841 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE; 2842 2843 for (type = 0; type < MAXQUOTAS; type++) { 2844 qf_inum = f2fs_qf_ino(sb, type); 2845 if (qf_inum) { 2846 err = f2fs_quota_enable(sb, type, QFMT_VFS_V1, 2847 DQUOT_USAGE_ENABLED | 2848 (quota_mopt[type] ? DQUOT_LIMITS_ENABLED : 0)); 2849 if (err) { 2850 f2fs_err(sbi, "Failed to enable quota tracking (type=%d, err=%d). Please run fsck to fix.", 2851 type, err); 2852 for (type--; type >= 0; type--) 2853 dquot_quota_off(sb, type); 2854 set_sbi_flag(F2FS_SB(sb), 2855 SBI_QUOTA_NEED_REPAIR); 2856 return err; 2857 } 2858 } 2859 } 2860 return 0; 2861 } 2862 2863 static int f2fs_quota_sync_file(struct f2fs_sb_info *sbi, int type) 2864 { 2865 struct quota_info *dqopt = sb_dqopt(sbi->sb); 2866 struct address_space *mapping = dqopt->files[type]->i_mapping; 2867 int ret = 0; 2868 2869 ret = dquot_writeback_dquots(sbi->sb, type); 2870 if (ret) 2871 goto out; 2872 2873 ret = filemap_fdatawrite(mapping); 2874 if (ret) 2875 goto out; 2876 2877 /* if we are using journalled quota */ 2878 if (is_journalled_quota(sbi)) 2879 goto out; 2880 2881 ret = filemap_fdatawait(mapping); 2882 2883 truncate_inode_pages(&dqopt->files[type]->i_data, 0); 2884 out: 2885 if (ret) 2886 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR); 2887 return ret; 2888 } 2889 2890 int f2fs_quota_sync(struct super_block *sb, int type) 2891 { 2892 struct f2fs_sb_info *sbi = F2FS_SB(sb); 2893 struct quota_info *dqopt = sb_dqopt(sb); 2894 int cnt; 2895 int ret = 0; 2896 2897 /* 2898 * Now when everything is written we can discard the pagecache so 2899 * that userspace sees the changes. 2900 */ 2901 for (cnt = 0; cnt < MAXQUOTAS; cnt++) { 2902 2903 if (type != -1 && cnt != type) 2904 continue; 2905 2906 if (!sb_has_quota_active(sb, cnt)) 2907 continue; 2908 2909 if (!f2fs_sb_has_quota_ino(sbi)) 2910 inode_lock(dqopt->files[cnt]); 2911 2912 /* 2913 * do_quotactl 2914 * f2fs_quota_sync 2915 * f2fs_down_read(quota_sem) 2916 * dquot_writeback_dquots() 2917 * f2fs_dquot_commit 2918 * block_operation 2919 * f2fs_down_read(quota_sem) 2920 */ 2921 f2fs_lock_op(sbi); 2922 f2fs_down_read(&sbi->quota_sem); 2923 2924 ret = f2fs_quota_sync_file(sbi, cnt); 2925 2926 f2fs_up_read(&sbi->quota_sem); 2927 f2fs_unlock_op(sbi); 2928 2929 if (!f2fs_sb_has_quota_ino(sbi)) 2930 inode_unlock(dqopt->files[cnt]); 2931 2932 if (ret) 2933 break; 2934 } 2935 return ret; 2936 } 2937 2938 static int f2fs_quota_on(struct super_block *sb, int type, int format_id, 2939 const struct path *path) 2940 { 2941 struct inode *inode; 2942 int err; 2943 2944 /* if quota sysfile exists, deny enabling quota with specific file */ 2945 if (f2fs_sb_has_quota_ino(F2FS_SB(sb))) { 2946 f2fs_err(F2FS_SB(sb), "quota sysfile already exists"); 2947 return -EBUSY; 2948 } 2949 2950 if (path->dentry->d_sb != sb) 2951 return -EXDEV; 2952 2953 err = f2fs_quota_sync(sb, type); 2954 if (err) 2955 return err; 2956 2957 inode = d_inode(path->dentry); 2958 2959 err = filemap_fdatawrite(inode->i_mapping); 2960 if (err) 2961 return err; 2962 2963 err = filemap_fdatawait(inode->i_mapping); 2964 if (err) 2965 return err; 2966 2967 err = dquot_quota_on(sb, type, format_id, path); 2968 if (err) 2969 return err; 2970 2971 inode_lock(inode); 2972 F2FS_I(inode)->i_flags |= F2FS_QUOTA_DEFAULT_FL; 2973 f2fs_set_inode_flags(inode); 2974 inode_unlock(inode); 2975 f2fs_mark_inode_dirty_sync(inode, false); 2976 2977 return 0; 2978 } 2979 2980 static int __f2fs_quota_off(struct super_block *sb, int type) 2981 { 2982 struct inode *inode = sb_dqopt(sb)->files[type]; 2983 int err; 2984 2985 if (!inode || !igrab(inode)) 2986 return dquot_quota_off(sb, type); 2987 2988 err = f2fs_quota_sync(sb, type); 2989 if (err) 2990 goto out_put; 2991 2992 err = dquot_quota_off(sb, type); 2993 if (err || f2fs_sb_has_quota_ino(F2FS_SB(sb))) 2994 goto out_put; 2995 2996 inode_lock(inode); 2997 F2FS_I(inode)->i_flags &= ~F2FS_QUOTA_DEFAULT_FL; 2998 f2fs_set_inode_flags(inode); 2999 inode_unlock(inode); 3000 f2fs_mark_inode_dirty_sync(inode, false); 3001 out_put: 3002 iput(inode); 3003 return err; 3004 } 3005 3006 static int f2fs_quota_off(struct super_block *sb, int type) 3007 { 3008 struct f2fs_sb_info *sbi = F2FS_SB(sb); 3009 int err; 3010 3011 err = __f2fs_quota_off(sb, type); 3012 3013 /* 3014 * quotactl can shutdown journalled quota, result in inconsistence 3015 * between quota record and fs data by following updates, tag the 3016 * flag to let fsck be aware of it. 3017 */ 3018 if (is_journalled_quota(sbi)) 3019 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR); 3020 return err; 3021 } 3022 3023 void f2fs_quota_off_umount(struct super_block *sb) 3024 { 3025 int type; 3026 int err; 3027 3028 for (type = 0; type < MAXQUOTAS; type++) { 3029 err = __f2fs_quota_off(sb, type); 3030 if (err) { 3031 int ret = dquot_quota_off(sb, type); 3032 3033 f2fs_err(F2FS_SB(sb), "Fail to turn off disk quota (type: %d, err: %d, ret:%d), Please run fsck to fix it.", 3034 type, err, ret); 3035 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR); 3036 } 3037 } 3038 /* 3039 * In case of checkpoint=disable, we must flush quota blocks. 3040 * This can cause NULL exception for node_inode in end_io, since 3041 * put_super already dropped it. 3042 */ 3043 sync_filesystem(sb); 3044 } 3045 3046 static void f2fs_truncate_quota_inode_pages(struct super_block *sb) 3047 { 3048 struct quota_info *dqopt = sb_dqopt(sb); 3049 int type; 3050 3051 for (type = 0; type < MAXQUOTAS; type++) { 3052 if (!dqopt->files[type]) 3053 continue; 3054 f2fs_inode_synced(dqopt->files[type]); 3055 } 3056 } 3057 3058 static int f2fs_dquot_commit(struct dquot *dquot) 3059 { 3060 struct f2fs_sb_info *sbi = F2FS_SB(dquot->dq_sb); 3061 int ret; 3062 3063 f2fs_down_read_nested(&sbi->quota_sem, SINGLE_DEPTH_NESTING); 3064 ret = dquot_commit(dquot); 3065 if (ret < 0) 3066 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR); 3067 f2fs_up_read(&sbi->quota_sem); 3068 return ret; 3069 } 3070 3071 static int f2fs_dquot_acquire(struct dquot *dquot) 3072 { 3073 struct f2fs_sb_info *sbi = F2FS_SB(dquot->dq_sb); 3074 int ret; 3075 3076 f2fs_down_read(&sbi->quota_sem); 3077 ret = dquot_acquire(dquot); 3078 if (ret < 0) 3079 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR); 3080 f2fs_up_read(&sbi->quota_sem); 3081 return ret; 3082 } 3083 3084 static int f2fs_dquot_release(struct dquot *dquot) 3085 { 3086 struct f2fs_sb_info *sbi = F2FS_SB(dquot->dq_sb); 3087 int ret = dquot_release(dquot); 3088 3089 if (ret < 0) 3090 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR); 3091 return ret; 3092 } 3093 3094 static int f2fs_dquot_mark_dquot_dirty(struct dquot *dquot) 3095 { 3096 struct super_block *sb = dquot->dq_sb; 3097 struct f2fs_sb_info *sbi = F2FS_SB(sb); 3098 int ret = dquot_mark_dquot_dirty(dquot); 3099 3100 /* if we are using journalled quota */ 3101 if (is_journalled_quota(sbi)) 3102 set_sbi_flag(sbi, SBI_QUOTA_NEED_FLUSH); 3103 3104 return ret; 3105 } 3106 3107 static int f2fs_dquot_commit_info(struct super_block *sb, int type) 3108 { 3109 struct f2fs_sb_info *sbi = F2FS_SB(sb); 3110 int ret = dquot_commit_info(sb, type); 3111 3112 if (ret < 0) 3113 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR); 3114 return ret; 3115 } 3116 3117 static int f2fs_get_projid(struct inode *inode, kprojid_t *projid) 3118 { 3119 *projid = F2FS_I(inode)->i_projid; 3120 return 0; 3121 } 3122 3123 static const struct dquot_operations f2fs_quota_operations = { 3124 .get_reserved_space = f2fs_get_reserved_space, 3125 .write_dquot = f2fs_dquot_commit, 3126 .acquire_dquot = f2fs_dquot_acquire, 3127 .release_dquot = f2fs_dquot_release, 3128 .mark_dirty = f2fs_dquot_mark_dquot_dirty, 3129 .write_info = f2fs_dquot_commit_info, 3130 .alloc_dquot = dquot_alloc, 3131 .destroy_dquot = dquot_destroy, 3132 .get_projid = f2fs_get_projid, 3133 .get_next_id = dquot_get_next_id, 3134 }; 3135 3136 static const struct quotactl_ops f2fs_quotactl_ops = { 3137 .quota_on = f2fs_quota_on, 3138 .quota_off = f2fs_quota_off, 3139 .quota_sync = f2fs_quota_sync, 3140 .get_state = dquot_get_state, 3141 .set_info = dquot_set_dqinfo, 3142 .get_dqblk = dquot_get_dqblk, 3143 .set_dqblk = dquot_set_dqblk, 3144 .get_nextdqblk = dquot_get_next_dqblk, 3145 }; 3146 #else 3147 int f2fs_dquot_initialize(struct inode *inode) 3148 { 3149 return 0; 3150 } 3151 3152 int f2fs_quota_sync(struct super_block *sb, int type) 3153 { 3154 return 0; 3155 } 3156 3157 void f2fs_quota_off_umount(struct