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/sched/mm.h> 12 #include <linux/statfs.h> 13 #include <linux/buffer_head.h> 14 #include <linux/kthread.h> 15 #include <linux/parser.h> 16 #include <linux/mount.h> 17 #include <linux/seq_file.h> 18 #include <linux/proc_fs.h> 19 #include <linux/random.h> 20 #include <linux/exportfs.h> 21 #include <linux/blkdev.h> 22 #include <linux/quotaops.h> 23 #include <linux/f2fs_fs.h> 24 #include <linux/sysfs.h> 25 #include <linux/quota.h> 26 #include <linux/unicode.h> 27 #include <linux/part_stat.h> 28 #include <linux/zstd.h> 29 #include <linux/lz4.h> 30 31 #include "f2fs.h" 32 #include "node.h" 33 #include "segment.h" 34 #include "xattr.h" 35 #include "gc.h" 36 #include "iostat.h" 37 38 #define CREATE_TRACE_POINTS 39 #include <trace/events/f2fs.h> 40 41 static struct kmem_cache *f2fs_inode_cachep; 42 43 #ifdef CONFIG_F2FS_FAULT_INJECTION 44 45 const char *f2fs_fault_name[FAULT_MAX] = { 46 [FAULT_KMALLOC] = "kmalloc", 47 [FAULT_KVMALLOC] = "kvmalloc", 48 [FAULT_PAGE_ALLOC] = "page alloc", 49 [FAULT_PAGE_GET] = "page get", 50 [FAULT_ALLOC_NID] = "alloc nid", 51 [FAULT_ORPHAN] = "orphan", 52 [FAULT_BLOCK] = "no more block", 53 [FAULT_DIR_DEPTH] = "too big dir depth", 54 [FAULT_EVICT_INODE] = "evict_inode fail", 55 [FAULT_TRUNCATE] = "truncate fail", 56 [FAULT_READ_IO] = "read IO error", 57 [FAULT_CHECKPOINT] = "checkpoint error", 58 [FAULT_DISCARD] = "discard error", 59 [FAULT_WRITE_IO] = "write IO error", 60 [FAULT_SLAB_ALLOC] = "slab alloc", 61 [FAULT_DQUOT_INIT] = "dquot initialize", 62 [FAULT_LOCK_OP] = "lock_op", 63 }; 64 65 void f2fs_build_fault_attr(struct f2fs_sb_info *sbi, unsigned int rate, 66 unsigned int type) 67 { 68 struct f2fs_fault_info *ffi = &F2FS_OPTION(sbi).fault_info; 69 70 if (rate) { 71 atomic_set(&ffi->inject_ops, 0); 72 ffi->inject_rate = rate; 73 } 74 75 if (type) 76 ffi->inject_type = type; 77 78 if (!rate && !type) 79 memset(ffi, 0, sizeof(struct f2fs_fault_info)); 80 } 81 #endif 82 83 /* f2fs-wide shrinker description */ 84 static struct shrinker f2fs_shrinker_info = { 85 .scan_objects = f2fs_shrink_scan, 86 .count_objects = f2fs_shrink_count, 87 .seeks = DEFAULT_SEEKS, 88 }; 89 90 enum { 91 Opt_gc_background, 92 Opt_disable_roll_forward, 93 Opt_norecovery, 94 Opt_discard, 95 Opt_nodiscard, 96 Opt_noheap, 97 Opt_heap, 98 Opt_user_xattr, 99 Opt_nouser_xattr, 100 Opt_acl, 101 Opt_noacl, 102 Opt_active_logs, 103 Opt_disable_ext_identify, 104 Opt_inline_xattr, 105 Opt_noinline_xattr, 106 Opt_inline_xattr_size, 107 Opt_inline_data, 108 Opt_inline_dentry, 109 Opt_noinline_dentry, 110 Opt_flush_merge, 111 Opt_noflush_merge, 112 Opt_nobarrier, 113 Opt_fastboot, 114 Opt_extent_cache, 115 Opt_noextent_cache, 116 Opt_noinline_data, 117 Opt_data_flush, 118 Opt_reserve_root, 119 Opt_resgid, 120 Opt_resuid, 121 Opt_mode, 122 Opt_io_size_bits, 123 Opt_fault_injection, 124 Opt_fault_type, 125 Opt_lazytime, 126 Opt_nolazytime, 127 Opt_quota, 128 Opt_noquota, 129 Opt_usrquota, 130 Opt_grpquota, 131 Opt_prjquota, 132 Opt_usrjquota, 133 Opt_grpjquota, 134 Opt_prjjquota, 135 Opt_offusrjquota, 136 Opt_offgrpjquota, 137 Opt_offprjjquota, 138 Opt_jqfmt_vfsold, 139 Opt_jqfmt_vfsv0, 140 Opt_jqfmt_vfsv1, 141 Opt_whint, 142 Opt_alloc, 143 Opt_fsync, 144 Opt_test_dummy_encryption, 145 Opt_inlinecrypt, 146 Opt_checkpoint_disable, 147 Opt_checkpoint_disable_cap, 148 Opt_checkpoint_disable_cap_perc, 149 Opt_checkpoint_enable, 150 Opt_checkpoint_merge, 151 Opt_nocheckpoint_merge, 152 Opt_compress_algorithm, 153 Opt_compress_log_size, 154 Opt_compress_extension, 155 Opt_nocompress_extension, 156 Opt_compress_chksum, 157 Opt_compress_mode, 158 Opt_compress_cache, 159 Opt_atgc, 160 Opt_gc_merge, 161 Opt_nogc_merge, 162 Opt_discard_unit, 163 Opt_err, 164 }; 165 166 static match_table_t f2fs_tokens = { 167 {Opt_gc_background, "background_gc=%s"}, 168 {Opt_disable_roll_forward, "disable_roll_forward"}, 169 {Opt_norecovery, "norecovery"}, 170 {Opt_discard, "discard"}, 171 {Opt_nodiscard, "nodiscard"}, 172 {Opt_noheap, "no_heap"}, 173 {Opt_heap, "heap"}, 174 {Opt_user_xattr, "user_xattr"}, 175 {Opt_nouser_xattr, "nouser_xattr"}, 176 {Opt_acl, "acl"}, 177 {Opt_noacl, "noacl"}, 178 {Opt_active_logs, "active_logs=%u"}, 179 {Opt_disable_ext_identify, "disable_ext_identify"}, 180 {Opt_inline_xattr, "inline_xattr"}, 181 {Opt_noinline_xattr, "noinline_xattr"}, 182 {Opt_inline_xattr_size, "inline_xattr_size=%u"}, 183 {Opt_inline_data, "inline_data"}, 184 {Opt_inline_dentry, "inline_dentry"}, 185 {Opt_noinline_dentry, "noinline_dentry"}, 186 {Opt_flush_merge, "flush_merge"}, 187 {Opt_noflush_merge, "noflush_merge"}, 188 {Opt_nobarrier, "nobarrier"}, 189 {Opt_fastboot, "fastboot"}, 190 {Opt_extent_cache, "extent_cache"}, 191 {Opt_noextent_cache, "noextent_cache"}, 192 {Opt_noinline_data, "noinline_data"}, 193 {Opt_data_flush, "data_flush"}, 194 {Opt_reserve_root, "reserve_root=%u"}, 195 {Opt_resgid, "resgid=%u"}, 196 {Opt_resuid, "resuid=%u"}, 197 {Opt_mode, "mode=%s"}, 198 {Opt_io_size_bits, "io_bits=%u"}, 199 {Opt_fault_injection, "fault_injection=%u"}, 200 {Opt_fault_type, "fault_type=%u"}, 201 {Opt_lazytime, "lazytime"}, 202 {Opt_nolazytime, "nolazytime"}, 203 {Opt_quota, "quota"}, 204 {Opt_noquota, "noquota"}, 205 {Opt_usrquota, "usrquota"}, 206 {Opt_grpquota, "grpquota"}, 207 {Opt_prjquota, "prjquota"}, 208 {Opt_usrjquota, "usrjquota=%s"}, 209 {Opt_grpjquota, "grpjquota=%s"}, 210 {Opt_prjjquota, "prjjquota=%s"}, 211 {Opt_offusrjquota, "usrjquota="}, 212 {Opt_offgrpjquota, "grpjquota="}, 213 {Opt_offprjjquota, "prjjquota="}, 214 {Opt_jqfmt_vfsold, "jqfmt=vfsold"}, 215 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"}, 216 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"}, 217 {Opt_whint, "whint_mode=%s"}, 218 {Opt_alloc, "alloc_mode=%s"}, 219 {Opt_fsync, "fsync_mode=%s"}, 220 {Opt_test_dummy_encryption, "test_dummy_encryption=%s"}, 221 {Opt_test_dummy_encryption, "test_dummy_encryption"}, 222 {Opt_inlinecrypt, "inlinecrypt"}, 223 {Opt_checkpoint_disable, "checkpoint=disable"}, 224 {Opt_checkpoint_disable_cap, "checkpoint=disable:%u"}, 225 {Opt_checkpoint_disable_cap_perc, "checkpoint=disable:%u%%"}, 226 {Opt_checkpoint_enable, "checkpoint=enable"}, 227 {Opt_checkpoint_merge, "checkpoint_merge"}, 228 {Opt_nocheckpoint_merge, "nocheckpoint_merge"}, 229 {Opt_compress_algorithm, "compress_algorithm=%s"}, 230 {Opt_compress_log_size, "compress_log_size=%u"}, 231 {Opt_compress_extension, "compress_extension=%s"}, 232 {Opt_nocompress_extension, "nocompress_extension=%s"}, 233 {Opt_compress_chksum, "compress_chksum"}, 234 {Opt_compress_mode, "compress_mode=%s"}, 235 {Opt_compress_cache, "compress_cache"}, 236 {Opt_atgc, "atgc"}, 237 {Opt_gc_merge, "gc_merge"}, 238 {Opt_nogc_merge, "nogc_merge"}, 239 {Opt_discard_unit, "discard_unit=%s"}, 240 {Opt_err, NULL}, 241 }; 242 243 void f2fs_printk(struct f2fs_sb_info *sbi, const char *fmt, ...) 244 { 245 struct va_format vaf; 246 va_list args; 247 int level; 248 249 va_start(args, fmt); 250 251 level = printk_get_level(fmt); 252 vaf.fmt = printk_skip_level(fmt); 253 vaf.va = &args; 254 printk("%c%cF2FS-fs (%s): %pV\n", 255 KERN_SOH_ASCII, level, sbi->sb->s_id, &vaf); 256 257 va_end(args); 258 } 259 260 #if IS_ENABLED(CONFIG_UNICODE) 261 static const struct f2fs_sb_encodings { 262 __u16 magic; 263 char *name; 264 unsigned int version; 265 } f2fs_sb_encoding_map[] = { 266 {F2FS_ENC_UTF8_12_1, "utf8", UNICODE_AGE(12, 1, 0)}, 267 }; 268 269 static const struct f2fs_sb_encodings * 270 f2fs_sb_read_encoding(const struct f2fs_super_block *sb) 271 { 272 __u16 magic = le16_to_cpu(sb->s_encoding); 273 int i; 274 275 for (i = 0; i < ARRAY_SIZE(f2fs_sb_encoding_map); i++) 276 if (magic == f2fs_sb_encoding_map[i].magic) 277 return &f2fs_sb_encoding_map[i]; 278 279 return NULL; 280 } 281 282 struct kmem_cache *f2fs_cf_name_slab; 283 static int __init f2fs_create_casefold_cache(void) 284 { 285 f2fs_cf_name_slab = f2fs_kmem_cache_create("f2fs_casefolded_name", 286 F2FS_NAME_LEN); 287 if (!f2fs_cf_name_slab) 288 return -ENOMEM; 289 return 0; 290 } 291 292 static void f2fs_destroy_casefold_cache(void) 293 { 294 kmem_cache_destroy(f2fs_cf_name_slab); 295 } 296 #else 297 static int __init f2fs_create_casefold_cache(void) { return 0; } 298 static void f2fs_destroy_casefold_cache(void) { } 299 #endif 300 301 static inline void limit_reserve_root(struct f2fs_sb_info *sbi) 302 { 303 block_t limit = min((sbi->user_block_count << 1) / 1000, 304 sbi->user_block_count - sbi->reserved_blocks); 305 306 /* limit is 0.2% */ 307 if (test_opt(sbi, RESERVE_ROOT) && 308 F2FS_OPTION(sbi).root_reserved_blocks > limit) { 309 F2FS_OPTION(sbi).root_reserved_blocks = limit; 310 f2fs_info(sbi, "Reduce reserved blocks for root = %u", 311 F2FS_OPTION(sbi).root_reserved_blocks); 312 } 313 if (!test_opt(sbi, RESERVE_ROOT) && 314 (!uid_eq(F2FS_OPTION(sbi).s_resuid, 315 make_kuid(&init_user_ns, F2FS_DEF_RESUID)) || 316 !gid_eq(F2FS_OPTION(sbi).s_resgid, 317 make_kgid(&init_user_ns, F2FS_DEF_RESGID)))) 318 f2fs_info(sbi, "Ignore s_resuid=%u, s_resgid=%u w/o reserve_root", 319 from_kuid_munged(&init_user_ns, 320 F2FS_OPTION(sbi).s_resuid), 321 from_kgid_munged(&init_user_ns, 322 F2FS_OPTION(sbi).s_resgid)); 323 } 324 325 static inline int adjust_reserved_segment(struct f2fs_sb_info *sbi) 326 { 327 unsigned int sec_blks = sbi->blocks_per_seg * sbi->segs_per_sec; 328 unsigned int avg_vblocks; 329 unsigned int wanted_reserved_segments; 330 block_t avail_user_block_count; 331 332 if (!F2FS_IO_ALIGNED(sbi)) 333 return 0; 334 335 /* average valid block count in section in worst case */ 336 avg_vblocks = sec_blks / F2FS_IO_SIZE(sbi); 337 338 /* 339 * we need enough free space when migrating one section in worst case 340 */ 341 wanted_reserved_segments = (F2FS_IO_SIZE(sbi) / avg_vblocks) * 342 reserved_segments(sbi); 343 wanted_reserved_segments -= reserved_segments(sbi); 344 345 avail_user_block_count = sbi->user_block_count - 346 sbi->current_reserved_blocks - 347 F2FS_OPTION(sbi).root_reserved_blocks; 348 349 if (wanted_reserved_segments * sbi->blocks_per_seg > 350 avail_user_block_count) { 351 f2fs_err(sbi, "IO align feature can't grab additional reserved segment: %u, available segments: %u", 352 wanted_reserved_segments, 353 avail_user_block_count >> sbi->log_blocks_per_seg); 354 return -ENOSPC; 355 } 356 357 SM_I(sbi)->additional_reserved_segments = wanted_reserved_segments; 358 359 f2fs_info(sbi, "IO align feature needs additional reserved segment: %u", 360 wanted_reserved_segments); 361 362 return 0; 363 } 364 365 static inline void adjust_unusable_cap_perc(struct f2fs_sb_info *sbi) 366 { 367 if (!F2FS_OPTION(sbi).unusable_cap_perc) 368 return; 369 370 if (F2FS_OPTION(sbi).unusable_cap_perc == 100) 371 F2FS_OPTION(sbi).unusable_cap = sbi->user_block_count; 372 else 373 F2FS_OPTION(sbi).unusable_cap = (sbi->user_block_count / 100) * 374 F2FS_OPTION(sbi).unusable_cap_perc; 375 376 f2fs_info(sbi, "Adjust unusable cap for checkpoint=disable = %u / %u%%", 377 F2FS_OPTION(sbi).unusable_cap, 378 F2FS_OPTION(sbi).unusable_cap_perc); 379 } 380 381 static void init_once(void *foo) 382 { 383 struct f2fs_inode_info *fi = (struct f2fs_inode_info *) foo; 384 385 inode_init_once(&fi->vfs_inode); 386 } 387 388 #ifdef CONFIG_QUOTA 389 static const char * const quotatypes[] = INITQFNAMES; 390 #define QTYPE2NAME(t) (quotatypes[t]) 391 static int f2fs_set_qf_name(struct super_block *sb, int qtype, 392 substring_t *args) 393 { 394 struct f2fs_sb_info *sbi = F2FS_SB(sb); 395 char *qname; 396 int ret = -EINVAL; 397 398 if (sb_any_quota_loaded(sb) && !F2FS_OPTION(sbi).s_qf_names[qtype]) { 399 f2fs_err(sbi, "Cannot change journaled quota options when quota turned on"); 400 return -EINVAL; 401 } 402 if (f2fs_sb_has_quota_ino(sbi)) { 403 f2fs_info(sbi, "QUOTA feature is enabled, so ignore qf_name"); 404 return 0; 405 } 406 407 qname = match_strdup(args); 408 if (!qname) { 409 f2fs_err(sbi, "Not enough memory for storing quotafile name"); 410 return -ENOMEM; 411 } 412 if (F2FS_OPTION(sbi).s_qf_names[qtype]) { 413 if (strcmp(F2FS_OPTION(sbi).s_qf_names[qtype], qname) == 0) 414 ret = 0; 415 else 416 f2fs_err(sbi, "%s quota file already specified", 417 QTYPE2NAME(qtype)); 418 goto errout; 419 } 420 if (strchr(qname, '/')) { 421 f2fs_err(sbi, "quotafile must be on filesystem root"); 422 goto errout; 423 } 424 F2FS_OPTION(sbi).s_qf_names[qtype] = qname; 425 set_opt(sbi, QUOTA); 426 return 0; 427 errout: 428 kfree(qname); 429 return ret; 430 } 431 432 static int f2fs_clear_qf_name(struct super_block *sb, int qtype) 433 { 434 struct f2fs_sb_info *sbi = F2FS_SB(sb); 435 436 if (sb_any_quota_loaded(sb) && F2FS_OPTION(sbi).s_qf_names[qtype]) { 437 f2fs_err(sbi, "Cannot change journaled quota options when quota turned on"); 438 return -EINVAL; 439 } 440 kfree(F2FS_OPTION(sbi).s_qf_names[qtype]); 441 F2FS_OPTION(sbi).s_qf_names[qtype] = NULL; 442 return 0; 443 } 444 445 static int f2fs_check_quota_options(struct f2fs_sb_info *sbi) 446 { 447 /* 448 * We do the test below only for project quotas. 'usrquota' and 449 * 'grpquota' mount options are allowed even without quota feature 450 * to support legacy quotas in quota files. 451 */ 452 if (test_opt(sbi, PRJQUOTA) && !f2fs_sb_has_project_quota(sbi)) { 453 f2fs_err(sbi, "Project quota feature not enabled. Cannot enable project quota enforcement."); 454 return -1; 455 } 456 if (F2FS_OPTION(sbi).s_qf_names[USRQUOTA] || 457 F2FS_OPTION(sbi).s_qf_names[GRPQUOTA] || 458 F2FS_OPTION(sbi).s_qf_names[PRJQUOTA]) { 459 if (test_opt(sbi, USRQUOTA) && 460 F2FS_OPTION(sbi).s_qf_names[USRQUOTA]) 461 clear_opt(sbi, USRQUOTA); 462 463 if (test_opt(sbi, GRPQUOTA) && 464 F2FS_OPTION(sbi).s_qf_names[GRPQUOTA]) 465 clear_opt(sbi, GRPQUOTA); 466 467 if (test_opt(sbi, PRJQUOTA) && 468 F2FS_OPTION(sbi).s_qf_names[PRJQUOTA]) 469 clear_opt(sbi, PRJQUOTA); 470 471 if (test_opt(sbi, GRPQUOTA) || test_opt(sbi, USRQUOTA) || 472 test_opt(sbi, PRJQUOTA)) { 473 f2fs_err(sbi, "old and new quota format mixing"); 474 return -1; 475 } 476 477 if (!F2FS_OPTION(sbi).s_jquota_fmt) { 478 f2fs_err(sbi, "journaled quota format not specified"); 479 return -1; 480 } 481 } 482 483 if (f2fs_sb_has_quota_ino(sbi) && F2FS_OPTION(sbi).s_jquota_fmt) { 484 f2fs_info(sbi, "QUOTA feature is enabled, so ignore jquota_fmt"); 485 F2FS_OPTION(sbi).s_jquota_fmt = 0; 486 } 487 return 0; 488 } 489 #endif 490 491 static int f2fs_set_test_dummy_encryption(struct super_block *sb, 492 const char *opt, 493 const substring_t *arg, 494 bool is_remount) 495 { 496 struct f2fs_sb_info *sbi = F2FS_SB(sb); 497 #ifdef CONFIG_FS_ENCRYPTION 498 int err; 499 500 if (!f2fs_sb_has_encrypt(sbi)) { 501 f2fs_err(sbi, "Encrypt feature is off"); 502 return -EINVAL; 503 } 504 505 /* 506 * This mount option is just for testing, and it's not worthwhile to 507 * implement the extra complexity (e.g. RCU protection) that would be 508 * needed to allow it to be set or changed during remount. We do allow 509 * it to be specified during remount, but only if there is no change. 510 */ 511 if (is_remount && !F2FS_OPTION(sbi).dummy_enc_policy.policy) { 512 f2fs_warn(sbi, "Can't set test_dummy_encryption on remount"); 513 return -EINVAL; 514 } 515 err = fscrypt_set_test_dummy_encryption( 516 sb, arg->from, &F2FS_OPTION(sbi).dummy_enc_policy); 517 if (err) { 518 if (err == -EEXIST) 519 f2fs_warn(sbi, 520 "Can't change test_dummy_encryption on remount"); 521 else if (err == -EINVAL) 522 f2fs_warn(sbi, "Value of option \"%s\" is unrecognized", 523 opt); 524 else 525 f2fs_warn(sbi, "Error processing option \"%s\" [%d]", 526 opt, err); 527 return -EINVAL; 528 } 529 f2fs_warn(sbi, "Test dummy encryption mode enabled"); 530 #else 531 f2fs_warn(sbi, "Test dummy encryption mount option ignored"); 532 #endif 533 return 0; 534 } 535 536 #ifdef CONFIG_F2FS_FS_COMPRESSION 537 /* 538 * 1. The same extension name cannot not appear in both compress and non-compress extension 539 * at the same time. 540 * 2. If the compress extension specifies all files, the types specified by the non-compress 541 * extension will be treated as special cases and will not be compressed. 542 * 3. Don't allow the non-compress extension specifies all files. 543 */ 544 static int f2fs_test_compress_extension(struct f2fs_sb_info *sbi) 545 { 546 unsigned char (*ext)[F2FS_EXTENSION_LEN]; 547 unsigned char (*noext)[F2FS_EXTENSION_LEN]; 548 int ext_cnt, noext_cnt, index = 0, no_index = 0; 549 550 ext = F2FS_OPTION(sbi).extensions; 551 ext_cnt = F2FS_OPTION(sbi).compress_ext_cnt; 552 noext = F2FS_OPTION(sbi).noextensions; 553 noext_cnt = F2FS_OPTION(sbi).nocompress_ext_cnt; 554 555 if (!noext_cnt) 556 return 0; 557 558 for (no_index = 0; no_index < noext_cnt; no_index++) { 559 if (!strcasecmp("*", noext[no_index])) { 560 f2fs_info(sbi, "Don't allow the nocompress extension specifies all files"); 561 return -EINVAL; 562 } 563 for (index = 0; index < ext_cnt; index++) { 564 if (!strcasecmp(ext[index], noext[no_index])) { 565 f2fs_info(sbi, "Don't allow the same extension %s appear in both compress and nocompress extension", 566 ext[index]); 567 return -EINVAL; 568 } 569 } 570 } 571 return 0; 572 } 573 574 #ifdef CONFIG_F2FS_FS_LZ4 575 static int f2fs_set_lz4hc_level(struct f2fs_sb_info *sbi, const char *str) 576 { 577 #ifdef CONFIG_F2FS_FS_LZ4HC 578 unsigned int level; 579 #endif 580 581 if (strlen(str) == 3) { 582 F2FS_OPTION(sbi).compress_level = 0; 583 return 0; 584 } 585 586 #ifdef CONFIG_F2FS_FS_LZ4HC 587 str += 3; 588 589 if (str[0] != ':') { 590 f2fs_info(sbi, "wrong format, e.g. <alg_name>:<compr_level>"); 591 return -EINVAL; 592 } 593 if (kstrtouint(str + 1, 10, &level)) 594 return -EINVAL; 595 596 if (level < LZ4HC_MIN_CLEVEL || level > LZ4HC_MAX_CLEVEL) { 597 f2fs_info(sbi, "invalid lz4hc compress level: %d", level); 598 return -EINVAL; 599 } 600 601 F2FS_OPTION(sbi).compress_level = level; 602 return 0; 603 #else 604 f2fs_info(sbi, "kernel doesn't support lz4hc compression"); 605 return -EINVAL; 606 #endif 607 } 608 #endif 609 610 #ifdef CONFIG_F2FS_FS_ZSTD 611 static int f2fs_set_zstd_level(struct f2fs_sb_info *sbi, const char *str) 612 { 613 unsigned int level; 614 int len = 4; 615 616 if (strlen(str) == len) { 617 F2FS_OPTION(sbi).compress_level = 0; 618 return 0; 619 } 620 621 str += len; 622 623 if (str[0] != ':') { 624 f2fs_info(sbi, "wrong format, e.g. <alg_name>:<compr_level>"); 625 return -EINVAL; 626 } 627 if (kstrtouint(str + 1, 10, &level)) 628 return -EINVAL; 629 630 if (!level || level > zstd_max_clevel()) { 631 f2fs_info(sbi, "invalid zstd compress level: %d", level); 632 return -EINVAL; 633 } 634 635 F2FS_OPTION(sbi).compress_level = level; 636 return 0; 637 } 638 #endif 639 #endif 640 641 static int parse_options(struct super_block *sb, char *options, bool is_remount) 642 { 643 struct f2fs_sb_info *sbi = F2FS_SB(sb); 644 substring_t args[MAX_OPT_ARGS]; 645 #ifdef CONFIG_F2FS_FS_COMPRESSION 646 unsigned char (*ext)[F2FS_EXTENSION_LEN]; 647 unsigned char (*noext)[F2FS_EXTENSION_LEN]; 648 int ext_cnt, noext_cnt; 649 #endif 650 char *p, *name; 651 int arg = 0; 652 kuid_t uid; 653 kgid_t gid; 654 int ret; 655 656 if (!options) 657 goto default_check; 658 659 while ((p = strsep(&options, ",")) != NULL) { 660 int token; 661 662 if (!*p) 663 continue; 664 /* 665 * Initialize args struct so we know whether arg was 666 * found; some options take optional arguments. 667 */ 668 args[0].to = args[0].from = NULL; 669 token = match_token(p, f2fs_tokens, args); 670 671 switch (token) { 672 case Opt_gc_background: 673 name = match_strdup(&args[0]); 674 675 if (!name) 676 return -ENOMEM; 677 if (!strcmp(name, "on")) { 678 F2FS_OPTION(sbi).bggc_mode = BGGC_MODE_ON; 679 } else if (!strcmp(name, "off")) { 680 F2FS_OPTION(sbi).bggc_mode = BGGC_MODE_OFF; 681 } else if (!strcmp(name, "sync")) { 682 F2FS_OPTION(sbi).bggc_mode = BGGC_MODE_SYNC; 683 } else { 684 kfree(name); 685 return -EINVAL; 686 } 687 kfree(name); 688 break; 689 case Opt_disable_roll_forward: 690 set_opt(sbi, DISABLE_ROLL_FORWARD); 691 break; 692 case Opt_norecovery: 693 /* this option mounts f2fs with ro */ 694 set_opt(sbi, NORECOVERY); 695 if (!f2fs_readonly(sb)) 696 return -EINVAL; 697 break; 698 case Opt_discard: 699 if (!f2fs_hw_support_discard(sbi)) { 700 f2fs_warn(sbi, "device does not support discard"); 701 break; 702 } 703 set_opt(sbi, DISCARD); 704 break; 705 case Opt_nodiscard: 706 if (f2fs_hw_should_discard(sbi)) { 707 f2fs_warn(sbi, "discard is required for zoned block devices"); 708 return -EINVAL; 709 } 710 clear_opt(sbi, DISCARD); 711 break; 712 case Opt_noheap: 713 set_opt(sbi, NOHEAP); 714 break; 715 case Opt_heap: 716 clear_opt(sbi, NOHEAP); 717 break; 718 #ifdef CONFIG_F2FS_FS_XATTR 719 case Opt_user_xattr: 720 set_opt(sbi, XATTR_USER); 721 break; 722 case Opt_nouser_xattr: 723 clear_opt(sbi, XATTR_USER); 724 break; 725 case Opt_inline_xattr: 726 set_opt(sbi, INLINE_XATTR); 727 break; 728 case Opt_noinline_xattr: 729 clear_opt(sbi, INLINE_XATTR); 730 break; 731 case Opt_inline_xattr_size: 732 if (args->from && match_int(args, &arg)) 733 return -EINVAL; 734 set_opt(sbi, INLINE_XATTR_SIZE); 735 F2FS_OPTION(sbi).inline_xattr_size = arg; 736 break; 737 #else 738 case Opt_user_xattr: 739 f2fs_info(sbi, "user_xattr options not supported"); 740 break; 741 case Opt_nouser_xattr: 742 f2fs_info(sbi, "nouser_xattr options not supported"); 743 break; 744 case Opt_inline_xattr: 745 f2fs_info(sbi, "inline_xattr options not supported"); 746 break; 747 case Opt_noinline_xattr: 748 f2fs_info(sbi, "noinline_xattr options not supported"); 749 break; 750 #endif 751 #ifdef CONFIG_F2FS_FS_POSIX_ACL 752 case Opt_acl: 753 set_opt(sbi, POSIX_ACL); 754 break; 755 case Opt_noacl: 756 clear_opt(sbi, POSIX_ACL); 757 break; 758 #else 759 case Opt_acl: 760 f2fs_info(sbi, "acl options not supported"); 761 break; 762 case Opt_noacl: 763 f2fs_info(sbi, "noacl options not supported"); 764 break; 765 #endif 766 case Opt_active_logs: 767 if (args->from && match_int(args, &arg)) 768 return -EINVAL; 769 if (arg != 2 && arg != 4 && 770 arg != NR_CURSEG_PERSIST_TYPE) 771 return -EINVAL; 772 F2FS_OPTION(sbi).active_logs = arg; 773 break; 774 case Opt_disable_ext_identify: 775 set_opt(sbi, DISABLE_EXT_IDENTIFY); 776 break; 777 case Opt_inline_data: 778 set_opt(sbi, INLINE_DATA); 779 break; 780 case Opt_inline_dentry: 781 set_opt(sbi, INLINE_DENTRY); 782 break; 783 case Opt_noinline_dentry: 784 clear_opt(sbi, INLINE_DENTRY); 785 break; 786 case Opt_flush_merge: 787 set_opt(sbi, FLUSH_MERGE); 788 break; 789 case Opt_noflush_merge: 790 clear_opt(sbi, FLUSH_MERGE); 791 break; 792 case Opt_nobarrier: 793 set_opt(sbi, NOBARRIER); 794 break; 795 case Opt_fastboot: 796 set_opt(sbi, FASTBOOT); 797 break; 798 case Opt_extent_cache: 799 set_opt(sbi, EXTENT_CACHE); 800 break; 801 case Opt_noextent_cache: 802 clear_opt(sbi, EXTENT_CACHE); 803 break; 804 case Opt_noinline_data: 805 clear_opt(sbi, INLINE_DATA); 806 break; 807 case Opt_data_flush: 808 set_opt(sbi, DATA_FLUSH); 809 break; 810 case Opt_reserve_root: 811 if (args->from && match_int(args, &arg)) 812 return -EINVAL; 813 if (test_opt(sbi, RESERVE_ROOT)) { 814 f2fs_info(sbi, "Preserve previous reserve_root=%u", 815 F2FS_OPTION(sbi).root_reserved_blocks); 816 } else { 817 F2FS_OPTION(sbi).root_reserved_blocks = arg; 818 set_opt(sbi, RESERVE_ROOT); 819 } 820 break; 821 case Opt_resuid: 822 if (args->from && match_int(args, &arg)) 823 return -EINVAL; 824 uid = make_kuid(current_user_ns(), arg); 825 if (!uid_valid(uid)) { 826 f2fs_err(sbi, "Invalid uid value %d", arg); 827 return -EINVAL; 828 } 829 F2FS_OPTION(sbi).s_resuid = uid; 830 break; 831 case Opt_resgid: 832 if (args->from && match_int(args, &arg)) 833 return -EINVAL; 834 gid = make_kgid(current_user_ns(), arg); 835 if (!gid_valid(gid)) { 836 f2fs_err(sbi, "Invalid gid value %d", arg); 837 return -EINVAL; 838 } 839 F2FS_OPTION(sbi).s_resgid = gid; 840 break; 841 case Opt_mode: 842 name = match_strdup(&args[0]); 843 844 if (!name) 845 return -ENOMEM; 846 if (!strcmp(name, "adaptive")) { 847 if (f2fs_sb_has_blkzoned(sbi)) { 848 f2fs_warn(sbi, "adaptive mode is not allowed with zoned block device feature"); 849 kfree(name); 850 return -EINVAL; 851 } 852 F2FS_OPTION(sbi).fs_mode = FS_MODE_ADAPTIVE; 853 } else if (!strcmp(name, "lfs")) { 854 F2FS_OPTION(sbi).fs_mode = FS_MODE_LFS; 855 } else if (!strcmp(name, "fragment:segment")) { 856 F2FS_OPTION(sbi).fs_mode = FS_MODE_FRAGMENT_SEG; 857 } else if (!strcmp(name, "fragment:block")) { 858 F2FS_OPTION(sbi).fs_mode = FS_MODE_FRAGMENT_BLK; 859 } else { 860 kfree(name); 861 return -EINVAL; 862 } 863 kfree(name); 864 break; 865 case Opt_io_size_bits: 866 if (args->from && match_int(args, &arg)) 867 return -EINVAL; 868 if (arg <= 0 || arg > __ilog2_u32(BIO_MAX_VECS)) { 869 f2fs_warn(sbi, "Not support %d, larger than %d", 870 1 << arg, BIO_MAX_VECS); 871 return -EINVAL; 872 } 873 F2FS_OPTION(sbi).write_io_size_bits = arg; 874 break; 875 #ifdef CONFIG_F2FS_FAULT_INJECTION 876 case Opt_fault_injection: 877 if (args->from && match_int(args, &arg)) 878 return -EINVAL; 879 f2fs_build_fault_attr(sbi, arg, F2FS_ALL_FAULT_TYPE); 880 set_opt(sbi, FAULT_INJECTION); 881 break; 882 883 case Opt_fault_type: 884 if (args->from && match_int(args, &arg)) 885 return -EINVAL; 886 f2fs_build_fault_attr(sbi, 0, arg); 887 set_opt(sbi, FAULT_INJECTION); 888 break; 889 #else 890 case Opt_fault_injection: 891 f2fs_info(sbi, "fault_injection options not supported"); 892 break; 893 894 case Opt_fault_type: 895 f2fs_info(sbi, "fault_type options not supported"); 896 break; 897 #endif 898 case Opt_lazytime: 899 sb->s_flags |= SB_LAZYTIME; 900 break; 901 case Opt_nolazytime: 902 sb->s_flags &= ~SB_LAZYTIME; 903 break; 904 #ifdef CONFIG_QUOTA 905 case Opt_quota: 906 case Opt_usrquota: 907 set_opt(sbi, USRQUOTA); 908 break; 909 case Opt_grpquota: 910 set_opt(sbi, GRPQUOTA); 911 break; 912 case Opt_prjquota: 913 set_opt(sbi, PRJQUOTA); 914 break; 915 case Opt_usrjquota: 916 ret = f2fs_set_qf_name(sb, USRQUOTA, &args[0]); 917 if (ret) 918 return ret; 919 break; 920 case Opt_grpjquota: 921 ret = f2fs_set_qf_name(sb, GRPQUOTA, &args[0]); 922 if (ret) 923 return ret; 924 break; 925 case Opt_prjjquota: 926 ret = f2fs_set_qf_name(sb, PRJQUOTA, &args[0]); 927 if (ret) 928 return ret; 929 break; 930 case Opt_offusrjquota: 931 ret = f2fs_clear_qf_name(sb, USRQUOTA); 932 if (ret) 933 return ret; 934 break; 935 case Opt_offgrpjquota: 936 