super_block *sb) 3158 { 3159 } 3160 #endif 3161 3162 static const struct super_operations f2fs_sops = { 3163 .alloc_inode = f2fs_alloc_inode, 3164 .free_inode = f2fs_free_inode, 3165 .drop_inode = f2fs_drop_inode, 3166 .write_inode = f2fs_write_inode, 3167 .dirty_inode = f2fs_dirty_inode, 3168 .show_options = f2fs_show_options, 3169 #ifdef CONFIG_QUOTA 3170 .quota_read = f2fs_quota_read, 3171 .quota_write = f2fs_quota_write, 3172 .get_dquots = f2fs_get_dquots, 3173 #endif 3174 .evict_inode = f2fs_evict_inode, 3175 .put_super = f2fs_put_super, 3176 .sync_fs = f2fs_sync_fs, 3177 .freeze_fs = f2fs_freeze, 3178 .unfreeze_fs = f2fs_unfreeze, 3179 .statfs = f2fs_statfs, 3180 .remount_fs = f2fs_remount, 3181 }; 3182 3183 #ifdef CONFIG_FS_ENCRYPTION 3184 static int f2fs_get_context(struct inode *inode, void *ctx, size_t len) 3185 { 3186 return f2fs_getxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION, 3187 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT, 3188 ctx, len, NULL); 3189 } 3190 3191 static int f2fs_set_context(struct inode *inode, const void *ctx, size_t len, 3192 void *fs_data) 3193 { 3194 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 3195 3196 /* 3197 * Encrypting the root directory is not allowed because fsck 3198 * expects lost+found directory to exist and remain unencrypted 3199 * if LOST_FOUND feature is enabled. 3200 * 3201 */ 3202 if (f2fs_sb_has_lost_found(sbi) && 3203 inode->i_ino == F2FS_ROOT_INO(sbi)) 3204 return -EPERM; 3205 3206 return f2fs_setxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION, 3207 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT, 3208 ctx, len, fs_data, XATTR_CREATE); 3209 } 3210 3211 static const union fscrypt_policy *f2fs_get_dummy_policy(struct super_block *sb) 3212 { 3213 return F2FS_OPTION(F2FS_SB(sb)).dummy_enc_policy.policy; 3214 } 3215 3216 static bool f2fs_has_stable_inodes(struct super_block *sb) 3217 { 3218 return true; 3219 } 3220 3221 static struct block_device **f2fs_get_devices(struct super_block *sb, 3222 unsigned int *num_devs) 3223 { 3224 struct f2fs_sb_info *sbi = F2FS_SB(sb); 3225 struct block_device **devs; 3226 int i; 3227 3228 if (!f2fs_is_multi_device(sbi)) 3229 return NULL; 3230 3231 devs = kmalloc_array(sbi->s_ndevs, sizeof(*devs), GFP_KERNEL); 3232 if (!devs) 3233 return ERR_PTR(-ENOMEM); 3234 3235 for (i = 0; i < sbi->s_ndevs; i++) 3236 devs[i] = FDEV(i).bdev; 3237 *num_devs = sbi->s_ndevs; 3238 return devs; 3239 } 3240 3241 static const struct fscrypt_operations f2fs_cryptops = { 3242 .needs_bounce_pages = 1, 3243 .has_32bit_inodes = 1, 3244 .supports_subblock_data_units = 1, 3245 .legacy_key_prefix = "f2fs:", 3246 .get_context = f2fs_get_context, 3247 .set_context = f2fs_set_context, 3248 .get_dummy_policy = f2fs_get_dummy_policy, 3249 .empty_dir = f2fs_empty_dir, 3250 .has_stable_inodes = f2fs_has_stable_inodes, 3251 .get_devices = f2fs_get_devices, 3252 }; 3253 #endif 3254 3255 static struct inode *f2fs_nfs_get_inode(struct super_block *sb, 3256 u64 ino, u32 generation) 3257 { 3258 struct f2fs_sb_info *sbi = F2FS_SB(sb); 3259 struct inode *inode; 3260 3261 if (f2fs_check_nid_range(sbi, ino)) 3262 return ERR_PTR(-ESTALE); 3263 3264 /* 3265 * f2fs_iget isn't quite right if the inode is currently unallocated! 3266 * However f2fs_iget currently does appropriate checks to handle stale 3267 * inodes so everything is OK. 3268 */ 3269 inode = f2fs_iget(sb, ino); 3270 if (IS_ERR(inode)) 3271 return ERR_CAST(inode); 3272 if (unlikely(generation && inode->i_generation != generation)) { 3273 /* we didn't find the right inode.. */ 3274 iput(inode); 3275 return ERR_PTR(-ESTALE); 3276 } 3277 return inode; 3278 } 3279 3280 static struct dentry *f2fs_fh_to_dentry(struct super_block *sb, struct fid *fid, 3281 int fh_len, int fh_type) 3282 { 3283 return generic_fh_to_dentry(sb, fid, fh_len, fh_type, 3284 f2fs_nfs_get_inode); 3285 } 3286 3287 static struct dentry *f2fs_fh_to_parent(struct super_block *sb, struct fid *fid, 3288 int fh_len, int fh_type) 3289 { 3290 return generic_fh_to_parent(sb, fid, fh_len, fh_type, 3291 f2fs_nfs_get_inode); 3292 } 3293 3294 static const struct export_operations f2fs_export_ops = { 3295 .fh_to_dentry = f2fs_fh_to_dentry, 3296 .fh_to_parent = f2fs_fh_to_parent, 3297 .get_parent = f2fs_get_parent, 3298 }; 3299 3300 loff_t max_file_blocks(struct inode *inode) 3301 { 3302 loff_t result = 0; 3303 loff_t leaf_count; 3304 3305 /* 3306 * note: previously, result is equal to (DEF_ADDRS_PER_INODE - 3307 * DEFAULT_INLINE_XATTR_ADDRS), but now f2fs try to reserve more 3308 * space in inode.i_addr, it will be more safe to reassign 3309 * result as zero. 3310 */ 3311 3312 if (inode && f2fs_compressed_file(inode)) 3313 leaf_count = ADDRS_PER_BLOCK(inode); 3314 else 3315 leaf_count = DEF_ADDRS_PER_BLOCK; 3316 3317 /* two direct node blocks */ 3318 result += (leaf_count * 2); 3319 3320 /* two indirect node blocks */ 3321 leaf_count *= NIDS_PER_BLOCK; 3322 result += (leaf_count * 2); 3323 3324 /* one double indirect node block */ 3325 leaf_count *= NIDS_PER_BLOCK; 3326 result += leaf_count; 3327 3328 return result; 3329 } 3330 3331 static int __f2fs_commit_super(struct buffer_head *bh, 3332 struct f2fs_super_block *super) 3333 { 3334 lock_buffer(bh); 3335 if (super) 3336 memcpy(bh->b_data + F2FS_SUPER_OFFSET, super, sizeof(*super)); 3337 set_buffer_dirty(bh); 3338 unlock_buffer(bh); 3339 3340 /* it's rare case, we can do fua all the time */ 3341 return __sync_dirty_buffer(bh, REQ_SYNC | REQ_PREFLUSH | REQ_FUA); 3342 } 3343 3344 static inline bool sanity_check_area_boundary(struct f2fs_sb_info *sbi, 3345 struct buffer_head *bh) 3346 { 3347 struct f2fs_super_block *raw_super = (struct f2fs_super_block *) 3348 (bh->b_data + F2FS_SUPER_OFFSET); 3349 struct super_block *sb = sbi->sb; 3350 u32 segment0_blkaddr = le32_to_cpu(raw_super->segment0_blkaddr); 3351 u32 cp_blkaddr = le32_to_cpu(raw_super->cp_blkaddr); 3352 u32 sit_blkaddr = le32_to_cpu(raw_super->sit_blkaddr); 3353 u32 nat_blkaddr = le32_to_cpu(raw_super->nat_blkaddr); 3354 u32 ssa_blkaddr = le32_to_cpu(raw_super->ssa_blkaddr); 3355 u32 main_blkaddr = le32_to_cpu(raw_super->main_blkaddr); 3356 u32 segment_count_ckpt = le32_to_cpu(raw_super->segment_count_ckpt); 3357 u32 segment_count_sit = le32_to_cpu(raw_super->segment_count_sit); 3358 u32 segment_count_nat = le32_to_cpu(raw_super->segment_count_nat); 3359 u32 segment_count_ssa = le32_to_cpu(raw_super->segment_count_ssa); 3360 u32 segment_count_main = le32_to_cpu(raw_super->segment_count_main); 3361 u32 segment_count = le32_to_cpu(raw_super->segment_count); 3362 u32 log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg); 3363 u64 main_end_blkaddr = main_blkaddr + 3364 (segment_count_main << log_blocks_per_seg); 3365 u64 seg_end_blkaddr = segment0_blkaddr + 3366 (segment_count << log_blocks_per_seg); 3367 3368 if (segment0_blkaddr != cp_blkaddr) { 3369 f2fs_info(sbi, "Mismatch start address, segment0(%u) cp_blkaddr(%u)", 3370 segment0_blkaddr, cp_blkaddr); 3371 return true; 3372 } 3373 3374 if (cp_blkaddr + (segment_count_ckpt << log_blocks_per_seg) != 3375 sit_blkaddr) { 3376 f2fs_info(sbi, "Wrong CP boundary, start(%u) end(%u) blocks(%u)", 3377 cp_blkaddr, sit_blkaddr, 3378 segment_count_ckpt << log_blocks_per_seg); 3379 return true; 3380 } 3381 3382 if (sit_blkaddr + (segment_count_sit << log_blocks_per_seg) != 3383 nat_blkaddr) { 3384 f2fs_info(sbi, "Wrong SIT boundary, start(%u) end(%u) blocks(%u)", 3385 sit_blkaddr, nat_blkaddr, 3386 segment_count_sit << log_blocks_per_seg); 3387 return true; 3388 } 3389 3390 if (nat_blkaddr + (segment_count_nat << log_blocks_per_seg) != 3391 ssa_blkaddr) { 3392 f2fs_info(sbi, "Wrong NAT boundary, start(%u) end(%u) blocks(%u)", 3393 nat_blkaddr, ssa_blkaddr, 3394 segment_count_nat << log_blocks_per_seg); 3395 return true; 3396 } 3397 3398 if (ssa_blkaddr + (segment_count_ssa << log_blocks_per_seg) != 3399 main_blkaddr) { 3400 f2fs_info(sbi, "Wrong SSA boundary, start(%u) end(%u) blocks(%u)", 3401 ssa_blkaddr, main_blkaddr, 3402 segment_count_ssa << log_blocks_per_seg); 3403 return true; 3404 } 3405 3406 if (main_end_blkaddr > seg_end_blkaddr) { 3407 f2fs_info(sbi, "Wrong MAIN_AREA boundary, start(%u) end(%llu) block(%u)", 3408 main_blkaddr, seg_end_blkaddr, 3409 segment_count_main << log_blocks_per_seg); 3410 return true; 3411 } else if (main_end_blkaddr < seg_end_blkaddr) { 3412 int err = 0; 3413 char *res; 3414 3415 /* fix in-memory information all the time */ 3416 