ret = f2fs_clear_qf_name(sb, GRPQUOTA); 937 if (ret) 938 return ret; 939 break; 940 case Opt_offprjjquota: 941 ret = f2fs_clear_qf_name(sb, PRJQUOTA); 942 if (ret) 943 return ret; 944 break; 945 case Opt_jqfmt_vfsold: 946 F2FS_OPTION(sbi).s_jquota_fmt = QFMT_VFS_OLD; 947 break; 948 case Opt_jqfmt_vfsv0: 949 F2FS_OPTION(sbi).s_jquota_fmt = QFMT_VFS_V0; 950 break; 951 case Opt_jqfmt_vfsv1: 952 F2FS_OPTION(sbi).s_jquota_fmt = QFMT_VFS_V1; 953 break; 954 case Opt_noquota: 955 clear_opt(sbi, QUOTA); 956 clear_opt(sbi, USRQUOTA); 957 clear_opt(sbi, GRPQUOTA); 958 clear_opt(sbi, PRJQUOTA); 959 break; 960 #else 961 case Opt_quota: 962 case Opt_usrquota: 963 case Opt_grpquota: 964 case Opt_prjquota: 965 case Opt_usrjquota: 966 case Opt_grpjquota: 967 case Opt_prjjquota: 968 case Opt_offusrjquota: 969 case Opt_offgrpjquota: 970 case Opt_offprjjquota: 971 case Opt_jqfmt_vfsold: 972 case Opt_jqfmt_vfsv0: 973 case Opt_jqfmt_vfsv1: 974 case Opt_noquota: 975 f2fs_info(sbi, "quota operations not supported"); 976 break; 977 #endif 978 case Opt_whint: 979 name = match_strdup(&args[0]); 980 if (!name) 981 return -ENOMEM; 982 if (!strcmp(name, "user-based")) { 983 F2FS_OPTION(sbi).whint_mode = WHINT_MODE_USER; 984 } else if (!strcmp(name, "off")) { 985 F2FS_OPTION(sbi).whint_mode = WHINT_MODE_OFF; 986 } else if (!strcmp(name, "fs-based")) { 987 F2FS_OPTION(sbi).whint_mode = WHINT_MODE_FS; 988 } else { 989 kfree(name); 990 return -EINVAL; 991 } 992 kfree(name); 993 break; 994 case Opt_alloc: 995 name = match_strdup(&args[0]); 996 if (!name) 997 return -ENOMEM; 998 999 if (!strcmp(name, "default")) { 1000 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_DEFAULT; 1001 } else if (!strcmp(name, "reuse")) { 1002 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_REUSE; 1003 } else { 1004 kfree(name); 1005 return -EINVAL; 1006 } 1007 kfree(name); 1008 break; 1009 case Opt_fsync: 1010 name = match_strdup(&args[0]); 1011 if (!name) 1012 return -ENOMEM; 1013 if (!strcmp(name, "posix")) { 1014 F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_POSIX; 1015 } else if (!strcmp(name, "strict")) { 1016 F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_STRICT; 1017 } else if (!strcmp(name, "nobarrier")) { 1018 F2FS_OPTION(sbi).fsync_mode = 1019 FSYNC_MODE_NOBARRIER; 1020 } else { 1021 kfree(name); 1022 return -EINVAL; 1023 } 1024 kfree(name); 1025 break; 1026 case Opt_test_dummy_encryption: 1027 ret = f2fs_set_test_dummy_encryption(sb, p, &args[0], 1028 is_remount); 1029 if (ret) 1030 return ret; 1031 break; 1032 case Opt_inlinecrypt: 1033 #ifdef CONFIG_FS_ENCRYPTION_INLINE_CRYPT 1034 sb->s_flags |= SB_INLINECRYPT; 1035 #else 1036 f2fs_info(sbi, "inline encryption not supported"); 1037 #endif 1038 break; 1039 case Opt_checkpoint_disable_cap_perc: 1040 if (args->from && match_int(args, &arg)) 1041 return -EINVAL; 1042 if (arg < 0 || arg > 100) 1043 return -EINVAL; 1044 F2FS_OPTION(sbi).unusable_cap_perc = arg; 1045 set_opt(sbi, DISABLE_CHECKPOINT); 1046 break; 1047 case Opt_checkpoint_disable_cap: 1048 if (args->from && match_int(args, &arg)) 1049 return -EINVAL; 1050 F2FS_OPTION(sbi).unusable_cap = arg; 1051 set_opt(sbi, DISABLE_CHECKPOINT); 1052 break; 1053 case Opt_checkpoint_disable: 1054 set_opt(sbi, DISABLE_CHECKPOINT); 1055 break; 1056 case Opt_checkpoint_enable: 1057 clear_opt(sbi, DISABLE_CHECKPOINT); 1058 break; 1059 case Opt_checkpoint_merge: 1060 set_opt(sbi, MERGE_CHECKPOINT); 1061 break; 1062 case Opt_nocheckpoint_merge: 1063 clear_opt(sbi, MERGE_CHECKPOINT); 1064 break; 1065 #ifdef CONFIG_F2FS_FS_COMPRESSION 1066 case Opt_compress_algorithm: 1067 if (!f2fs_sb_has_compression(sbi)) { 1068 f2fs_info(sbi, "Image doesn't support compression"); 1069 break; 1070 } 1071 name = match_strdup(&args[0]); 1072 if (!name) 1073 return -ENOMEM; 1074 if (!strcmp(name, "lzo")) { 1075 #ifdef CONFIG_F2FS_FS_LZO 1076 F2FS_OPTION(sbi).compress_level = 0; 1077 F2FS_OPTION(sbi).compress_algorithm = 1078 COMPRESS_LZO; 1079 #else 1080 f2fs_info(sbi, "kernel doesn't support lzo compression"); 1081 #endif 1082 } else if (!strncmp(name, "lz4", 3)) { 1083 #ifdef CONFIG_F2FS_FS_LZ4 1084 ret = f2fs_set_lz4hc_level(sbi, name); 1085 if (ret) { 1086 kfree(name); 1087 return -EINVAL; 1088 } 1089 F2FS_OPTION(sbi).compress_algorithm = 1090 COMPRESS_LZ4; 1091 #else 1092 f2fs_info(sbi, "kernel doesn't support lz4 compression"); 1093 #endif 1094 } else if (!strncmp(name, "zstd", 4)) { 1095 #ifdef CONFIG_F2FS_FS_ZSTD 1096 ret = f2fs_set_zstd_level(sbi, name); 1097 if (ret) { 1098 kfree(name); 1099 return -EINVAL; 1100 } 1101 F2FS_OPTION(sbi).compress_algorithm = 1102 COMPRESS_ZSTD; 1103 #else 1104 f2fs_info(sbi, "kernel doesn't support zstd compression"); 1105 #endif 1106 } else if (!strcmp(name, "lzo-rle")) { 1107 #ifdef CONFIG_F2FS_FS_LZORLE 1108 F2FS_OPTION(sbi).compress_level = 0; 1109 F2FS_OPTION(sbi).compress_algorithm = 1110 COMPRESS_LZORLE; 1111 #else 1112 f2fs_info(sbi, "kernel doesn't support lzorle compression"); 1113 #endif 1114 } else { 1115 kfree(name); 1116 return -EINVAL; 1117 } 1118 kfree(name); 1119 break; 1120 case Opt_compress_log_size: 1121 if (!f2fs_sb_has_compression(sbi)) { 1122 f2fs_info(sbi, "Image doesn't support compression"); 1123 break; 1124 } 1125 if (args->from && match_int(args, &arg)) 1126 return -EINVAL; 1127 if (arg < MIN_COMPRESS_LOG_SIZE || 1128 arg > MAX_COMPRESS_LOG_SIZE) { 1129 f2fs_err(sbi, 1130 "Compress cluster log size is out of range"); 1131 return -EINVAL; 1132 } 1133 F2FS_OPTION(sbi).compress_log_size = arg; 1134 break; 1135 case Opt_compress_extension: 1136 if (!f2fs_sb_has_compression(sbi)) { 1137 f2fs_info(sbi, "Image doesn't support compression"); 1138 break; 1139 } 1140 name = match_strdup(&args[0]); 1141 if (!name) 1142 return -ENOMEM; 1143 1144 ext = F2FS_OPTION(sbi).extensions; 1145 ext_cnt = F2FS_OPTION(sbi).compress_ext_cnt; 1146 1147 if (strlen(name) >= F2FS_EXTENSION_LEN || 1148 ext_cnt >= COMPRESS_EXT_NUM) { 1149 f2fs_err(sbi, 1150 "invalid extension length/number"); 1151 kfree(name); 1152 return -EINVAL; 1153 } 1154 1155 strcpy(ext[ext_cnt], name); 1156 F2FS_OPTION(sbi).compress_ext_cnt++; 1157 kfree(name); 1158 break; 1159 case Opt_nocompress_extension: 1160 if (!f2fs_sb_has_compression(sbi)) { 1161 f2fs_info(sbi, "Image doesn't support compression"); 1162 break; 1163 } 1164 name = match_strdup(&args[0]); 1165 if (!name) 1166 return -ENOMEM; 1167 1168 noext = F2FS_OPTION(sbi).noextensions; 1169 noext_cnt = F2FS_OPTION(sbi).nocompress_ext_cnt; 1170 1171 if (strlen(name) >= F2FS_EXTENSION_LEN || 1172 noext_cnt >= COMPRESS_EXT_NUM) { 1173 f2fs_err(sbi, 1174 "invalid extension length/number"); 1175 kfree(name); 1176 return -EINVAL; 1177 } 1178 1179 strcpy(noext[noext_cnt], name); 1180 F2FS_OPTION(sbi).nocompress_ext_cnt++; 1181 kfree(name); 1182 break; 1183 case Opt_compress_chksum: 1184 F2FS_OPTION(sbi).compress_chksum = true; 1185 break; 1186 case Opt_compress_mode: 1187 name = match_strdup(&args[0]); 1188 if (!name) 1189 return -ENOMEM; 1190 if (!strcmp(name, "fs")) { 1191 F2FS_OPTION(sbi).compress_mode = COMPR_MODE_FS; 1192 } else if (!strcmp(name, "user")) { 1193 F2FS_OPTION(sbi).compress_mode = COMPR_MODE_USER; 1194 } else { 1195 kfree(name); 1196 return -EINVAL; 1197 } 1198 kfree(name); 1199 break; 1200 case Opt_compress_cache: 1201 set_opt(sbi, COMPRESS_CACHE); 1202 break; 1203 #else 1204 case Opt_compress_algorithm: 1205 case Opt_compress_log_size: 1206 case Opt_compress_extension: 1207 case Opt_nocompress_extension: 1208 case Opt_compress_chksum: 1209 case Opt_compress_mode: 1210 case Opt_compress_cache: 1211 f2fs_info(sbi, "compression options not supported"); 1212 break; 1213 #endif 1214 case Opt_atgc: 1215 set_opt(sbi, ATGC); 1216 break; 1217 case Opt_gc_merge: 1218 set_opt(sbi, GC_MERGE); 1219 break; 1220 case Opt_nogc_merge: 1221 clear_opt(sbi, GC_MERGE); 1222 break; 1223 case Opt_discard_unit: 1224 name = match_strdup(&args[0]); 1225 if (!name) 1226 return -ENOMEM; 1227 if (!strcmp(name, "block")) { 1228 F2FS_OPTION(sbi).discard_unit = 1229 DISCARD_UNIT_BLOCK; 1230 } else if (!strcmp(name, "segment")) { 1231 F2FS_OPTION(sbi).discard_unit = 1232 DISCARD_UNIT_SEGMENT; 1233 } else if (!strcmp(name, "section")) { 1234 F2FS_OPTION(sbi).discard_unit = 1235 DISCARD_UNIT_SECTION; 1236 } else { 1237 kfree(name); 1238 return -EINVAL; 1239 } 1240 kfree(name); 1241 break; 1242 default: 1243 f2fs_err(sbi, "Unrecognized mount option \"%s\" or missing value", 1244 p); 1245 return -EINVAL; 1246 } 1247 } 1248 default_check: 1249 #ifdef CONFIG_QUOTA 1250 if (f2fs_check_quota_options(sbi)) 1251 return -EINVAL; 1252 #else 1253 if (f2fs_sb_has_quota_ino(sbi) && !f2fs_readonly(sbi->sb)) { 1254 f2fs_info(sbi, "Filesystem with quota feature cannot be mounted RDWR without CONFIG_QUOTA"); 1255 return -EINVAL; 1256 } 1257 if (f2fs_sb_has_project_quota(sbi) && !f2fs_readonly(sbi->sb)) { 1258 f2fs_err(sbi, "Filesystem with project quota feature cannot be mounted RDWR without CONFIG_QUOTA"); 1259 return -EINVAL; 1260 } 1261 #endif 1262 #if !IS_ENABLED(CONFIG_UNICODE) 1263 if (f2fs_sb_has_casefold(sbi)) { 1264 f2fs_err(sbi, 1265 "Filesystem with casefold feature cannot be mounted without CONFIG_UNICODE"); 1266 return -EINVAL; 1267 } 1268 #endif 1269 /* 1270 * The BLKZONED feature indicates that the drive was formatted with 1271 * zone alignment optimization. This is optional for host-aware 1272 * devices, but mandatory for host-managed zoned block devices. 