raw_super->segment_count = cpu_to_le32((main_end_blkaddr - 3417 segment0_blkaddr) >> log_blocks_per_seg); 3418 3419 if (f2fs_readonly(sb) || f2fs_hw_is_readonly(sbi)) { 3420 set_sbi_flag(sbi, SBI_NEED_SB_WRITE); 3421 res = "internally"; 3422 } else { 3423 err = __f2fs_commit_super(bh, NULL); 3424 res = err ? "failed" : "done"; 3425 } 3426 f2fs_info(sbi, "Fix alignment : %s, start(%u) end(%llu) block(%u)", 3427 res, main_blkaddr, seg_end_blkaddr, 3428 segment_count_main << log_blocks_per_seg); 3429 if (err) 3430 return true; 3431 } 3432 return false; 3433 } 3434 3435 static int sanity_check_raw_super(struct f2fs_sb_info *sbi, 3436 struct buffer_head *bh) 3437 { 3438 block_t segment_count, segs_per_sec, secs_per_zone, segment_count_main; 3439 block_t total_sections, blocks_per_seg; 3440 struct f2fs_super_block *raw_super = (struct f2fs_super_block *) 3441 (bh->b_data + F2FS_SUPER_OFFSET); 3442 size_t crc_offset = 0; 3443 __u32 crc = 0; 3444 3445 if (le32_to_cpu(raw_super->magic) != F2FS_SUPER_MAGIC) { 3446 f2fs_info(sbi, "Magic Mismatch, valid(0x%x) - read(0x%x)", 3447 F2FS_SUPER_MAGIC, le32_to_cpu(raw_super->magic)); 3448 return -EINVAL; 3449 } 3450 3451 /* Check checksum_offset and crc in superblock */ 3452 if (__F2FS_HAS_FEATURE(raw_super, F2FS_FEATURE_SB_CHKSUM)) { 3453 crc_offset = le32_to_cpu(raw_super->checksum_offset); 3454 if (crc_offset != 3455 offsetof(struct f2fs_super_block, crc)) { 3456 f2fs_info(sbi, "Invalid SB checksum offset: %zu", 3457 crc_offset); 3458 return -EFSCORRUPTED; 3459 } 3460 crc = le32_to_cpu(raw_super->crc); 3461 if (!f2fs_crc_valid(sbi, crc, raw_super, crc_offset)) { 3462 f2fs_info(sbi, "Invalid SB checksum value: %u", crc); 3463 return -EFSCORRUPTED; 3464 } 3465 } 3466 3467 /* Currently, support only 4KB block size */ 3468 if (le32_to_cpu(raw_super->log_blocksize) != F2FS_BLKSIZE_BITS) { 3469 f2fs_info(sbi, "Invalid log_blocksize (%u), supports only %u", 3470 le32_to_cpu(raw_super->log_blocksize), 3471 F2FS_BLKSIZE_BITS); 3472 return -EFSCORRUPTED; 3473 } 3474 3475 /* check log blocks per segment */ 3476 if (le32_to_cpu(raw_super->log_blocks_per_seg) != 9) { 3477 f2fs_info(sbi, "Invalid log blocks per segment (%u)", 3478 le32_to_cpu(raw_super->log_blocks_per_seg)); 3479 return -EFSCORRUPTED; 3480 } 3481 3482 /* Currently, support 512/1024/2048/4096 bytes sector size */ 3483 if (le32_to_cpu(raw_super->log_sectorsize) > 3484 F2FS_MAX_LOG_SECTOR_SIZE || 3485 le32_to_cpu(raw_super->log_sectorsize) < 3486 F2FS_MIN_LOG_SECTOR_SIZE) { 3487 f2fs_info(sbi, "Invalid log sectorsize (%u)", 3488 le32_to_cpu(raw_super->log_sectorsize)); 3489 return -EFSCORRUPTED; 3490 } 3491 if (le32_to_cpu(raw_super->log_sectors_per_block) + 3492 le32_to_cpu(raw_super->log_sectorsize) != 3493 F2FS_MAX_LOG_SECTOR_SIZE) { 3494 f2fs_info(sbi, "Invalid log sectors per block(%u) log sectorsize(%u)", 3495 le32_to_cpu(raw_super->log_sectors_per_block), 3496 le32_to_cpu(raw_super->log_sectorsize)); 3497 return -EFSCORRUPTED; 3498 } 3499 3500 segment_count = le32_to_cpu(raw_super->segment_count); 3501 segment_count_main = le32_to_cpu(raw_super->segment_count_main); 3502 segs_per_sec = le32_to_cpu(raw_super->segs_per_sec); 3503 secs_per_zone = le32_to_cpu(raw_super->secs_per_zone); 3504 total_sections = le32_to_cpu(raw_super->section_count); 3505 3506 /* blocks_per_seg should be 512, given the above check */ 3507 blocks_per_seg = BIT(le32_to_cpu(raw_super->log_blocks_per_seg)); 3508 3509 if (segment_count > F2FS_MAX_SEGMENT || 3510 segment_count < F2FS_MIN_SEGMENTS) { 3511 f2fs_info(sbi, "Invalid segment count (%u)", segment_count); 3512 return -EFSCORRUPTED; 3513 } 3514 3515 if (total_sections > segment_count_main || total_sections < 1 || 3516 segs_per_sec > segment_count || !segs_per_sec) { 3517 f2fs_info(sbi, "Invalid segment/section count (%u, %u x %u)", 3518 segment_count, total_sections, segs_per_sec); 3519 return -EFSCORRUPTED; 3520 } 3521 3522 if (segment_count_main != total_sections * segs_per_sec) { 3523 f2fs_info(sbi, "Invalid segment/section count (%u != %u * %u)", 3524 segment_count_main, total_sections, segs_per_sec); 3525 return -EFSCORRUPTED; 3526 } 3527 3528 if ((segment_count / segs_per_sec) < total_sections) { 3529 f2fs_info(sbi, "Small segment_count (%u < %u * %u)", 3530 segment_count, segs_per_sec, total_sections); 3531 return -EFSCORRUPTED; 3532 } 3533 3534 if (segment_count > (le64_to_cpu(raw_super->block_count) >> 9)) { 3535 f2fs_info(sbi, "Wrong segment_count / block_count (%u > %llu)", 3536 segment_count, le64_to_cpu(raw_super->block_count)); 3537 return -EFSCORRUPTED; 3538 } 3539 3540 if (RDEV(0).path[0]) { 3541 block_t dev_seg_count = le32_to_cpu(RDEV(0).total_segments); 3542 int i = 1; 3543 3544 while (i < MAX_DEVICES && RDEV(i).path[0]) { 3545 dev_seg_count += le32_to_cpu(RDEV(i).total_segments); 3546 i++; 3547 } 3548 if (segment_count != dev_seg_count) { 3549 f2fs_info(sbi, "Segment count (%u) mismatch with total segments from devices (%u)", 3550 segment_count, dev_seg_count); 3551 return -EFSCORRUPTED; 3552 } 3553 } else { 3554 if (__F2FS_HAS_FEATURE(raw_super, F2FS_FEATURE_BLKZONED) && 3555 !bdev_is_zoned(sbi->sb->s_bdev)) { 3556 f2fs_info(sbi, "Zoned block device path is missing"); 3557 return -EFSCORRUPTED; 3558 } 3559 } 3560 3561 if (secs_per_zone > total_sections || !secs_per_zone) { 3562 f2fs_info(sbi, "Wrong secs_per_zone / total_sections (%u, %u)", 3563 secs_per_zone, total_sections); 3564 return -EFSCORRUPTED; 3565 } 3566 if (le32_to_cpu(raw_super->extension_count) > F2FS_MAX_EXTENSION || 3567 raw_super->hot_ext_count > F2FS_MAX_EXTENSION || 3568 (le32_to_cpu(raw_super->extension_count) + 3569 raw_super->hot_ext_count) > F2FS_MAX_EXTENSION) { 3570 f2fs_info(sbi, "Corrupted extension count (%u + %u > %u)", 3571 le32_to_cpu(raw_super->extension_count), 3572 raw_super->hot_ext_count, 3573 F2FS_MAX_EXTENSION); 3574 return -EFSCORRUPTED; 3575 } 3576 3577 if (le32_to_cpu(raw_super->cp_payload) >= 3578 (blocks_per_seg - F2FS_CP_PACKS - 3579 NR_CURSEG_PERSIST_TYPE)) { 3580 f2fs_info(sbi, "Insane cp_payload (%u >= %u)", 3581 le32_to_cpu(raw_super->cp_payload), 3582 blocks_per_seg - F2FS_CP_PACKS - 3583 NR_CURSEG_PERSIST_TYPE); 3584 return -EFSCORRUPTED; 3585 } 3586 3587 /* check reserved ino info */ 3588 if (le32_to_cpu(raw_super->node_ino) != 1 || 3589 le32_to_cpu(raw_super->meta_ino) != 2 || 3590 le32_to_cpu(raw_super->root_ino) != 3) { 3591 f2fs_info(sbi, "Invalid Fs Meta Ino: node(%u) meta(%u) root(%u)", 3592 le32_to_cpu(raw_super->node_ino), 3593 le32_to_cpu(raw_super->meta_ino), 3594 le32_to_cpu(raw_super->root_ino)); 3595 return -EFSCORRUPTED; 3596 } 3597 3598 /* check CP/SIT/NAT/SSA/MAIN_AREA area boundary */ 3599 if (sanity_check_area_boundary(sbi, bh)) 3600 return -EFSCORRUPTED; 3601 3602 return 0; 3603 } 3604 3605 int f2fs_sanity_check_ckpt(struct f2fs_sb_info *sbi) 3606 { 3607 unsigned int total, fsmeta; 3608 struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi); 3609 struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi); 3610 unsigned int ovp_segments, reserved_segments; 3611 unsigned int main_segs, blocks_per_seg; 3612 unsigned int sit_segs, nat_segs; 3613 unsigned int sit_bitmap_size, nat_bitmap_size; 3614 unsigned int log_blocks_per_seg; 3615 unsigned int segment_count_main; 3616 unsigned int cp_pack_start_sum, cp_payload; 3617 block_t user_block_count, valid_user_blocks; 3618 block_t avail_node_count, valid_node_count; 3619 unsigned int nat_blocks, nat_bits_bytes, nat_bits_blocks; 3620 int i, j; 3621 3622 total = le32_to_cpu(raw_super->segment_count); 3623 fsmeta = le32_to_cpu(raw_super->segment_count_ckpt); 3624 sit_segs = le32_to_cpu(raw_super->segment_count_sit); 3625 fsmeta += sit_segs; 3626 nat_segs = le32_to_cpu(raw_super->segment_count_nat); 3627 fsmeta += nat_segs; 3628 fsmeta += le32_to_cpu(ckpt->rsvd_segment_count); 3629 fsmeta += le32_to_cpu(raw_super->segment_count_ssa); 3630 3631 if (unlikely(fsmeta >= total)) 3632 return 1; 3633 3634 ovp_segments = le32_to_cpu(ckpt->overprov_segment_count); 3635 reserved_segments = le32_to_cpu(ckpt->rsvd_segment_count); 3636 3637 if (!f2fs_sb_has_readonly(sbi) && 3638 unlikely(fsmeta < F2FS_MIN_META_SEGMENTS || 3639 ovp_segments == 0 || reserved_segments == 0)) { 3640 f2fs_err(sbi, "Wrong layout: check mkfs.f2fs version"); 3641 return 1; 3642 } 3643 user_block_count = le64_to_cpu(ckpt->user_block_count); 3644 segment_count_main = le32_to_cpu(raw_super->segment_count_main) + 3645 (f2fs_sb_has_readonly(sbi) ? 