1273 */ 1274 #ifndef CONFIG_BLK_DEV_ZONED 1275 if (f2fs_sb_has_blkzoned(sbi)) { 1276 f2fs_err(sbi, "Zoned block device support is not enabled"); 1277 return -EINVAL; 1278 } 1279 #endif 1280 if (f2fs_sb_has_blkzoned(sbi)) { 1281 if (F2FS_OPTION(sbi).discard_unit != 1282 DISCARD_UNIT_SECTION) { 1283 f2fs_info(sbi, "Zoned block device doesn't need small discard, set discard_unit=section by default"); 1284 F2FS_OPTION(sbi).discard_unit = 1285 DISCARD_UNIT_SECTION; 1286 } 1287 } 1288 1289 #ifdef CONFIG_F2FS_FS_COMPRESSION 1290 if (f2fs_test_compress_extension(sbi)) { 1291 f2fs_err(sbi, "invalid compress or nocompress extension"); 1292 return -EINVAL; 1293 } 1294 #endif 1295 1296 if (F2FS_IO_SIZE_BITS(sbi) && !f2fs_lfs_mode(sbi)) { 1297 f2fs_err(sbi, "Should set mode=lfs with %uKB-sized IO", 1298 F2FS_IO_SIZE_KB(sbi)); 1299 return -EINVAL; 1300 } 1301 1302 if (test_opt(sbi, INLINE_XATTR_SIZE)) { 1303 int min_size, max_size; 1304 1305 if (!f2fs_sb_has_extra_attr(sbi) || 1306 !f2fs_sb_has_flexible_inline_xattr(sbi)) { 1307 f2fs_err(sbi, "extra_attr or flexible_inline_xattr feature is off"); 1308 return -EINVAL; 1309 } 1310 if (!test_opt(sbi, INLINE_XATTR)) { 1311 f2fs_err(sbi, "inline_xattr_size option should be set with inline_xattr option"); 1312 return -EINVAL; 1313 } 1314 1315 min_size = sizeof(struct f2fs_xattr_header) / sizeof(__le32); 1316 max_size = MAX_INLINE_XATTR_SIZE; 1317 1318 if (F2FS_OPTION(sbi).inline_xattr_size < min_size || 1319 F2FS_OPTION(sbi).inline_xattr_size > max_size) { 1320 f2fs_err(sbi, "inline xattr size is out of range: %d ~ %d", 1321 min_size, max_size); 1322 return -EINVAL; 1323 } 1324 } 1325 1326 if (test_opt(sbi, DISABLE_CHECKPOINT) && f2fs_lfs_mode(sbi)) { 1327 f2fs_err(sbi, "LFS not compatible with checkpoint=disable"); 1328 return -EINVAL; 1329 } 1330 1331 /* Not pass down write hints if the number of active logs is lesser 1332 * than NR_CURSEG_PERSIST_TYPE. 1333 */ 1334 if (F2FS_OPTION(sbi).active_logs != NR_CURSEG_PERSIST_TYPE) 1335 F2FS_OPTION(sbi).whint_mode = WHINT_MODE_OFF; 1336 1337 if (f2fs_sb_has_readonly(sbi) && !f2fs_readonly(sbi->sb)) { 1338 f2fs_err(sbi, "Allow to mount readonly mode only"); 1339 return -EROFS; 1340 } 1341 return 0; 1342 } 1343 1344 static struct inode *f2fs_alloc_inode(struct super_block *sb) 1345 { 1346 struct f2fs_inode_info *fi; 1347 1348 fi = f2fs_kmem_cache_alloc(f2fs_inode_cachep, 1349 GFP_F2FS_ZERO, false, F2FS_SB(sb)); 1350 if (!fi) 1351 return NULL; 1352 1353 init_once((void *) fi); 1354 1355 /* Initialize f2fs-specific inode info */ 1356 atomic_set(&fi->dirty_pages, 0); 1357 atomic_set(&fi->i_compr_blocks, 0); 1358 init_rwsem(&fi->i_sem); 1359 spin_lock_init(&fi->i_size_lock); 1360 INIT_LIST_HEAD(&fi->dirty_list); 1361 INIT_LIST_HEAD(&fi->gdirty_list); 1362 INIT_LIST_HEAD(&fi->inmem_ilist); 1363 INIT_LIST_HEAD(&fi->inmem_pages); 1364 mutex_init(&fi->inmem_lock); 1365 init_rwsem(&fi->i_gc_rwsem[READ]); 1366 init_rwsem(&fi->i_gc_rwsem[WRITE]); 1367 init_rwsem(&fi->i_xattr_sem); 1368 1369 /* Will be used by directory only */ 1370 fi->i_dir_level = F2FS_SB(sb)->dir_level; 1371 1372 return &fi->vfs_inode; 1373 } 1374 1375 static int f2fs_drop_inode(struct inode *inode) 1376 { 1377 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1378 int ret; 1379 1380 /* 1381 * during filesystem shutdown, if checkpoint is disabled, 1382 * drop useless meta/node dirty pages. 1383 */ 1384 if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED))) { 1385 if (inode->i_ino == F2FS_NODE_INO(sbi) || 1386 inode->i_ino == F2FS_META_INO(sbi)) { 1387 trace_f2fs_drop_inode(inode, 1); 1388 return 1; 1389 } 1390 } 1391 1392 /* 1393 * This is to avoid a deadlock condition like below. 1394 * writeback_single_inode(inode) 1395 * - f2fs_write_data_page 1396 * - f2fs_gc -> iput -> evict 1397 * - inode_wait_for_writeback(inode) 1398 */ 1399 if ((!inode_unhashed(inode) && inode->i_state & I_SYNC)) { 1400 if (!inode->i_nlink && !is_bad_inode(inode)) { 1401 /* to avoid evict_inode call simultaneously */ 1402 atomic_inc(&inode->i_count); 1403 spin_unlock(&inode->i_lock); 1404 1405 /* some remained atomic pages should discarded */ 1406 if (f2fs_is_atomic_file(inode)) 1407 f2fs_drop_inmem_pages(inode); 1408 1409 /* should remain fi->extent_tree for writepage */ 1410 f2fs_destroy_extent_node(inode); 1411 1412 sb_start_intwrite(inode->i_sb); 1413 f2fs_i_size_write(inode, 0); 1414 1415 f2fs_submit_merged_write_cond(F2FS_I_SB(inode), 1416 inode, NULL, 0, DATA); 1417 truncate_inode_pages_final(inode->i_mapping); 1418 1419 if (F2FS_HAS_BLOCKS(inode)) 1420 f2fs_truncate(inode); 1421 1422 sb_end_intwrite(inode->i_sb); 1423 1424 spin_lock(&inode->i_lock); 1425 atomic_dec(&inode->i_count); 1426 } 1427 trace_f2fs_drop_inode(inode, 0); 1428 return 0; 1429 } 1430 ret = generic_drop_inode(inode); 1431 if (!ret) 1432 ret = fscrypt_drop_inode(inode); 1433 trace_f2fs_drop_inode(inode, ret); 1434 return ret; 1435 } 1436 1437 int f2fs_inode_dirtied(struct inode *inode, bool sync) 1438 { 1439 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1440 int ret = 0; 1441 1442 spin_lock(&sbi->inode_lock[DIRTY_META]); 1443 if (is_inode_flag_set(inode, FI_DIRTY_INODE)) { 1444 ret = 1; 1445 } else { 1446 set_inode_flag(inode, FI_DIRTY_INODE); 1447 stat_inc_dirty_inode(sbi, DIRTY_META); 1448 } 1449 if (sync && list_empty(&F2FS_I(inode)->gdirty_list)) { 1450 list_add_tail(&F2FS_I(inode)->gdirty_list, 1451 &sbi->inode_list[DIRTY_META]); 1452 inc_page_count(sbi, F2FS_DIRTY_IMETA); 1453 } 1454 spin_unlock(&sbi->inode_lock[DIRTY_META]); 1455 return ret; 1456 } 1457 1458 void f2fs_inode_synced(struct inode *inode) 1459 { 1460 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1461 1462 spin_lock(&sbi->inode_lock[DIRTY_META]); 1463 if (!is_inode_flag_set(inode, FI_DIRTY_INODE)) { 1464 spin_unlock(&sbi->inode_lock[DIRTY_META]); 1465 return; 1466 } 1467 if (!list_empty(&F2FS_I(inode)->gdirty_list)) { 1468 list_del_init(&F2FS_I(inode)->gdirty_list); 1469 dec_page_count(sbi, F2FS_DIRTY_IMETA); 1470 } 1471 clear_inode_flag(inode, FI_DIRTY_INODE); 1472 clear_inode_flag(inode, FI_AUTO_RECOVER); 1473 stat_dec_dirty_inode(F2FS_I_SB(inode), DIRTY_META); 1474 spin_unlock(&sbi->inode_lock[DIRTY_META]); 1475 } 1476 1477 /* 1478 * f2fs_dirty_inode() is called from __mark_inode_dirty() 1479 * 1480 * We should call set_dirty_inode to write the dirty inode through write_inode. 1481 */ 1482 static void f2fs_dirty_inode(struct inode *inode, int flags) 1483 { 1484 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1485 1486 if (inode->i_ino == F2FS_NODE_INO(sbi) || 1487 inode->i_ino == F2FS_META_INO(sbi)) 1488 return; 1489 1490 if (is_inode_flag_set(inode, FI_AUTO_RECOVER)) 1491 clear_inode_flag(inode, FI_AUTO_RECOVER); 1492 1493 f2fs_inode_dirtied(inode, false); 1494 } 1495 1496 static void f2fs_free_inode(struct inode *inode) 1497 { 1498 fscrypt_free_inode(inode); 1499 kmem_cache_free(f2fs_inode_cachep, F2FS_I(inode)); 1500 } 1501 1502 static void destroy_percpu_info(struct f2fs_sb_info *sbi) 1503 { 1504 percpu_counter_destroy(&sbi->alloc_valid_block_count); 1505 percpu_counter_destroy(&sbi->total_valid_inode_count); 1506 } 1507 1508 static void destroy_device_list(struct f2fs_sb_info *sbi) 1509 { 1510 int i; 1511 1512 for (i = 0; i < sbi->s_ndevs; i++) { 1513 blkdev_put(FDEV(i).bdev, FMODE_EXCL); 1514 #ifdef CONFIG_BLK_DEV_ZONED 1515 kvfree(FDEV(i).blkz_seq); 1516 kfree(FDEV(i).zone_capacity_blocks); 1517 #endif 1518 } 1519 kvfree(sbi->devs); 1520 } 1521 1522 static void f2fs_put_super(struct super_block *sb) 1523 { 1524 struct f2fs_sb_info *sbi = F2FS_SB(sb); 1525 int i; 1526 bool dropped; 1527 1528 /* unregister procfs/sysfs entries in advance to avoid race case */ 1529 f2fs_unregister_sysfs(sbi); 1530 1531 f2fs_quota_off_umount(sb); 1532 1533 /* prevent remaining shrinker jobs */ 1534 mutex_lock(&sbi->umount_mutex); 1535 1536 /* 1537 * flush all issued checkpoints and stop checkpoint issue thread. 1538 * after then, all checkpoints should be done by each process context. 1539 */ 1540 f2fs_stop_ckpt_thread(sbi); 1541 1542 /* 1543 * We don't need to do checkpoint when superblock is clean. 1544 * But, the previous checkpoint was not done by umount, it needs to do 1545 * clean checkpoint again. 1546 */ 1547 if ((is_sbi_flag_set(sbi, SBI_IS_DIRTY) || 1548 !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG))) { 1549 struct cp_control cpc = { 1550 .reason = CP_UMOUNT, 1551 }; 1552 f2fs_write_checkpoint(sbi, &cpc); 1553 } 1554 1555 /* be sure to wait for any on-going discard commands */ 1556 dropped = f2fs_issue_discard_timeout(sbi); 1557 1558 if ((f2fs_hw_support_discard(sbi) || f2fs_hw_should_discard(sbi)) && 1559 !sbi->discard_blks && !dropped) { 1560 struct cp_control cpc = { 1561 .reason = CP_UMOUNT | CP_TRIMMED, 1562 }; 1563 f2fs_write_checkpoint(sbi, &cpc); 1564 } 1565 1566 /* 1567 * normally superblock is clean, so we need to release this. 1568 * In addition, EIO will skip do checkpoint, we need this as well. 1569 */ 1570 f2fs_release_ino_entry(sbi, true); 1571 1572 f2fs_leave_shrinker(sbi); 1573 mutex_unlock(&sbi->umount_mutex); 1574 1575 /* our cp_error case, we can wait for any writeback page */ 1576 f2fs_flush_merged_writes(sbi); 1577 1578 f2fs_wait_on_all_pages(sbi, F2FS_WB_CP_DATA); 1579 1580 f2fs_bug_on(sbi, sbi->fsync_node_num); 1581 1582 f2fs_destroy_compress_inode(sbi); 1583 1584 iput(sbi->node_inode); 1585 sbi->node_inode = NULL; 1586 1587 iput(sbi->meta_inode); 1588 sbi->meta_inode = NULL; 1589 1590 /* 1591 * iput() can update stat information, if f2fs_write_checkpoint() 1592 * above failed with error. 1593 */ 1594 f2fs_destroy_stats(sbi); 1595 1596 /* destroy f2fs internal modules */ 1597 f2fs_destroy_node_manager(sbi); 1598 f2fs_destroy_segment_manager(sbi); 1599 1600 f2fs_destroy_post_read_wq(sbi); 1601 1602 kvfree(sbi->ckpt); 1603 1604 sb->s_fs_info = NULL; 1605 if (sbi->s_chksum_driver) 1606 crypto_free_shash(sbi->s_chksum_driver); 1607 kfree(sbi->raw_super); 1608 1609 destroy_device_list(sbi); 1610 f2fs_destroy_page_array_cache(sbi); 1611 f2fs_destroy_xattr_caches(sbi); 1612 mempool_destroy(sbi->write_io_dummy); 1613 #ifdef CONFIG_QUOTA 1614 for (i = 0; i < MAXQUOTAS; i++) 1615 kfree(F2FS_OPTION(sbi).s_qf_names[i]); 1616 #endif 1617 fscrypt_free_dummy_policy(&F2FS_OPTION(sbi).dummy_enc_policy); 1618 destroy_percpu_info(sbi); 1619 f2fs_destroy_iostat(sbi); 1620 for (i = 0; i < NR_PAGE_TYPE; i++) 1621 kvfree(sbi->write_io[i]); 1622 #if IS_ENABLED(CONFIG_UNICODE) 1623 utf8_unload(sb->s_encoding); 1624 #endif 1625 kfree(sbi); 1626 } 1627 1628 int f2fs_sync_fs(struct super_block *sb, int sync) 1629 { 1630 struct f2fs_sb_info *sbi = F2FS_SB(sb); 1631 int err = 0; 1632 1633 if (unlikely(f2fs_cp_error(sbi))) 1634 return 0; 1635 if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED))) 1636 return 0; 1637 1638 trace_f2fs_sync_fs(sb, sync); 1639 1640 if (unlikely(is_sbi_flag_set(sbi, SBI_POR_DOING))) 1641 return -EAGAIN; 1642 1643 if (sync) 1644 err = f2fs_issue_checkpoint(sbi); 1645 1646 return err; 1647 } 1648 1649 static int f2fs_freeze(struct super_block *sb) 1650 { 1651 if (f2fs_readonly(sb)) 1652 return 0; 1653 1654 /* IO error happened before */ 1655 if (unlikely(f2fs_cp_error(F2FS_SB(sb)))) 1656 return -EIO; 1657 1658 /* must be clean, since sync_filesystem() was already called */ 1659 if (is_sbi_flag_set(F2FS_SB(sb), SBI_IS_DIRTY)) 1660 return -EINVAL; 1661 1662 /* ensure no checkpoint required */ 1663 if (!llist_empty(&F2FS_SB(sb)->cprc_info.issue_list)) 1664 return -EINVAL; 1665 return 0; 1666 } 1667 1668 static int f2fs_unfreeze(struct super_block *sb) 1669 { 1670 return 0; 1671 } 1672 1673 #ifdef CONFIG_QUOTA 1674 static int f2fs_statfs_project(struct super_block *sb, 1675 kprojid_t projid, struct kstatfs *buf) 1676 { 1677 struct kqid qid; 1678 struct dquot *dquot; 1679 u64 limit; 1680 u64 curblock; 1681 1682 qid = make_kqid_projid(projid); 1683 dquot = dqget(sb, qid); 1684 if (IS_ERR(dquot)) 1685 return PTR_ERR(dquot); 1686 spin_lock(&dquot->dq_dqb_lock); 1687 1688 limit = min_not_zero(dquot->dq_dqb.dqb_bsoftlimit, 1689 dquot->dq_dqb.dqb_bhardlimit); 1690 if (limit) 1691 limit >>= sb->s_blocksize_bits; 1692 1693 if (limit && buf->f_blocks > limit) { 1694 curblock = (dquot->dq_dqb.dqb_curspace + 1695 dquot->dq_dqb.dqb_rsvspace) >> sb->s_blocksize_bits; 1696 buf->f_blocks = limit; 1697 buf->f_bfree = buf->f_bavail = 1698 (buf->f_blocks > curblock) ? 1699 (buf->f_blocks - curblock) : 0; 1700 } 1701 1702 limit = min_not_zero(dquot->dq_dqb.dqb_isoftlimit, 1703 dquot->dq_dqb.dqb_ihardlimit); 1704 1705 if (limit && buf->f_files > limit) { 1706 buf->f_files = limit; 1707 buf->f_ffree = 1708 (buf->f_files > dquot->dq_dqb.dqb_curinodes) ? 1709 (buf->f_files - dquot->dq_dqb.dqb_curinodes) : 0; 1710 } 1711 1712 spin_unlock(&dquot->dq_dqb_lock); 1713 dqput(dquot); 1714 return 0; 1715 } 1716 #endif 1717 1718 static int f2fs_statfs(struct dentry *dentry, struct kstatfs *buf) 1719 { 1720 struct super_block *sb = dentry->d_sb; 1721 struct f2fs_sb_info *sbi = F2FS_SB(sb); 1722 u64 id = huge_encode_dev(sb->s_bdev->bd_dev); 1723 block_t total_count, user_block_count, start_count; 1724 u64 avail_node_count; 1725 1726 total_count = le64_to_cpu(sbi->raw_super->block_count); 1727 user_block_count = sbi->user_block_count; 1728 start_count = le32_to_cpu(sbi->raw_super->segment0_blkaddr); 1729 buf->f_type = F2FS_SUPER_MAGIC; 1730 buf->f_bsize = sbi->blocksize; 1731 1732 buf->f_blocks = total_count - start_count; 1733 buf->f_bfree = user_block_count - valid_user_blocks(sbi) - 1734 sbi->current_reserved_blocks; 1735 1736 spin_lock(&sbi->stat_lock); 1737 if (unlikely(buf->f_bfree <= sbi->unusable_block_count)) 1738 buf->f_bfree = 0; 1739 else 1740 buf->f_bfree -= sbi->unusable_block_count; 1741 spin_unlock(&sbi->stat_lock); 1742 1743 if (buf->f_bfree > F2FS_OPTION(sbi).root_reserved_blocks) 1744 buf->f_bavail = buf->f_bfree - 1745 F2FS_OPTION(sbi).root_reserved_blocks; 1746 else 1747 buf->f_bavail = 0; 1748 1749 avail_node_count = sbi->total_node_count - F2FS_RESERVED_NODE_NUM; 1750 1751 if (avail_node_count > user_block_count) { 1752 buf->f_files = user_block_count; 1753 buf->f_ffree = buf->f_bavail; 1754 } else { 1755 buf->f_files = avail_node_count; 1756 buf->f_ffree = min(avail_node_count - valid_node_count(sbi), 1757 buf->f_bavail); 1758 } 1759 1760 buf->f_namelen = F2FS_NAME_LEN; 1761 buf->f_fsid = u64_to_fsid(id); 1762 1763 #ifdef CONFIG_QUOTA 1764 if (is_inode_flag_set(dentry->d_inode, FI_PROJ_INHERIT) && 1765 sb_has_quota_limits_enabled(sb, PRJQUOTA)) { 1766 f2fs_statfs_project(sb, F2FS_I(dentry->d_inode)->i_projid, buf); 1767 } 1768 #endif 1769 return 0; 1770 } 1771 1772 static inline void f2fs_show_quota_options(struct seq_file *seq, 1773 struct super_block *sb) 1774 { 1775 #ifdef CONFIG_QUOTA 1776 struct f2fs_sb_info *sbi = F2FS_SB(sb); 1777 1778 if (F2FS_OPTION(sbi).s_jquota_fmt) { 1779 char *fmtname = ""; 1780 1781 switch (F2FS_OPTION(sbi).s_jquota_fmt) { 1782 case QFMT_VFS_OLD: 1783 fmtname = "vfsold"; 1784 break; 1785 case QFMT_VFS_V0: 1786 fmtname = "vfsv0"; 1787 break; 1788 case QFMT_VFS_V1: 1789 fmtname = "vfsv1"; 1790 break; 1791 } 1792 seq_printf(seq, ",jqfmt=%s", fmtname); 1793 } 1794 1795 if (F2FS_OPTION(sbi).s_qf_names[USRQUOTA]) 1796 seq_show_option(seq, "usrjquota", 1797 F2FS_OPTION(sbi).s_qf_names[USRQUOTA]); 1798 1799 if (F2FS_OPTION(sbi).s_qf_names[GRPQUOTA]) 1800 seq_show_option(seq, "grpjquota", 1801 F2FS_OPTION(sbi).s_qf_names[GRPQUOTA]); 1802 1803 if (F2FS_OPTION(sbi).s_qf_names[PRJQUOTA]) 1804 seq_show_option(seq, "prjjquota", 1805 F2FS_OPTION(sbi).s_qf_names[PRJQUOTA]); 1806 #endif 1807 } 1808 1809 #ifdef CONFIG_F2FS_FS_COMPRESSION 1810 static inline void f2fs_show_compress_options(struct seq_file *seq, 1811 struct super_block *sb) 1812 { 1813 struct f2fs_sb_info *sbi = F2FS_SB(sb); 1814 char *algtype = ""; 1815 int i; 1816 1817 if (!f2fs_sb_has_compression(sbi)) 1818 return; 1819 1820 switch (F2FS_OPTION(sbi).compress_algorithm) { 1821 case COMPRESS_LZO: 1822 algtype = "lzo"; 1823 break; 1824 case COMPRESS_LZ4: 1825 algtype = "lz4"; 1826 break; 1827 case COMPRESS_ZSTD: 1828 algtype = "zstd"; 1829 break; 1830 case COMPRESS_LZORLE: 1831 algtype = "lzo-rle"; 1832 break; 1833 } 1834 seq_printf(seq, ",compress_algorithm=%s", algtype); 1835 1836 if (F2FS_OPTION(sbi).compress_level) 1837 seq_printf(seq, ":%d", F2FS_OPTION(sbi).compress_level); 1838 1839 seq_printf(seq, ",compress_log_size=%u", 1840 F2FS_OPTION(sbi).compress_log_size); 1841 1842 for (i = 0; i < F2FS_OPTION(sbi).compress_ext_cnt; i++) { 1843 seq_printf(seq, ",compress_extension=%s", 1844 F2FS_OPTION(sbi).extensions[i]); 1845 } 1846 1847 for (i = 0; i < F2FS_OPTION(sbi).nocompress_ext_cnt; i++) { 1848 seq_printf(seq, ",nocompress_extension=%s", 1849 F2FS_OPTION(sbi).noextensions[i]); 1850 } 1851 1852 if (F2FS_OPTION(sbi).compress_chksum) 1853 seq_puts(seq, ",compress_chksum"); 1854 1855 if (F2FS_OPTION(sbi).compress_mode == COMPR_MODE_FS) 