1 : 0); 3646 log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg); 3647 if (!user_block_count || user_block_count >= 3648 segment_count_main << log_blocks_per_seg) { 3649 f2fs_err(sbi, "Wrong user_block_count: %u", 3650 user_block_count); 3651 return 1; 3652 } 3653 3654 valid_user_blocks = le64_to_cpu(ckpt->valid_block_count); 3655 if (valid_user_blocks > user_block_count) { 3656 f2fs_err(sbi, "Wrong valid_user_blocks: %u, user_block_count: %u", 3657 valid_user_blocks, user_block_count); 3658 return 1; 3659 } 3660 3661 valid_node_count = le32_to_cpu(ckpt->valid_node_count); 3662 avail_node_count = sbi->total_node_count - F2FS_RESERVED_NODE_NUM; 3663 if (valid_node_count > avail_node_count) { 3664 f2fs_err(sbi, "Wrong valid_node_count: %u, avail_node_count: %u", 3665 valid_node_count, avail_node_count); 3666 return 1; 3667 } 3668 3669 main_segs = le32_to_cpu(raw_super->segment_count_main); 3670 blocks_per_seg = sbi->blocks_per_seg; 3671 3672 for (i = 0; i < NR_CURSEG_NODE_TYPE; i++) { 3673 if (le32_to_cpu(ckpt->cur_node_segno[i]) >= main_segs || 3674 le16_to_cpu(ckpt->cur_node_blkoff[i]) >= blocks_per_seg) 3675 return 1; 3676 3677 if (f2fs_sb_has_readonly(sbi)) 3678 goto check_data; 3679 3680 for (j = i + 1; j < NR_CURSEG_NODE_TYPE; j++) { 3681 if (le32_to_cpu(ckpt->cur_node_segno[i]) == 3682 le32_to_cpu(ckpt->cur_node_segno[j])) { 3683 f2fs_err(sbi, "Node segment (%u, %u) has the same segno: %u", 3684 i, j, 3685 le32_to_cpu(ckpt->cur_node_segno[i])); 3686 return 1; 3687 } 3688 } 3689 } 3690 check_data: 3691 for (i = 0; i < NR_CURSEG_DATA_TYPE; i++) { 3692 if (le32_to_cpu(ckpt->cur_data_segno[i]) >= main_segs || 3693 le16_to_cpu(ckpt->cur_data_blkoff[i]) >= blocks_per_seg) 3694 return 1; 3695 3696 if (f2fs_sb_has_readonly(sbi)) 3697 goto skip_cross; 3698 3699 for (j = i + 1; j < NR_CURSEG_DATA_TYPE; j++) { 3700 if (le32_to_cpu(ckpt->cur_data_segno[i]) == 3701 le32_to_cpu(ckpt->cur_data_segno[j])) { 3702 f2fs_err(sbi, "Data segment (%u, %u) has the same segno: %u", 3703 i, j, 3704 le32_to_cpu(ckpt->cur_data_segno[i])); 3705 return 1; 3706 } 3707 } 3708 } 3709 for (i = 0; i < NR_CURSEG_NODE_TYPE; i++) { 3710 for (j = 0; j < NR_CURSEG_DATA_TYPE; j++) { 3711 if (le32_to_cpu(ckpt->cur_node_segno[i]) == 3712 le32_to_cpu(ckpt->cur_data_segno[j])) { 3713 f2fs_err(sbi, "Node segment (%u) and Data segment (%u) has the same segno: %u", 3714 i, j, 3715 le32_to_cpu(ckpt->cur_node_segno[i])); 3716 return 1; 3717 } 3718 } 3719 } 3720 skip_cross: 3721 sit_bitmap_size = le32_to_cpu(ckpt->sit_ver_bitmap_bytesize); 3722 nat_bitmap_size = le32_to_cpu(ckpt->nat_ver_bitmap_bytesize); 3723 3724 if (sit_bitmap_size != ((sit_segs / 2) << log_blocks_per_seg) / 8 || 3725 nat_bitmap_size != ((nat_segs / 2) << log_blocks_per_seg) / 8) { 3726 f2fs_err(sbi, "Wrong bitmap size: sit: %u, nat:%u", 3727 sit_bitmap_size, nat_bitmap_size); 3728 return 1; 3729 } 3730 3731 cp_pack_start_sum = __start_sum_addr(sbi); 3732 cp_payload = __cp_payload(sbi); 3733 if (cp_pack_start_sum < cp_payload + 1 || 3734 cp_pack_start_sum > blocks_per_seg - 1 - 3735 NR_CURSEG_PERSIST_TYPE) { 3736 f2fs_err(sbi, "Wrong cp_pack_start_sum: %u", 3737 cp_pack_start_sum); 3738 return 1; 3739 } 3740 3741 if (__is_set_ckpt_flags(ckpt, CP_LARGE_NAT_BITMAP_FLAG) && 3742 le32_to_cpu(ckpt->checksum_offset) != CP_MIN_CHKSUM_OFFSET) { 3743 f2fs_warn(sbi, "using deprecated layout of large_nat_bitmap, " 3744 "please run fsck v1.13.0 or higher to repair, chksum_offset: %u, " 3745 "fixed with patch: \"f2fs-tools: relocate chksum_offset for large_nat_bitmap feature\"", 3746 le32_to_cpu(ckpt->checksum_offset)); 3747 return 1; 3748 } 3749 3750 nat_blocks = nat_segs << log_blocks_per_seg; 3751 nat_bits_bytes = nat_blocks / BITS_PER_BYTE; 3752 nat_bits_blocks = F2FS_BLK_ALIGN((nat_bits_bytes << 1) + 8); 3753 if (__is_set_ckpt_flags(ckpt, CP_NAT_BITS_FLAG) && 3754 (cp_payload + F2FS_CP_PACKS + 3755 NR_CURSEG_PERSIST_TYPE + nat_bits_blocks >= blocks_per_seg)) { 3756 f2fs_warn(sbi, "Insane cp_payload: %u, nat_bits_blocks: %u)", 3757 cp_payload, nat_bits_blocks); 3758 return 1; 3759 } 3760 3761 if (unlikely(f2fs_cp_error(sbi))) { 3762 f2fs_err(sbi, "A bug case: need to run fsck"); 3763 return 1; 3764 } 3765 return 0; 3766 } 3767 3768 static void init_sb_info(struct f2fs_sb_info *sbi) 3769 { 3770 struct f2fs_super_block *raw_super = sbi->raw_super; 3771 int i; 3772 3773 sbi->log_sectors_per_block = 3774 le32_to_cpu(raw_super->log_sectors_per_block); 3775 sbi->log_blocksize = le32_to_cpu(raw_super->log_blocksize); 3776 sbi->blocksize = BIT(sbi->log_blocksize); 3777 sbi->log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg); 3778 sbi->blocks_per_seg = BIT(sbi->log_blocks_per_seg); 3779 sbi->segs_per_sec = le32_to_cpu(raw_super->segs_per_sec); 3780 sbi->secs_per_zone = le32_to_cpu(raw_super->secs_per_zone); 3781 sbi->total_sections = le32_to_cpu(raw_super->section_count); 3782 sbi->total_node_count = 3783 (le32_to_cpu(raw_super->segment_count_nat) / 2) 3784 * sbi->blocks_per_seg * NAT_ENTRY_PER_BLOCK; 3785 F2FS_ROOT_INO(sbi) = le32_to_cpu(raw_super->root_ino); 3786 F2FS_NODE_INO(sbi) = le32_to_cpu(raw_super->node_ino); 3787 F2FS_META_INO(sbi) = le32_to_cpu(raw_super->meta_ino); 3788 sbi->cur_victim_sec = NULL_SECNO; 3789 sbi->gc_mode = GC_NORMAL; 3790 sbi->next_victim_seg[BG_GC] = NULL_SEGNO; 3791 sbi->next_victim_seg[FG_GC] = NULL_SEGNO; 3792 sbi->max_victim_search = DEF_MAX_VICTIM_SEARCH; 3793 sbi->migration_granularity = sbi->segs_per_sec; 3794 sbi->seq_file_ra_mul = MIN_RA_MUL; 3795 sbi->max_fragment_chunk = DEF_FRAGMENT_SIZE; 3796 sbi->max_fragment_hole = DEF_FRAGMENT_SIZE; 3797 spin_lock_init(&sbi->gc_remaining_trials_lock); 3798 atomic64_set(&sbi->current_atomic_write, 0); 3799 3800 sbi->dir_level = DEF_DIR_LEVEL; 3801 sbi->interval_time[CP_TIME] = DEF_CP_INTERVAL; 3802 sbi->interval_time[REQ_TIME] = DEF_IDLE_INTERVAL; 3803 sbi->interval_time[DISCARD_TIME] = DEF_IDLE_INTERVAL; 3804 sbi->interval_time[GC_TIME] = DEF_IDLE_INTERVAL; 3805 sbi->interval_time[DISABLE_TIME] = DEF_DISABLE_INTERVAL; 3806 sbi->interval_time[UMOUNT_DISCARD_TIMEOUT] = 3807 DEF_UMOUNT_DISCARD_TIMEOUT; 3808 clear_sbi_flag(sbi, SBI_NEED_FSCK); 3809 3810 for (i = 0; i < NR_COUNT_TYPE; i++) 3811 atomic_set(&sbi->nr_pages[i], 0); 3812 3813 for (i = 0; i < META; i++) 3814 atomic_set(&sbi->wb_sync_req[i], 0); 3815 3816 INIT_LIST_HEAD(&sbi->s_list); 3817 mutex_init(&sbi->umount_mutex); 3818 init_f2fs_rwsem(&sbi->io_order_lock); 3819 spin_lock_init(&sbi->cp_lock); 3820 3821 sbi->dirty_device = 0; 3822 spin_lock_init(&sbi->dev_lock); 3823 3824 init_f2fs_rwsem(&sbi->sb_lock); 3825 init_f2fs_rwsem(&sbi->pin_sem); 3826 } 3827 3828 static int init_percpu_info(struct f2fs_sb_info *sbi) 3829 { 3830 int err; 3831 3832 err = percpu_counter_init(&sbi->alloc_valid_block_count, 0, GFP_KERNEL); 3833 if (err) 3834 return err; 3835 3836 err = percpu_counter_init(&sbi->rf_node_block_count, 0, GFP_KERNEL); 3837 if (err) 3838 goto err_valid_block; 3839 3840 err = percpu_counter_init(&sbi->total_valid_inode_count, 0, 3841 GFP_KERNEL); 3842 if (err) 3843 goto err_node_block; 3844 return 0; 3845 3846 err_node_block: 3847 percpu_counter_destroy(&sbi->rf_node_block_count); 3848 err_valid_block: 3849 percpu_counter_destroy(&sbi->alloc_valid_block_count); 3850 return err; 3851 } 3852 3853 #ifdef CONFIG_BLK_DEV_ZONED 3854 3855 struct f2fs_report_zones_args { 3856 struct f2fs_sb_info *sbi; 3857 struct f2fs_dev_info *dev; 3858 }; 3859 3860 static int f2fs_report_zone_cb(struct blk_zone *zone, unsigned int idx, 3861 void *data) 3862 { 3863 struct f2fs_report_zones_args *rz_args = data; 3864 block_t unusable_blocks = (zone->len - zone->capacity) >> 3865 F2FS_LOG_SECTORS_PER_BLOCK; 3866 3867 if (zone->type == BLK_ZONE_TYPE_CONVENTIONAL) 3868 return 0; 3869 3870 set_bit(idx, rz_args->dev->blkz_seq); 3871 if (!rz_args->sbi->unusable_blocks_per_sec) { 3872 rz_args->sbi->unusable_blocks_per_sec = unusable_blocks; 3873 return 0; 3874 } 3875 if (rz_args->sbi->unusable_blocks_per_sec != unusable_blocks) { 3876 f2fs_err(rz_args->sbi, "F2FS supports single zone capacity\n"); 3877 return -EINVAL; 3878 } 3879 return 0; 3880 } 3881 3882 static int init_blkz_info(struct f2fs_sb_info *sbi, int devi) 3883 { 3884 struct block_device *bdev = FDEV(devi).bdev; 3885 sector_t nr_sectors = bdev_nr_sectors(bdev); 3886 struct f2fs_report_zones_args rep_zone_arg; 3887 u64 zone_sectors; 3888 int ret; 3889 3890 if (!f2fs_sb_has_blkzoned(sbi)) 3891 return 0; 3892 3893 zone_sectors = bdev_zone_sectors(bdev); 3894 if (!is_power_of_2(zone_sectors)) { 3895 f2fs_err(sbi, "F2FS does not support non power of 2 zone sizes\n"); 3896 return -EINVAL; 3897 } 3898 3899 if (sbi->blocks_per_blkz && sbi->blocks_per_blkz != 3900 SECTOR_TO_BLOCK(zone_sectors)) 3901 return -EINVAL; 3902 sbi->blocks_per_blkz = SECTOR_TO_BLOCK(zone_sectors); 3903 FDEV(devi).nr_blkz = div_u64(SECTOR_TO_BLOCK(nr_sectors), 3904 sbi->blocks_per_blkz); 3905 if (nr_sectors & (zone_sectors - 1)) 3906 FDEV(devi).nr_blkz++; 3907 3908 FDEV(devi).blkz_seq = f2fs_kvzalloc(sbi, 3909 BITS_TO_LONGS(FDEV(devi).nr_blkz) 3910 * sizeof(unsigned long), 3911 GFP_KERNEL); 3912 if (!FDEV(devi).blkz_seq) 3913 return -ENOMEM; 3914 3915 rep_zone_arg.sbi = sbi; 3916 rep_zone_arg.dev = &FDEV(devi); 3917 3918 ret = blkdev_report_zones(bdev, 0, BLK_ALL_ZONES, f2fs_report_zone_cb, 3919 &rep_zone_arg); 3920 if (ret < 0) 3921 return ret; 3922 return 0; 3923 } 3924 #endif 3925 3926 /* 3927 * Read f2fs raw super block. 3928 * Because we have two copies of super block, so read both of them 3929 * to get the first valid one. If any one of them is broken, we pass 3930 * them recovery flag back to the caller. 3931 */ 3932 static int read_raw_super_block(struct f2fs_sb_info *sbi, 3933 struct f2fs_super_block **raw_super, 3934 int *valid_super_block, int *recovery) 3935 { 3936 struct super_block *sb = sbi->sb; 3937 int block; 3938 struct buffer_head *bh; 3939 struct f2fs_super_block *super; 3940 int err = 0; 3941 3942 super = kzalloc(sizeof(struct f2fs_super_block), GFP_KERNEL); 3943 if (!super) 3944 return -ENOMEM; 3945 3946 for (block = 0; block < 2; block++) { 3947 bh = sb_bread(sb, block); 3948 if (!bh) { 3949 f2fs_err(sbi, "Unable to read %dth superblock", 3950 block + 1); 3951 err = -EIO; 3952 *recovery = 1; 3953 continue; 3954 } 3955 3956 /* sanity checking of raw super */ 3957 err = sanity_check_raw_super(sbi, bh); 3958 if (err) { 3959 f2fs_err(sbi, "Can't find valid F2FS filesystem in %dth superblock", 3960 block + 1); 3961 brelse(bh); 3962 *recovery = 1; 3963 continue; 3964 } 3965 3966 if (!*raw_super) { 3967 memcpy(super, bh->b_data + F2FS_SUPER_OFFSET, 3968 sizeof(*super)); 3969 *valid_super_block = block; 3970 *raw_super = super; 3971 } 3972 brelse(bh); 3973 } 3974 3975 /* No valid superblock */ 3976 if (!*raw_super) 3977 kfree(super); 3978 else 3979 err = 0; 3980 3981 return err; 3982 } 3983 3984 int f2fs_commit_super(struct f2fs_sb_info *sbi, bool recover) 3985 { 3986 struct buffer_head *bh; 3987 __u32 crc = 0; 3988 int err; 3989 3990 if ((recover && f2fs_readonly(sbi->sb)) || 3991 f2fs_hw_is_readonly(sbi)) { 3992 set_sbi_flag(sbi, SBI_NEED_SB_WRITE); 3993 return -EROFS; 3994 } 3995 3996 /* we should update superblock crc here */ 3997 if (!recover && f2fs_sb_has_sb_chksum(sbi)) { 3998 crc = f2fs_crc32(sbi, F2FS_RAW_SUPER(sbi), 3999 offsetof(struct f2fs_super_block, crc)); 4000 F2FS_RAW_SUPER(sbi)->crc = cpu_to_le32(crc); 4001 } 4002 4003 /* write back-up superblock first */ 4004 bh = sb_bread(sbi->sb, sbi->valid_super_block ? 0 : 1); 4005 if (!bh) 4006 return -EIO; 4007 err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi)); 4008 brelse(bh); 4009 4010 /* if we are in recovery path, skip writing valid superblock */ 4011 if (recover || err) 4012 return err; 4013 4014 /* write current valid superblock */ 4015 bh = sb_bread(sbi->sb, sbi->valid_super_block); 4016 if (!bh) 4017 return -EIO; 4018 err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi)); 4019 brelse(bh); 4020 return err; 4021 } 4022 4023 static void save_stop_reason(struct f2fs_sb_info *sbi, unsigned char reason) 4024 { 4025 unsigned long flags; 4026 4027 spin_lock_irqsave(&sbi->error_lock, flags); 4028 if (sbi->stop_reason[reason] < GENMASK(BITS_PER_BYTE - 1, 0)) 4029 sbi->stop_reason[reason]++; 4030 spin_unlock_irqrestore(&sbi->error_lock, flags); 4031 } 4032 4033 static void f2fs_record_stop_reason(struct f2fs_sb_info *sbi) 4034 { 4035 struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi); 4036 unsigned long flags; 4037 int err; 4038 4039 f2fs_down_write(&sbi->sb_lock); 4040 4041 spin_lock_irqsave(&sbi->error_lock, flags); 4042 if (sbi->error_dirty) { 4043 memcpy(F2FS_RAW_SUPER(sbi)->s_errors, sbi->errors, 4044 MAX_F2FS_ERRORS); 4045 sbi->error_dirty = false; 4046 } 4047 memcpy(raw_super->s_stop_reason, sbi->stop_reason, MAX_STOP_REASON); 4048 spin_unlock_irqrestore(&sbi->error_lock, flags); 4049 4050 err = f2fs_commit_super(sbi, false); 4051 4052 f2fs_up_write(&sbi->sb_lock); 4053 if (err) 4054 f2fs_err(sbi, "f2fs_commit_super fails to record err:%d", err); 4055 } 4056 4057 void f2fs_save_errors(struct f2fs_sb_info *sbi, unsigned char flag) 4058 { 4059 unsigned long flags; 4060 4061 spin_lock_irqsave(&sbi->error_lock, flags); 4062 if (!test_bit(flag, (unsigned long *)sbi->errors)) { 4063 set_bit(flag, (unsigned long *)sbi->errors); 4064 sbi->error_dirty = true; 4065 } 4066 spin_unlock_irqrestore(&sbi->error_lock, flags); 4067 } 4068 4069 static bool f2fs_update_errors(struct f2fs_sb_info *sbi) 4070 { 4071 unsigned long flags; 4072 bool need_update = false; 4073 4074 spin_lock_irqsave(&sbi->error_lock, flags); 4075 if (sbi->error_dirty) { 4076 memcpy(F2FS_RAW_SUPER(sbi)->s_errors, sbi->errors, 4077 MAX_F2FS_ERRORS); 4078 sbi->error_dirty = false; 4079 need_update = true; 4080 } 4081 spin_unlock_irqrestore(&sbi->error_lock, flags); 4082 4083 return need_update; 4084 } 4085 4086 static void f2fs_record_errors(struct f2fs_sb_info *sbi, unsigned char error) 4087 { 4088 int err; 4089 4090 f2fs_down_write(&sbi->sb_lock); 4091 4092 if (!f2fs_update_errors(sbi)) 4093 goto out_unlock; 4094 4095 err = f2fs_commit_super(sbi, false); 4096 if (err) 4097 f2fs_err(sbi, "f2fs_commit_super fails to record errors:%u, err:%d", 4098 error, err); 4099 out_unlock: 4100 f2fs_up_write(&sbi->sb_lock); 4101 } 4102 4103 void f2fs_handle_error(struct f2fs_sb_info *sbi, unsigned char error) 4104 { 4105 f2fs_save_errors(sbi, error); 4106 f2fs_record_errors(sbi, error); 4107 } 4108 4109 void f2fs_handle_error_async(struct f2fs_sb_info *sbi, unsigned char error) 4110 { 4111 f2fs_save_errors(sbi, error); 4112 4113 if (!sbi->error_dirty) 4114 return; 4115 if (!test_bit(error, (unsigned long *)sbi->errors)) 4116 return; 4117 schedule_work(&sbi->s_error_work); 4118 } 4119 4120 static bool system_going_down(void) 4121 { 4122 return system_state == SYSTEM_HALT || system_state == SYSTEM_POWER_OFF 4123 || system_state == SYSTEM_RESTART; 4124 } 4125 4126 void f2fs_handle_critical_error(struct f2fs_sb_info *sbi, unsigned char reason, 4127 bool irq_context) 4128 { 4129 struct super_block *sb = sbi->sb; 4130 bool shutdown = reason == STOP_CP_REASON_SHUTDOWN; 4131 bool continue_fs = !shutdown && 4132 F2FS_OPTION(sbi).errors == MOUNT_ERRORS_CONTINUE; 4133 4134 set_ckpt_flags(sbi, CP_ERROR_FLAG); 4135 4136 if (!f2fs_hw_is_readonly(sbi)) { 4137 save_stop_reason(sbi, reason); 4138 4139 if (irq_context && !shutdown) 4140 schedule_work(&sbi->s_error_work); 4141 else 4142 f2fs_record_stop_reason(sbi); 4143 } 4144 4145 /* 4146 * We force ERRORS_RO behavior when system is rebooting. Otherwise we 4147 * could panic during 'reboot -f' as the underlying device got already 4148 * disabled. 4149 */ 4150 if (F2FS_OPTION(sbi).errors == MOUNT_ERRORS_PANIC && 4151 !shutdown && !system_going_down() && 4152 !is_sbi_flag_set(sbi, SBI_IS_SHUTDOWN)) 4153 panic("F2FS-fs (device %s): panic forced after error\n", 4154 sb->s_id); 4155 4156 if (shutdown) 4157 set_sbi_flag(sbi, SBI_IS_SHUTDOWN); 4158 4159 /* continue filesystem operators if errors=continue */ 4160 if (continue_fs || f2fs_readonly(sb)) 4161 return; 4162 4163 f2fs_warn(sbi, "Remounting filesystem read-only"); 4164 /* 4165 * Make sure updated value of ->s_mount_flags will be visible before 4166 * ->s_flags update 4167 */ 4168 smp_wmb(); 4169 sb->s_flags |= SB_RDONLY; 4170 } 4171 4172 static void f2fs_record_error_work(struct work_struct *work) 4173 { 4174 struct f2fs_sb_info *sbi = container_of(work, 4175 struct f2fs_sb_info, s_error_work); 4176 4177 f2fs_record_stop_reason(sbi); 4178 } 4179 4180 static int f2fs_scan_devices(struct f2fs_sb_info *sbi) 4181 { 4182 struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi); 4183 unsigned int max_devices = MAX_DEVICES; 4184 unsigned int logical_blksize; 4185 blk_mode_t mode = sb_open_mode(sbi->sb->s_flags); 4186 int i; 4187 4188 /* Initialize single device information */ 4189 if (!RDEV(0).path[0]) { 4190 if (!bdev_is_zoned(sbi->sb->s_bdev)) 4191 return 0; 4192 max_devices = 1; 4193 } 4194 4195 /* 4196 * Initialize multiple devices information, or single 4197 * zoned block device information. 4198 */ 4199 sbi->devs = f2fs_kzalloc(sbi, 4200 array_size(max_devices, 4201 sizeof(struct f2fs_dev_info)), 4202 GFP_KERNEL); 4203 if (!sbi->devs) 4204 return -ENOMEM; 4205 4206 logical_blksize = bdev_logical_block_size(sbi->sb->s_bdev); 4207 sbi->aligned_blksize = true; 4208 4209 for (i = 0; i < max_devices; i++) { 4210 if (i == 0) 4211 FDEV(0).bdev_handle = sbi->sb->s_bdev_handle; 4212 else if (!RDEV(i).path[0]) 4213 break; 4214 4215 if (max_devices > 1) { 4216 /* Multi-device mount */ 4217 memcpy(FDEV(i).path, RDEV(i).path, MAX_PATH_LEN); 4218 FDEV(i).total_segments = 4219 le32_to_cpu(RDEV(i).total_segments); 4220 if (i == 0) { 4221 FDEV(i).start_blk = 0; 4222 FDEV(i).end_blk = FDEV(i).start_blk + 4223 (FDEV(i).total_segments << 4224 sbi->log_blocks_per_seg) - 1 + 4225 le32_to_cpu(raw_super->segment0_blkaddr); 4226 } else { 4227 FDEV(i).start_blk = FDEV(i - 1).end_blk + 1; 4228 FDEV(i).end_blk = FDEV(i).start_blk + 4229 (FDEV(i).total_segments << 4230 sbi->log_blocks_per_seg) - 1; 4231 FDEV(i).bdev_handle = bdev_open_by_path( 4232 FDEV(i).path, mode, sbi->sb, NULL); 4233 } 4234 } 4235 if (IS_ERR(FDEV(i).bdev_handle)) 4236 return PTR_ERR(FDEV(i).bdev_handle); 4237 4238 FDEV(i).bdev = FDEV(i).bdev_handle->bdev; 4239 /* to release errored devices */ 4240 sbi->s_ndevs = i + 1; 4241 4242 if (logical_blksize != bdev_logical_block_size(FDEV(i).bdev)) 4243 sbi->aligned_blksize = false; 4244 4245 #ifdef CONFIG_BLK_DEV_ZONED 4246 if (bdev_zoned_model(FDEV(i).bdev) == BLK_ZONED_HM && 4247 !f2fs_sb_has_blkzoned(sbi)) { 4248 f2fs_err(sbi, "Zoned block device feature not enabled"); 4249 return -EINVAL; 4250 } 4251 if (bdev_zoned_model(FDEV(i).bdev) != BLK_ZONED_NONE) { 4252 if (init_blkz_info(sbi, i)) { 4253 f2fs_err(sbi, "Failed to initialize F2FS blkzone information"); 4254 return -EINVAL; 4255 } 4256 if (max_devices == 1) 4257 break; 4258 f2fs_info(sbi, "Mount Device [%2d]: %20s, %8u, %8x - %8x (zone: %s)", 4259 i, FDEV(i).path, 4260 FDEV(i).total_segments, 4261 FDEV(i).start_blk, FDEV(i).end_blk, 4262 bdev_zoned_model(FDEV(i).bdev) == BLK_ZONED_HA ? 4263 "Host-aware" : "Host-managed"); 4264 continue; 4265 } 4266 #endif 4267 f2fs_info(sbi, "Mount Device [%2d]: %20s, %8u, %8x - %8x", 4268 i, FDEV(i).path, 4269 FDEV(i).total_segments, 4270 FDEV(i).start_blk, FDEV(i).end_blk); 4271 } 4272 f2fs_info(sbi, 4273 "IO Block Size: %8ld KB", F2FS_IO_SIZE_KB(sbi)); 4274 return 0; 4275 } 4276 4277 static int f2fs_setup_casefold(struct f2fs_sb_info *sbi) 4278 { 4279 #if IS_ENABLED(CONFIG_UNICODE) 4280 if (f2fs_sb_has_casefold(sbi) && !sbi->sb->s_encoding) { 4281 const struct f2fs_sb_encodings *encoding_info; 4282 struct unicode_map *encoding; 4283 __u16 encoding_flags; 4284 4285 encoding_info = f2fs_sb_read_encoding(sbi->raw_super); 4286 if (!encoding_info) { 4287 f2fs_err(sbi, 4288 "Encoding requested by superblock is unknown"); 4289 return -EINVAL; 4290 } 4291 4292 encoding_flags = le16_to_cpu(sbi->raw_super->s_encoding_flags); 4293 encoding = utf8_load(encoding_info->version); 4294 if (IS_ERR(encoding)) { 4295 f2fs_err(sbi, 4296 "can't mount with superblock charset: %s-%u.%u.%u " 4297 "not supported by the kernel. flags: 0x%x.", 4298 encoding_info->name, 4299 unicode_major(encoding_info->version), 4300 unicode_minor(encoding_info->version), 4301 unicode_rev(encoding_info->version), 4302 encoding_flags); 4303 return PTR_ERR(encoding); 4304 } 4305 f2fs_info(sbi, "Using encoding defined by superblock: " 4306 "%s-%u.%u.%u with flags 0x%hx", encoding_info->name, 4307 unicode_major(encoding_info->version), 4308 unicode_minor(encoding_info->version), 4309 unicode_rev(encoding_info->version), 4310 encoding_flags); 4311 4312 sbi->sb->s_encoding = encoding; 4313 sbi->sb->s_encoding_flags = encoding_flags; 4314 } 4315 #else 4316 if (f2fs_sb_has_casefold(sbi)) { 4317 f2fs_err(sbi, "Filesystem with casefold feature cannot be mounted without CONFIG_UNICODE"); 4318 return -EINVAL; 4319 } 4320 #endif 4321 return 0; 4322 } 4323 4324 static void f2fs_tuning_parameters(struct f2fs_sb_info *sbi) 4325 { 4326 /* adjust parameters according to the volume size */ 4327 if (MAIN_SEGS(sbi) <= SMALL_VOLUME_SEGMENTS) { 4328 if (f2fs_block_unit_discard(sbi)) 4329 SM_I(sbi)->dcc_info->discard_granularity = 4330 MIN_DISCARD_GRANULARITY; 4331 if (!f2fs_lfs_mode(sbi)) 4332 SM_I(sbi)->ipu_policy = BIT(F2FS_IPU_FORCE) | 4333 BIT(F2FS_IPU_HONOR_OPU_WRITE); 4334 } 4335 4336 sbi->readdir_ra = true; 4337 } 4338 4339 static int f2fs_fill_super(struct super_block *sb, void *data, int silent) 4340 { 4341 struct f2fs_sb_info *sbi; 4342 struct f2fs_super_block *raw_super; 4343 struct inode *root; 4344 int err; 4345 bool skip_recovery = false, need_fsck = false; 4346 char *options = NULL; 4347 int recovery, i, valid_super_block; 4348 struct curseg_info *seg_i; 4349 int retry_cnt = 1; 4350 #ifdef CONFIG_QUOTA 4351 bool quota_enabled = false; 4352 #endif 4353 4354 try_onemore: 4355 err = -EINVAL; 4356 raw_super = NULL; 4357 valid_super_block = -1; 4358 recovery = 0; 4359 4360 /* allocate memory for f2fs-specific super block info */ 4361 sbi = kzalloc(sizeof(struct f2fs_sb_info), GFP_KERNEL); 4362 if (!sbi) 4363 return -ENOMEM; 4364 4365 sbi->sb = sb; 4366 4367 /* initialize locks within allocated memory */ 4368 init_f2fs_rwsem(&sbi->gc_lock); 4369 mutex_init(&sbi->writepages); 4370 init_f2fs_rwsem(&sbi->cp_global_sem); 4371 init_f2fs_rwsem(&sbi->node_write); 4372 init_f2fs_rwsem(&sbi->node_change); 4373 spin_lock_init(&sbi->stat_lock); 4374 init_f2fs_rwsem(&sbi->cp_rwsem); 4375 init_f2fs_rwsem(&sbi->quota_sem); 4376 init_waitqueue_head(&sbi->cp_wait); 4377 spin_lock_init(&sbi->error_lock); 4378 4379 for (i = 0; i < NR_INODE_TYPE; i++) { 4380 INIT_LIST_HEAD(&sbi->inode_list[i]); 4381 spin_lock_init(&sbi->inode_lock[i]); 4382 } 4383 mutex_init(&sbi->flush_lock); 4384 4385 /* Load the checksum driver */ 4386 sbi->s_chksum_driver = crypto_alloc_shash("crc32", 0, 0); 4387 if (IS_ERR(sbi->s_chksum_driver)) { 4388 f2fs_err(sbi, "Cannot load crc32 driver."); 4389 err = PTR_ERR(sbi->s_chksum_driver); 4390 sbi->s_chksum_driver = NULL; 4391 goto free_sbi; 4392 } 4393 4394 /* set a block size */ 4395 if (unlikely(!sb_set_blocksize(sb, F2FS_BLKSIZE))) { 4396 f2fs_err(sbi, "unable to set blocksize"); 4397 goto free_sbi; 4398 } 4399 4400 err = read_raw_super_block(sbi, &raw_super, &valid_super_block, 4401 &recovery); 4402 if (err) 4403 goto free_sbi; 4404 4405 sb->s_fs_info = sbi; 4406 sbi->raw_super = raw_super; 4407 4408 INIT_WORK(&sbi->s_error_work, f2fs_record_error_work); 4409 memcpy(sbi->errors, raw_super->s_errors, MAX_F2FS_ERRORS); 4410 memcpy(sbi->stop_reason, raw_super->s_stop_reason, MAX_STOP_REASON); 4411 4412 /* precompute checksum seed for metadata */ 4413 if (f2fs_sb_has_inode_chksum(sbi)) 4414 sbi->s_chksum_seed = f2fs_chksum(sbi, ~0, raw_super->uuid, 4415 sizeof(raw_super->uuid)); 4416 4417 default_options(sbi, false); 4418 /* parse mount options */ 4419 options = kstrdup((const char *)data, GFP_KERNEL); 4420 if (data && !options) { 4421 err = -ENOMEM; 4422 goto free_sb_buf; 4423 } 4424 4425 err = parse_options(sb, options, false); 4426 if (err) 4427 goto free_options; 4428 4429 sb->s_maxbytes = max_file_blocks(NULL) << 4430 le32_to_cpu(raw_super->log_blocksize); 4431 sb->s_max_links = F2FS_LINK_MAX; 4432 4433 err = f2fs_setup_casefold(sbi); 4434 if (err) 4435 goto free_options; 4436 4437 #ifdef CONFIG_QUOTA 4438 sb->dq_op = &f2fs_quota_operations; 4439 sb->s_qcop = &f2fs_quotactl_ops; 4440 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ; 4441 4442 if (f2fs_sb_has_quota_ino(sbi)) { 4443 for (i = 0; i < MAXQUOTAS; i++) { 4444 if (f2fs_qf_ino(sbi->sb, i)) 4445 sbi->nquota_files++; 4446 } 4447 } 4448 #endif 4449 4450 sb->s_op = &f2fs_sops; 4451 #ifdef CONFIG_FS_ENCRYPTION 4452 sb->s_cop = &f2fs_cryptops; 4453 #endif 4454 #ifdef CONFIG_FS_VERITY 4455 sb->s_vop = &f2fs_verityops; 4456 #endif 4457 sb->s_xattr = f2fs_xattr_handlers; 4458 sb->s_export_op = &f2fs_export_ops; 4459 sb->s_magic = F2FS_SUPER_MAGIC; 4460 sb->s_time_gran = 1; 4461 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) | 4462 (test_opt(sbi, POSIX_ACL) ? SB_POSIXACL : 0); 4463 memcpy(&sb->s_uuid, raw_super->uuid, sizeof(raw_super->uuid)); 4464 sb->s_iflags |= SB_I_CGROUPWB; 4465 4466 /* init f2fs-specific super block info */ 4467 sbi->valid_super_block = valid_super_block; 4468 4469 /* disallow all the data/node/meta page writes */ 4470 set_sbi_flag(sbi, SBI_POR_DOING); 4471 4472 err = f2fs_init_write_merge_io(sbi); 4473 if (err) 4474 goto free_bio_info; 4475 4476 init_sb_info(sbi); 4477 4478 err = f2fs_init_iostat(sbi); 4479 if (err) 4480 goto free_bio_info; 4481 4482 err = init_percpu_info(sbi); 4483 if (err) 4484 goto free_iostat; 4485 4486 if (F2FS_IO_ALIGNED(sbi)) { 4487 sbi->write_io_dummy = 4488 mempool_create_page_pool(2 * (F2FS_IO_SIZE(sbi) - 1), 0); 4489 if (!sbi->write_io_dummy) { 4490 err = -ENOMEM; 4491 goto free_percpu; 4492 } 4493 } 4494 4495 /* init per sbi slab cache */ 4496 err = f2fs_init_xattr_caches(sbi); 4497 if (err) 4498 goto free_io_dummy; 4499 err = f2fs_init_page_array_cache(sbi); 4500 if (err) 4501 goto free_xattr_cache; 4502 4503 /* get an inode for meta space */ 4504 sbi->meta_inode = f2fs_iget(sb, F2FS_META_INO(sbi)); 4505 if (IS_ERR(sbi->meta_inode)) { 4506 f2fs_err(sbi, "Failed to read F2FS meta data inode"); 4507 err = PTR_ERR(sbi->meta_inode); 4508 goto free_page_array_cache; 4509 } 4510 4511 err = f2fs_get_valid_checkpoint(sbi); 4512 if (err) { 4513 f2fs_err(sbi, "Failed to get valid F2FS checkpoint"); 4514 goto free_meta_inode; 4515 } 4516 4517 if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_QUOTA_NEED_FSCK_FLAG)) 4518 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR); 4519 if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_DISABLED_QUICK_FLAG)) { 4520 set_sbi_flag(sbi, SBI_CP_DISABLED_QUICK); 4521 sbi->interval_time[DISABLE_TIME] = DEF_DISABLE_QUICK_INTERVAL; 4522 } 4523 4524 if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_FSCK_FLAG)) 4525 set_sbi_flag(sbi, SBI_NEED_FSCK); 4526 4527 /* Initialize device list */ 4528 err = f2fs_scan_devices(sbi); 4529 if (err) { 4530 f2fs_err(sbi, "Failed to find devices"); 4531 goto free_devices; 4532 } 4533 4534 err = f2fs_init_post_read_wq(sbi); 4535 if (err) { 4536 f2fs_err(sbi, "Failed to initialize post read workqueue"); 4537 goto free_devices; 4538 } 4539 4540 sbi->total_valid_node_count = 4541 le32_to_cpu(sbi->ckpt->valid_node_count); 4542 percpu_counter_set(&sbi->total_valid_inode_count, 4543 le32_to_cpu(sbi->ckpt->valid_inode_count)); 4544 sbi->user_block_count = le64_to_cpu(sbi->ckpt->user_block_count); 4545 sbi->total_valid_block_count = 4546 le64_to_cpu(sbi->ckpt->valid_block_count); 4547 sbi->last_valid_block_count = sbi->total_valid_block_count; 4548 sbi->reserved_blocks = 0; 4549 sbi->current_reserved_blocks = 0; 4550 limit_reserve_root(sbi); 4551 adjust_unusable_cap_perc(sbi); 4552 4553 f2fs_init_extent_cache_info(sbi); 4554 4555 f2fs_init_ino_entry_info(sbi); 4556 4557 f2fs_init_fsync_node_info(sbi); 4558 4559 /* setup checkpoint request control and start checkpoint issue thread */ 4560 f2fs_init_ckpt_req_control(sbi); 4561 if (!f2fs_readonly(sb) && !test_opt(sbi, DISABLE_CHECKPOINT) && 4562 test_opt(sbi, MERGE_CHECKPOINT)) { 4563 err = f2fs_start_ckpt_thread(sbi); 4564 if (err) { 4565 f2fs_err(sbi, 4566 "Failed to start F2FS issue_checkpoint_thread (%d)", 4567 err); 4568 goto stop_ckpt_thread; 4569 } 4570 } 4571 4572 /* setup f2fs internal modules */ 4573 err = f2fs_build_segment_manager(sbi); 4574 if (err) { 4575 f2fs_err(sbi, "Failed to initialize F2FS segment manager (%d)", 4576 err); 4577 goto free_sm; 4578 } 4579 err = f2fs_build_node_manager(sbi); 4580 if (err) { 4581 f2fs_err(sbi, "Failed to initialize F2FS node manager (%d)", 4582 err); 4583 goto free_nm; 4584 } 4585 4586 err = adjust_reserved_segment(sbi); 4587 if (err) 4588 goto free_nm; 4589 4590 /* For write statistics */ 4591 sbi->sectors_written_start = f2fs_get_sectors_written(sbi); 4592 4593 /* Read accumulated write IO statistics if exists */ 4594 seg_i = CURSEG_I(sbi, CURSEG_HOT_NODE); 4595 if (__exist_node_summaries(sbi)) 4596 sbi->kbytes_written = 4597 le64_to_cpu(seg_i->journal->info.kbytes_written); 4598 4599 f2fs_build_gc_manager(sbi); 4600 4601 err = f2fs_build_stats(sbi); 4602 if (err) 4603 goto free_nm; 4604 4605 /* get an inode for node space */ 4606 sbi->node_inode = f2fs_iget(sb, F2FS_NODE_INO(sbi)); 4607 if (IS_ERR(sbi->node_inode)) { 4608 f2fs_err(sbi, "Failed to read node inode"); 4609 err = PTR_ERR(sbi->node_inode); 4610 goto free_stats; 4611 } 4612 4613 /* read root inode and dentry */ 4614 root = f2fs_iget(sb, F2FS_ROOT_INO(sbi)); 4615 if (IS_ERR(root)) { 4616 f2fs_err(sbi, "Failed to read root inode"); 4617 err = PTR_ERR(root); 4618 goto free_node_inode; 4619 } 4620 if (!S_ISDIR(root->i_mode) || !root->i_blocks || 4621 !root->i_size || !root->i_nlink) { 4622 iput(root); 4623 err = -EINVAL; 4624 goto free_node_inode; 4625 } 4626 4627 sb->s_root = d_make_root(root); /* allocate root dentry */ 4628 if (!sb->s_root) { 4629 err = -ENOMEM; 4630 goto free_node_inode; 4631 } 4632 4633 err = f2fs_init_compress_inode(sbi); 4634 if (err) 4635 goto free_root_inode; 4636 4637 err = f2fs_register_sysfs(sbi); 4638 if (err) 4639 goto free_compress_inode; 4640 4641 #ifdef CONFIG_QUOTA 4642 /* Enable quota usage during mount */ 4643 if (f2fs_sb_has_quota_ino(sbi) && !f2fs_readonly(sb)) { 4644 err = f2fs_enable_quotas(sb); 4645 if (err) 4646 f2fs_err(sbi, "Cannot turn on quotas: error %d", err); 4647 } 4648 4649 quota_enabled = f2fs_recover_quota_begin(sbi); 4650 #endif 4651 /* if there are any orphan inodes, free them */ 4652 err = f2fs_recover_orphan_inodes(sbi); 4653 if (err) 4654 goto free_meta; 4655 4656 if (unlikely(is_set_ckpt_flags(sbi, CP_DISABLED_FLAG))) 4657 goto reset_checkpoint; 4658 4659 /* recover fsynced data */ 4660 if (!test_opt(sbi, DISABLE_ROLL_FORWARD) && 4661 !test_opt(sbi, NORECOVERY)) { 4662 /* 4663 * mount should be failed, when device has readonly mode, and 4664 * previous checkpoint was not done by clean system shutdown. 4665 */ 4666 if (f2fs_hw_is_readonly(sbi)) { 4667 if (!is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG)) { 4668 err = f2fs_recover_fsync_data(sbi, true); 4669 if (err > 0) { 4670 err = -EROFS; 4671 f2fs_err(sbi, "Need to recover fsync data, but " 4672 "write access unavailable, please try " 4673 "mount w/ disable_roll_forward or norecovery"); 4674 } 4675 if (err < 0) 4676 goto free_meta; 4677 } 4678 f2fs_info(sbi, "write access unavailable, skipping recovery"); 4679 goto reset_checkpoint; 4680 } 4681 4682 if (need_fsck) 4683 set_sbi_flag(sbi, SBI_NEED_FSCK); 4684 4685 if (skip_recovery) 4686 goto reset_checkpoint; 4687 4688 err = f2fs_recover_fsync_data(sbi, false); 4689 if (err < 0) { 4690 if (err != -ENOMEM) 4691 skip_recovery = true; 4692 need_fsck = true; 4693 f2fs_err(sbi, "Cannot recover all fsync data errno=%d", 4694 err); 4695 goto free_meta; 4696 } 4697 } else { 4698 err = f2fs_recover_fsync_data(sbi, true); 4699 4700 if (!f2fs_readonly(sb) && err > 0) { 4701 err = -EINVAL; 4702 f2fs_err(sbi, "Need to recover fsync data"); 4703 goto free_meta; 4704 } 4705 } 4706 4707 #ifdef CONFIG_QUOTA 4708 f2fs_recover_quota_end(sbi, quota_enabled); 4709 #endif 4710 4711 /* 4712 * If the f2fs is not readonly and fsync data recovery succeeds, 4713 * check zoned block devices' write pointer consistency. 4714 */ 4715 if (!err && !f2fs_readonly(sb) && f2fs_sb_has_blkzoned(sbi)) { 4716 err = f2fs_check_write_pointer(sbi); 4717 if (err) 4718 goto free_meta; 4719 } 4720 4721 reset_checkpoint: 4722 f2fs_init_inmem_curseg(sbi); 4723 4724 /* f2fs_recover_fsync_data() cleared this already */ 4725 clear_sbi_flag(sbi, SBI_POR_DOING); 4726 4727 if (test_opt(sbi, DISABLE_CHECKPOINT)) { 4728 err = f2fs_disable_checkpoint(sbi); 4729 if (err) 4730 goto sync_free_meta; 4731 } else if (is_set_ckpt_flags(sbi, CP_DISABLED_FLAG)) { 4732 f2fs_enable_checkpoint(sbi); 4733 } 4734 4735 /* 4736 * If filesystem is not mounted as read-only then 4737 * do start the gc_thread. 4738 */ 4739 if ((F2FS_OPTION(sbi).bggc_mode != BGGC_MODE_OFF || 4740 test_opt(sbi, GC_MERGE)) && !f2fs_readonly(sb)) { 4741 /* After POR, we can run background GC thread.*/ 4742 err = f2fs_start_gc_thread(sbi); 4743 if (err) 4744 goto sync_free_meta; 4745 } 4746 kvfree(options); 4747 4748 /* recover broken superblock */ 4749 if (recovery) { 4750 err = f2fs_commit_super(sbi, true); 4751 f2fs_info(sbi, "Try to recover %dth superblock, ret: %d", 4752 sbi->valid_super_block ? 1 : 2, err); 4753 } 4754 4755 f2fs_join_shrinker(sbi); 4756 4757 f2fs_tuning_parameters(sbi); 4758 4759 f2fs_notice(sbi, "Mounted with checkpoint version = %llx", 4760 cur_cp_version(F2FS_CKPT(sbi))); 4761 f2fs_update_time(sbi, CP_TIME); 4762 f2fs_update_time(sbi, REQ_TIME); 4763 clear_sbi_flag(sbi, SBI_CP_DISABLED_QUICK); 4764 return 0; 4765 4766 sync_free_meta: 4767 /* safe to flush all the data */ 4768 sync_filesystem(sbi->sb); 4769 retry_cnt = 0; 4770 4771 free_meta: 4772 #ifdef CONFIG_QUOTA 4773 f2fs_truncate_quota_inode_pages(sb); 4774 if (f2fs_sb_has_quota_ino(sbi) && !f2fs_readonly(sb)) 4775 f2fs_quota_off_umount(sbi->sb); 4776 #endif 4777 /* 4778 * Some dirty meta pages can be produced by f2fs_recover_orphan_inodes() 4779 * failed by EIO. Then, iput(node_inode) can trigger balance_fs_bg() 4780 * followed by f2fs_write_checkpoint() through f2fs_write_node_pages(), which 4781 * falls into an infinite loop in f2fs_sync_meta_pages(). 4782 */ 4783 truncate_inode_pages_final(META_MAPPING(sbi)); 4784 /* evict some inodes being cached by GC */ 4785 evict_inodes(sb); 4786 f2fs_unregister_sysfs(sbi); 4787 free_compress_inode: 4788 f2fs_destroy_compress_inode(sbi); 4789 free_root_inode: 4790 dput(sb->s_root); 4791 sb->s_root = NULL; 4792 free_node_inode: 4793 f2fs_release_ino_entry(sbi, true); 4794 truncate_inode_pages_final(NODE_MAPPING(sbi)); 4795 iput(sbi->node_inode); 4796 sbi->node_inode = NULL; 4797 free_stats: 4798 f2fs_destroy_stats(sbi); 4799 free_nm: 4800 /* stop discard thread before destroying node manager */ 4801 f2fs_stop_discard_thread(sbi); 4802 f2fs_destroy_node_manager(sbi); 4803 free_sm: 4804 f2fs_destroy_segment_manager(sbi); 4805 stop_ckpt_thread: 4806 f2fs_stop_ckpt_thread(sbi); 4807 /* flush s_error_work before sbi destroy */ 4808 flush_work(&sbi->s_error_work); 4809 f2fs_destroy_post_read_wq(sbi); 4810 free_devices: 4811 destroy_device_list(sbi); 4812 kvfree(sbi->ckpt); 4813 free_meta_inode: 4814 make_bad_inode(sbi->meta_inode); 4815 iput(sbi->meta_inode); 4816 sbi->meta_inode = NULL; 4817 free_page_array_cache: 4818 f2fs_destroy_page_array_cache(sbi); 4819 free_xattr_cache: 4820 f2fs_destroy_xattr_caches(sbi); 4821 free_io_dummy: 4822 mempool_destroy(sbi->write_io_dummy); 4823 free_percpu: 4824 destroy_percpu_info(sbi); 4825 free_iostat: 4826 f2fs_destroy_iostat(sbi); 4827 free_bio_info: 4828 for (i = 0; i < NR_PAGE_TYPE; i++) 4829 kvfree(sbi->write_io[i]); 4830 4831 #if IS_ENABLED(CONFIG_UNICODE) 4832 utf8_unload(sb->s_encoding); 4833 sb->s_encoding = NULL; 4834 #endif 4835 free_options: 4836 #ifdef CONFIG_QUOTA 4837 for (i = 0; i < MAXQUOTAS; i++) 4838 kfree(F2FS_OPTION(sbi).s_qf_names[i]); 4839 #endif 4840 fscrypt_free_dummy_policy(&F2FS_OPTION(sbi).dummy_enc_policy); 4841 kvfree(options); 4842 free_sb_buf: 4843 kfree(raw_super); 4844 free_sbi: 4845 if (sbi->s_chksum_driver) 4846 crypto_free_shash(sbi->s_chksum_driver); 4847 kfree(sbi); 4848 4849 /* give only one another chance */ 4850 if (retry_cnt > 0 && skip_recovery) { 4851 retry_cnt--; 4852 shrink_dcache_sb(sb); 4853 goto try_onemore; 4854 } 4855 return err; 4856 } 4857 4858 static struct dentry *f2fs_mount(struct file_system_type *fs_type, int flags, 4859 const char *dev_name, void *data) 4860 { 4861 return mount_bdev(fs_type, flags, dev_name, data, f2fs_fill_super); 4862 } 4863 4864 static void kill_f2fs_super(struct super_block *sb) 4865 { 4866 if (sb->s_root) { 4867 struct f2fs_sb_info *sbi = F2FS_SB(sb); 4868 4869 set_sbi_flag(sbi, SBI_IS_CLOSE); 4870 f2fs_stop_gc_thread(sbi); 4871 f2fs_stop_discard_thread(sbi); 4872 4873 #ifdef CONFIG_F2FS_FS_COMPRESSION 4874 /* 4875 * latter evict_inode() can bypass checking and invalidating 4876 * compress inode cache. 4877 */ 4878 if (test_opt(sbi, COMPRESS_CACHE)) 4879 truncate_inode_pages_final(COMPRESS_MAPPING(sbi)); 4880 #endif 4881 4882 if (is_sbi_flag_set(sbi, SBI_IS_DIRTY) || 4883 !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG)) { 4884 struct cp_control cpc = { 4885 .reason = CP_UMOUNT, 4886 }; 4887 stat_inc_cp_call_count(sbi, TOTAL_CALL); 4888 f2fs_write_checkpoint(sbi, &cpc); 4889 } 4890 4891 if (is_sbi_flag_set(sbi, SBI_IS_RECOVERED) && f2fs_readonly(sb)) 4892 sb->s_flags &= ~SB_RDONLY; 4893 } 4894 kill_block_super(sb); 4895 } 4896 4897 static struct file_system_type f2fs_fs_type = { 4898 .owner = THIS_MODULE, 4899 .name = "f2fs", 4900 .mount = f2fs_mount, 4901 .kill_sb = kill_f2fs_super, 4902 .fs_flags = FS_REQUIRES_DEV | FS_ALLOW_IDMAP, 4903 }; 4904 MODULE_ALIAS_FS("f2fs"); 4905 4906 static int __init init_inodecache(void) 4907 { 4908 f2fs_inode_cachep = kmem_cache_create("f2fs_inode_cache", 4909 sizeof(struct f2fs_inode_info), 0, 4910 SLAB_RECLAIM_ACCOUNT|SLAB_ACCOUNT, NULL); 4911 return f2fs_inode_cachep ? 0 : -ENOMEM; 4912 } 4913 4914 static void destroy_inodecache(void) 4915 { 4916 /* 4917 * Make sure all delayed rcu free inodes are flushed before we 4918 * destroy cache. 4919 */ 4920 rcu_barrier(); 4921 kmem_cache_destroy(f2fs_inode_cachep); 4922 } 4923 4924 static int __init init_f2fs_fs(void) 4925 { 4926 int err; 4927 4928 if (PAGE_SIZE != F2FS_BLKSIZE) { 4929 printk("F2FS not supported on PAGE_SIZE(%lu) != %d\n", 4930 PAGE_SIZE, F2FS_BLKSIZE); 4931 return -EINVAL; 4932 } 4933 4934 err = init_inodecache(); 4935 if (err) 4936 goto fail; 4937 err = f2fs_create_node_manager_caches(); 4938 if (err) 4939 goto free_inodecache; 4940 err = f2fs_create_segment_manager_caches(); 4941 if (err) 4942 goto free_node_manager_caches; 4943 err = f2fs_create_checkpoint_caches(); 4944 if (err) 4945 goto free_segment_manager_caches; 4946 err = f2fs_create_recovery_cache(); 4947 if (err) 4948 goto free_checkpoint_caches; 4949 err = f2fs_create_extent_cache(); 4950 if (err) 4951 goto free_recovery_cache; 4952 err = f2fs_create_garbage_collection_cache(); 4953 if (err) 4954 goto free_extent_cache; 4955 err = f2fs_init_sysfs(); 4956 if (err) 4957 goto free_garbage_collection_cache; 4958 err = f2fs_init_shrinker(); 4959 if (err) 4960 goto free_sysfs; 4961 err = register_filesystem(&f2fs_fs_type); 4962 if (err) 4963 goto free_shrinker; 4964 f2fs_create_root_stats(); 4965 err = f2fs_init_post_read_processing(); 4966 if (err) 4967 goto free_root_stats; 4968 err = f2fs_init_iostat_processing(); 4969 if (err) 4970 goto free_post_read; 4971 err = f2fs_init_bio_entry_cache(); 4972 if (err) 4973 goto free_iostat; 4974 err = f2fs_init_bioset(); 4975 if (err) 4976 goto free_bio_entry_cache; 4977 err = f2fs_init_compress_mempool(); 4978 if (err) 4979 goto free_bioset; 4980 err = f2fs_init_compress_cache(); 4981 if (err) 4982 goto free_compress_mempool; 4983 err = f2fs_create_casefold_cache(); 4984 if (err) 4985 goto free_compress_cache; 4986 return 0; 4987 free_compress_cache: 4988 f2fs_destroy_compress_cache(); 4989 free_compress_mempool: 4990 f2fs_destroy_compress_mempool(); 4991 free_bioset: 4992 f2fs_destroy_bioset(); 4993 free_bio_entry_cache: 4994 f2fs_destroy_bio_entry_cache(); 4995 free_iostat: 4996 f2fs_destroy_iostat_processing(); 4997 free_post_read: 4998 f2fs_destroy_post_read_processing(); 4999 free_root_stats: 5000 f2fs_destroy_root_stats(); 5001 unregister_filesystem(&f2fs_fs_type); 5002 free_shrinker: 5003 f2fs_exit_shrinker(); 5004 free_sysfs: 5005 f2fs_exit_sysfs(); 5006 free_garbage_collection_cache: 5007 f2fs_destroy_garbage_collection_cache(); 5008 free_extent_cache: 5009 f2fs_destroy_extent_cache(); 5010 free_recovery_cache: 5011 f2fs_destroy_recovery_cache(); 5012 free_checkpoint_caches: 5013 f2fs_destroy_checkpoint_caches(); 5014 free_segment_manager_caches: 5015 f2fs_destroy_segment_manager_caches(); 5016 free_node_manager_caches: 5017 f2fs_destroy_node_manager_caches(); 5018 free_inodecache: 5019 destroy_inodecache(); 5020 fail: 5021 return err; 5022 } 5023 5024 static void __exit exit_f2fs_fs(void) 5025 { 5026 f2fs_destroy_casefold_cache(); 5027 f2fs_destroy_compress_cache(); 5028 f2fs_destroy_compress_mempool(); 5029 f2fs_destroy_bioset(); 5030 f2fs_destroy_bio_entry_cache(); 5031 f2fs_destroy_iostat_processing(); 5032 f2fs_destroy_post_read_processing(); 5033 f2fs_destroy_root_stats(); 5034 unregister_filesystem(&f2fs_fs_type); 5035 f2fs_exit_shrinker(); 5036 f2fs_exit_sysfs(); 5037 f2fs_destroy_garbage_collection_cache(); 5038 f2fs_destroy_extent_cache(); 5039 f2fs_destroy_recovery_cache(); 5040 f2fs_destroy_checkpoint_caches(); 5041 f2fs_destroy_segment_manager_caches(); 5042 f2fs_destroy_node_manager_caches(); 5043 destroy_inodecache(); 5044 } 5045 5046 module_init(init_f2fs_fs) 5047 module_exit(exit_f2fs_fs) 5048 5049 MODULE_AUTHOR("Samsung Electronics's Praesto Team"); 5050 MODULE_DESCRIPTION("Flash Friendly File System"); 5051 MODULE_LICENSE("GPL"); 5052 MODULE_SOFTDEP("pre: crc32"); 5053 5054