1856 seq_printf(seq, ",compress_mode=%s", "fs"); 1857 else if (F2FS_OPTION(sbi).compress_mode == COMPR_MODE_USER) 1858 seq_printf(seq, ",compress_mode=%s", "user"); 1859 1860 if (test_opt(sbi, COMPRESS_CACHE)) 1861 seq_puts(seq, ",compress_cache"); 1862 } 1863 #endif 1864 1865 static int f2fs_show_options(struct seq_file *seq, struct dentry *root) 1866 { 1867 struct f2fs_sb_info *sbi = F2FS_SB(root->d_sb); 1868 1869 if (F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_SYNC) 1870 seq_printf(seq, ",background_gc=%s", "sync"); 1871 else if (F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_ON) 1872 seq_printf(seq, ",background_gc=%s", "on"); 1873 else if (F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_OFF) 1874 seq_printf(seq, ",background_gc=%s", "off"); 1875 1876 if (test_opt(sbi, GC_MERGE)) 1877 seq_puts(seq, ",gc_merge"); 1878 1879 if (test_opt(sbi, DISABLE_ROLL_FORWARD)) 1880 seq_puts(seq, ",disable_roll_forward"); 1881 if (test_opt(sbi, NORECOVERY)) 1882 seq_puts(seq, ",norecovery"); 1883 if (test_opt(sbi, DISCARD)) 1884 seq_puts(seq, ",discard"); 1885 else 1886 seq_puts(seq, ",nodiscard"); 1887 if (test_opt(sbi, NOHEAP)) 1888 seq_puts(seq, ",no_heap"); 1889 else 1890 seq_puts(seq, ",heap"); 1891 #ifdef CONFIG_F2FS_FS_XATTR 1892 if (test_opt(sbi, XATTR_USER)) 1893 seq_puts(seq, ",user_xattr"); 1894 else 1895 seq_puts(seq, ",nouser_xattr"); 1896 if (test_opt(sbi, INLINE_XATTR)) 1897 seq_puts(seq, ",inline_xattr"); 1898 else 1899 seq_puts(seq, ",noinline_xattr"); 1900 if (test_opt(sbi, INLINE_XATTR_SIZE)) 1901 seq_printf(seq, ",inline_xattr_size=%u", 1902 F2FS_OPTION(sbi).inline_xattr_size); 1903 #endif 1904 #ifdef CONFIG_F2FS_FS_POSIX_ACL 1905 if (test_opt(sbi, POSIX_ACL)) 1906 seq_puts(seq, ",acl"); 1907 else 1908 seq_puts(seq, ",noacl"); 1909 #endif 1910 if (test_opt(sbi, DISABLE_EXT_IDENTIFY)) 1911 seq_puts(seq, ",disable_ext_identify"); 1912 if (test_opt(sbi, INLINE_DATA)) 1913 seq_puts(seq, ",inline_data"); 1914 else 1915 seq_puts(seq, ",noinline_data"); 1916 if (test_opt(sbi, INLINE_DENTRY)) 1917 seq_puts(seq, ",inline_dentry"); 1918 else 1919 seq_puts(seq, ",noinline_dentry"); 1920 if (!f2fs_readonly(sbi->sb) && test_opt(sbi, FLUSH_MERGE)) 1921 seq_puts(seq, ",flush_merge"); 1922 if (test_opt(sbi, NOBARRIER)) 1923 seq_puts(seq, ",nobarrier"); 1924 if (test_opt(sbi, FASTBOOT)) 1925 seq_puts(seq, ",fastboot"); 1926 if (test_opt(sbi, EXTENT_CACHE)) 1927 seq_puts(seq, ",extent_cache"); 1928 else 1929 seq_puts(seq, ",noextent_cache"); 1930 if (test_opt(sbi, DATA_FLUSH)) 1931 seq_puts(seq, ",data_flush"); 1932 1933 seq_puts(seq, ",mode="); 1934 if (F2FS_OPTION(sbi).fs_mode == FS_MODE_ADAPTIVE) 1935 seq_puts(seq, "adaptive"); 1936 else if (F2FS_OPTION(sbi).fs_mode == FS_MODE_LFS) 1937 seq_puts(seq, "lfs"); 1938 else if (F2FS_OPTION(sbi).fs_mode == FS_MODE_FRAGMENT_SEG) 1939 seq_puts(seq, "fragment:segment"); 1940 else if (F2FS_OPTION(sbi).fs_mode == FS_MODE_FRAGMENT_BLK) 1941 seq_puts(seq, "fragment:block"); 1942 seq_printf(seq, ",active_logs=%u", F2FS_OPTION(sbi).active_logs); 1943 if (test_opt(sbi, RESERVE_ROOT)) 1944 seq_printf(seq, ",reserve_root=%u,resuid=%u,resgid=%u", 1945 F2FS_OPTION(sbi).root_reserved_blocks, 1946 from_kuid_munged(&init_user_ns, 1947 F2FS_OPTION(sbi).s_resuid), 1948 from_kgid_munged(&init_user_ns, 1949 F2FS_OPTION(sbi).s_resgid)); 1950 if (F2FS_IO_SIZE_BITS(sbi)) 1951 seq_printf(seq, ",io_bits=%u", 1952 F2FS_OPTION(sbi).write_io_size_bits); 1953 #ifdef CONFIG_F2FS_FAULT_INJECTION 1954 if (test_opt(sbi, FAULT_INJECTION)) { 1955 seq_printf(seq, ",fault_injection=%u", 1956 F2FS_OPTION(sbi).fault_info.inject_rate); 1957 seq_printf(seq, ",fault_type=%u", 1958 F2FS_OPTION(sbi).fault_info.inject_type); 1959 } 1960 #endif 1961 #ifdef CONFIG_QUOTA 1962 if (test_opt(sbi, QUOTA)) 1963 seq_puts(seq, ",quota"); 1964 if (test_opt(sbi, USRQUOTA)) 1965 seq_puts(seq, ",usrquota"); 1966 if (test_opt(sbi, GRPQUOTA)) 1967 seq_puts(seq, ",grpquota"); 1968 if (test_opt(sbi, PRJQUOTA)) 1969 seq_puts(seq, ",prjquota"); 1970 #endif 1971 f2fs_show_quota_options(seq, sbi->sb); 1972 if (F2FS_OPTION(sbi).whint_mode == WHINT_MODE_USER) 1973 seq_printf(seq, ",whint_mode=%s", "user-based"); 1974 else if (F2FS_OPTION(sbi).whint_mode == WHINT_MODE_FS) 1975 seq_printf(seq, ",whint_mode=%s", "fs-based"); 1976 1977 fscrypt_show_test_dummy_encryption(seq, ',', sbi->sb); 1978 1979 if (sbi->sb->s_flags & SB_INLINECRYPT) 1980 seq_puts(seq, ",inlinecrypt"); 1981 1982 if (F2FS_OPTION(sbi).alloc_mode == ALLOC_MODE_DEFAULT) 1983 seq_printf(seq, ",alloc_mode=%s", "default"); 1984 else if (F2FS_OPTION(sbi).alloc_mode == ALLOC_MODE_REUSE) 1985 seq_printf(seq, ",alloc_mode=%s", "reuse"); 1986 1987 if (test_opt(sbi, DISABLE_CHECKPOINT)) 1988 seq_printf(seq, ",checkpoint=disable:%u", 1989 F2FS_OPTION(sbi).unusable_cap); 1990 if (test_opt(sbi, MERGE_CHECKPOINT)) 1991 seq_puts(seq, ",checkpoint_merge"); 1992 else 1993 seq_puts(seq, ",nocheckpoint_merge"); 1994 if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_POSIX) 1995 seq_printf(seq, ",fsync_mode=%s", "posix"); 1996 else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_STRICT) 1997 seq_printf(seq, ",fsync_mode=%s", "strict"); 1998 else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_NOBARRIER) 1999 seq_printf(seq, ",fsync_mode=%s", "nobarrier"); 2000 2001 #ifdef CONFIG_F2FS_FS_COMPRESSION 2002 f2fs_show_compress_options(seq, sbi->sb); 2003 #endif 2004 2005 if (test_opt(sbi, ATGC)) 2006 seq_puts(seq, ",atgc"); 2007 2008 if (F2FS_OPTION(sbi).discard_unit == DISCARD_UNIT_BLOCK) 2009 seq_printf(seq, ",discard_unit=%s", "block"); 2010 else if (F2FS_OPTION(sbi).discard_unit == DISCARD_UNIT_SEGMENT) 2011 seq_printf(seq, ",discard_unit=%s", "segment"); 2012 else if (F2FS_OPTION(sbi).discard_unit == DISCARD_UNIT_SECTION) 2013 seq_printf(seq, ",discard_unit=%s", "section"); 2014 2015 return 0; 2016 } 2017 2018 static void default_options(struct f2fs_sb_info *sbi) 2019 { 2020 /* init some FS parameters */ 2021 if (f2fs_sb_has_readonly(sbi)) 2022 F2FS_OPTION(sbi).active_logs = NR_CURSEG_RO_TYPE; 2023 else 2024 F2FS_OPTION(sbi).active_logs = NR_CURSEG_PERSIST_TYPE; 2025 2026 F2FS_OPTION(sbi).inline_xattr_size = DEFAULT_INLINE_XATTR_ADDRS; 2027 F2FS_OPTION(sbi).whint_mode = WHINT_MODE_OFF; 2028 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_DEFAULT; 2029 F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_POSIX; 2030 F2FS_OPTION(sbi).s_resuid = make_kuid(&init_user_ns, F2FS_DEF_RESUID); 2031 F2FS_OPTION(sbi).s_resgid = make_kgid(&init_user_ns, F2FS_DEF_RESGID); 2032 F2FS_OPTION(sbi).compress_algorithm = COMPRESS_LZ4; 2033 F2FS_OPTION(sbi).compress_log_size = MIN_COMPRESS_LOG_SIZE; 2034 F2FS_OPTION(sbi).compress_ext_cnt = 0; 2035 F2FS_OPTION(sbi).compress_mode = COMPR_MODE_FS; 2036 F2FS_OPTION(sbi).bggc_mode = BGGC_MODE_ON; 2037 2038 sbi->sb->s_flags &= ~SB_INLINECRYPT; 2039 2040 set_opt(sbi, INLINE_XATTR); 2041 set_opt(sbi, INLINE_DATA); 2042 set_opt(sbi, INLINE_DENTRY); 2043 set_opt(sbi, EXTENT_CACHE); 2044 set_opt(sbi, NOHEAP); 2045 clear_opt(sbi, DISABLE_CHECKPOINT); 2046 set_opt(sbi, MERGE_CHECKPOINT); 2047 F2FS_OPTION(sbi).unusable_cap = 0; 2048 sbi->sb->s_flags |= SB_LAZYTIME; 2049 set_opt(sbi, FLUSH_MERGE); 2050 if (f2fs_hw_support_discard(sbi) || f2fs_hw_should_discard(sbi)) 2051 set_opt(sbi, DISCARD); 2052 if (f2fs_sb_has_blkzoned(sbi)) { 2053 F2FS_OPTION(sbi).fs_mode = FS_MODE_LFS; 2054 F2FS_OPTION(sbi).discard_unit = DISCARD_UNIT_SECTION; 2055 } else { 2056 F2FS_OPTION(sbi).fs_mode = FS_MODE_ADAPTIVE; 2057 F2FS_OPTION(sbi).discard_unit = DISCARD_UNIT_BLOCK; 2058 } 2059 2060 #ifdef CONFIG_F2FS_FS_XATTR 2061 set_opt(sbi, XATTR_USER); 2062 #endif 2063 #ifdef CONFIG_F2FS_FS_POSIX_ACL 2064 set_opt(sbi, POSIX_ACL); 2065 #endif 2066 2067 f2fs_build_fault_attr(sbi, 0, 0); 2068 } 2069 2070 #ifdef CONFIG_QUOTA 2071 static int f2fs_enable_quotas(struct super_block *sb); 2072 #endif 2073 2074 static int f2fs_disable_checkpoint(struct f2fs_sb_info *sbi) 2075 { 2076 unsigned int s_flags = sbi->sb->s_flags; 2077 struct cp_control cpc; 2078 int err = 0; 2079 int ret; 2080 block_t unusable; 2081 2082 if (s_flags & SB_RDONLY) { 2083 f2fs_err(sbi, "checkpoint=disable on readonly fs"); 2084 return -EINVAL; 2085 } 2086 sbi->sb->s_flags |= SB_ACTIVE; 2087 2088 f2fs_update_time(sbi, DISABLE_TIME); 2089 2090 while (!f2fs_time_over(sbi, DISABLE_TIME)) { 2091 down_write(&sbi->gc_lock); 2092 err = f2fs_gc(sbi, true, false, false, NULL_SEGNO); 2093 if (err == -ENODATA) { 2094 err = 0; 2095 break; 2096 } 2097 if (err && err != -EAGAIN) 2098 break; 2099 } 2100 2101 ret = sync_filesystem(sbi->sb); 2102 if (ret || err) { 2103 err = ret ? ret : err; 2104 goto restore_flag; 2105 } 2106 2107 unusable = f2fs_get_unusable_blocks(sbi); 2108 if (f2fs_disable_cp_again(sbi, unusable)) { 2109 err = -EAGAIN; 2110 goto restore_flag; 2111 } 2112 2113 down_write(&sbi->gc_lock); 2114 cpc.reason = CP_PAUSE; 2115 set_sbi_flag(sbi, SBI_CP_DISABLED); 2116 err = f2fs_write_checkpoint(sbi, &cpc); 2117 if (err) 2118 goto out_unlock; 2119 2120 spin_lock(&sbi->stat_lock); 2121 sbi->unusable_block_count = unusable; 2122 spin_unlock(&sbi->stat_lock); 2123 2124 out_unlock: 2125 up_write(&sbi->gc_lock); 2126 restore_flag: 2127 sbi->sb->s_flags = s_flags; /* Restore SB_RDONLY status */ 2128 return err; 2129 } 2130 2131 static void f2fs_enable_checkpoint(struct f2fs_sb_info *sbi) 2132 { 2133 int retry = DEFAULT_RETRY_IO_COUNT; 2134 2135 /* we should flush all the data to keep data consistency */ 2136 do { 2137 sync_inodes_sb(sbi->sb); 2138 cond_resched(); 2139 congestion_wait(BLK_RW_ASYNC, DEFAULT_IO_TIMEOUT); 2140 } while (get_pages(sbi, F2FS_DIRTY_DATA) && retry--); 2141 2142 if (unlikely(retry < 0)) 2143 f2fs_warn(sbi, "checkpoint=enable has some unwritten data."); 2144 2145 down_write(&sbi->gc_lock); 2146 f2fs_dirty_to_prefree(sbi); 2147 2148 clear_sbi_flag(sbi, SBI_CP_DISABLED); 2149 set_sbi_flag(sbi, SBI_IS_DIRTY); 2150 up_write(&sbi->gc_lock); 2151 2152 f2fs_sync_fs(sbi->sb, 1); 2153 } 2154 2155 static int f2fs_remount(struct super_block *sb, int *flags, char *data) 2156 { 2157 struct f2fs_sb_info *sbi = F2FS_SB(sb); 2158 struct f2fs_mount_info org_mount_opt; 2159 unsigned long old_sb_flags; 2160 int err; 2161 bool need_restart_gc = false, need_stop_gc = false; 2162 bool need_restart_ckpt = false, need_stop_ckpt = false; 2163 bool need_restart_flush = false, need_stop_flush = false; 2164 bool need_restart_discard = false, need_stop_discard = false; 2165 bool no_extent_cache = !test_opt(sbi, EXTENT_CACHE); 2166 bool enable_checkpoint = !test_opt(sbi, DISABLE_CHECKPOINT); 2167 bool no_io_align = !F2FS_IO_ALIGNED(sbi); 2168 bool no_atgc = !test_opt(sbi, ATGC); 2169 bool no_discard = !test_opt(sbi, DISCARD); 2170 bool no_compress_cache = !test_opt(sbi, COMPRESS_CACHE); 2171 bool block_unit_discard = f2fs_block_unit_discard(sbi); 2172 struct discard_cmd_control *dcc; 2173 #ifdef CONFIG_QUOTA 2174 int i, j; 2175 #endif 2176 2177 /* 2178 * Save the old mount options in case we 2179 * need to restore them. 2180 */ 2181 org_mount_opt = sbi->mount_opt; 2182 old_sb_flags = sb->s_flags; 2183 2184 #ifdef CONFIG_QUOTA 2185 org_mount_opt.s_jquota_fmt = F2FS_OPTION(sbi).s_jquota_fmt; 2186 for (i = 0; i < MAXQUOTAS; i++) { 2187 if (F2FS_OPTION(sbi).s_qf_names[i]) { 2188 org_mount_opt.s_qf_names[i] = 2189 kstrdup(F2FS_OPTION(sbi).s_qf_names[i], 2190 GFP_KERNEL); 2191 if (!org_mount_opt.s_qf_names[i]) { 2192 for (j = 0; j < i; j++) 2193 kfree(org_mount_opt.s_qf_names[j]); 2194 return -ENOMEM; 2195 } 2196 } else { 2197 org_mount_opt.s_qf_names[i] = NULL; 2198 } 2199 } 2200 #endif 2201 2202 /* recover superblocks we couldn't write due to previous RO mount */ 2203 if (!(*flags & SB_RDONLY) && is_sbi_flag_set(sbi, SBI_NEED_SB_WRITE)) { 2204 err = f2fs_commit_super(sbi, false); 2205 f2fs_info(sbi, "Try to recover all the superblocks, ret: %d", 2206 err); 2207 if (!err) 2208 clear_sbi_flag(sbi, SBI_NEED_SB_WRITE); 2209 } 2210 2211 default_options(sbi); 2212 2213 /* parse mount options */ 2214 err = parse_options(sb, data, true); 2215 if (err) 2216 goto restore_opts; 2217 2218 /* 2219 * Previous and new state of filesystem is RO, 2220 * so skip checking GC and FLUSH_MERGE conditions. 2221 */ 2222 if (f2fs_readonly(sb) && (*flags & SB_RDONLY)) 2223 goto skip; 2224 2225 if (f2fs_sb_has_readonly(sbi) && !(*flags & SB_RDONLY)) { 2226 err = -EROFS; 2227 goto restore_opts; 2228 } 2229 2230 #ifdef CONFIG_QUOTA 2231 if (!f2fs_readonly(sb) && (*flags & SB_RDONLY)) { 2232 err = dquot_suspend(sb, -1); 2233 if (err < 0) 2234 goto restore_opts; 2235 } else if (f2fs_readonly(sb) && !(*flags & SB_RDONLY)) { 2236 /* dquot_resume needs RW */ 2237 sb->s_flags &= ~SB_RDONLY; 2238 if (sb_any_quota_suspended(sb)) { 2239 dquot_resume(sb, -1); 2240 } else if (f2fs_sb_has_quota_ino(sbi)) { 2241 err = f2fs_enable_quotas(sb); 2242 if (err) 2243 goto restore_opts; 2244 } 2245 } 2246 #endif 2247 /* disallow enable atgc dynamically */ 2248 if (no_atgc == !!test_opt(sbi, ATGC)) { 2249 err = -EINVAL; 2250 f2fs_warn(sbi, "switch atgc option is not allowed"); 2251 goto restore_opts; 2252 } 2253 2254 /* disallow enable/disable extent_cache dynamically */ 2255 if (no_extent_cache == !!test_opt(sbi, EXTENT_CACHE)) { 2256 err = -EINVAL; 2257 f2fs_warn(sbi, "switch extent_cache option is not allowed"); 2258 goto restore_opts; 2259 } 2260 2261 if (no_io_align == !!F2FS_IO_ALIGNED(sbi)) { 2262 err = -EINVAL; 2263 f2fs_warn(sbi, "switch io_bits option is not allowed"); 2264 goto restore_opts; 2265 } 2266 2267 if (no_compress_cache == !!test_opt(sbi, COMPRESS_CACHE)) { 2268 err = -EINVAL; 2269 f2fs_warn(sbi, "switch compress_cache option is not allowed"); 2270 goto restore_opts; 2271 } 2272 2273 if (block_unit_discard != f2fs_block_unit_discard(sbi)) { 2274 err = -EINVAL; 2275 f2fs_warn(sbi, "switch discard_unit option is not allowed"); 2276 goto restore_opts; 2277 } 2278 2279 if ((*flags & SB_RDONLY) && test_opt(sbi, DISABLE_CHECKPOINT)) { 2280 err = -EINVAL; 2281 f2fs_warn(sbi, "disabling checkpoint not compatible with read-only"); 2282 goto restore_opts; 2283 } 2284 2285 /* 2286 * We stop the GC thread if FS is mounted as RO 2287 * or if background_gc = off is passed in mount 2288 * option. Also sync the filesystem. 2289 */ 2290 if ((*flags & SB_RDONLY) || 2291 (F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_OFF && 2292 !test_opt(sbi, GC_MERGE))) { 2293 if (sbi->gc_thread) { 2294 f2fs_stop_gc_thread(sbi); 2295 need_restart_gc = true; 2296 } 2297 } else if (!sbi->gc_thread) { 2298 err = f2fs_start_gc_thread(sbi); 2299 if (err) 2300 goto restore_opts; 2301 need_stop_gc = true; 2302 } 2303 2304 if (*flags & SB_RDONLY || 2305 F2FS_OPTION(sbi).whint_mode != org_mount_opt.whint_mode) { 2306 sync_inodes_sb(sb); 2307 2308 set_sbi_flag(sbi, SBI_IS_DIRTY); 2309 set_sbi_flag(sbi, SBI_IS_CLOSE); 2310 f2fs_sync_fs(sb, 1); 2311 clear_sbi_flag(sbi, SBI_IS_CLOSE); 2312 } 2313 2314 if ((*flags & SB_RDONLY) || test_opt(sbi, DISABLE_CHECKPOINT) || 2315 !test_opt(sbi, MERGE_CHECKPOINT)) { 2316 f2fs_stop_ckpt_thread(sbi); 2317 need_restart_ckpt = true; 2318 } else { 2319 err = f2fs_start_ckpt_thread(sbi); 2320 if (err) { 2321 f2fs_err(sbi, 2322 "Failed to start F2FS issue_checkpoint_thread (%d)", 2323 err); 2324 goto restore_gc; 2325 } 2326 need_stop_ckpt = true; 2327 } 2328 2329 /* 2330 * We stop issue flush thread if FS is mounted as RO 2331 * or if flush_merge is not passed in mount option. 2332 */ 2333 if ((*flags & SB_RDONLY) || !test_opt(sbi, FLUSH_MERGE)) { 2334 clear_opt(sbi, FLUSH_MERGE); 2335 f2fs_destroy_flush_cmd_control(sbi, false); 2336 need_restart_flush = true; 2337 } else { 2338 err = f2fs_create_flush_cmd_control(sbi); 2339 if (err) 2340 goto restore_ckpt; 2341 need_stop_flush = true; 2342 } 2343 2344 if (no_discard == !!test_opt(sbi, DISCARD)) { 2345 if (test_opt(sbi, DISCARD)) { 2346 err = f2fs_start_discard_thread(sbi); 2347 if (err) 2348 goto restore_flush; 2349 need_stop_discard = true; 2350 } else { 2351 dcc = SM_I(sbi)->dcc_info; 2352 f2fs_stop_discard_thread(sbi); 2353 if (atomic_read(&dcc->discard_cmd_cnt)) 2354 f2fs_issue_discard_timeout(sbi); 2355 need_restart_discard = true; 2356 } 2357 } 2358 2359 if (enable_checkpoint == !!test_opt(sbi, DISABLE_CHECKPOINT)) { 2360 if (test_opt(sbi, DISABLE_CHECKPOINT)) { 2361 err = f2fs_disable_checkpoint(sbi); 2362 if (err) 2363 goto restore_discard; 2364 } else { 2365 f2fs_enable_checkpoint(sbi); 2366 } 2367 } 2368 2369 skip: 2370 #ifdef CONFIG_QUOTA 2371 /* Release old quota file names */ 2372 for (i = 0; i < MAXQUOTAS; i++) 2373 kfree(org_mount_opt.s_qf_names[i]); 2374 #endif 2375 /* Update the POSIXACL Flag */ 2376 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) | 2377 (test_opt(sbi, POSIX_ACL) ? SB_POSIXACL : 0); 2378 2379 limit_reserve_root(sbi); 2380 adjust_unusable_cap_perc(sbi); 2381 *flags = (*flags & ~SB_LAZYTIME) | (sb->s_flags & SB_LAZYTIME); 2382 return 0; 2383 restore_discard: 2384 if (need_restart_discard) { 2385 if (f2fs_start_discard_thread(sbi)) 2386 f2fs_warn(sbi, "discard has been stopped"); 2387 } else if (need_stop_discard) { 2388 f2fs_stop_discard_thread(sbi); 2389 } 2390 restore_flush: 2391 if (need_restart_flush) { 2392 if (f2fs_create_flush_cmd_control(sbi)) 2393 f2fs_warn(sbi, "background flush thread has stopped"); 2394 } else if (need_stop_flush) { 2395 clear_opt(sbi, FLUSH_MERGE); 2396 f2fs_destroy_flush_cmd_control(sbi, false); 2397 } 2398 restore_ckpt: 2399 if (need_restart_ckpt) { 2400 if (f2fs_start_ckpt_thread(sbi)) 2401 f2fs_warn(sbi, "background ckpt thread has stopped"); 2402 } else if (need_stop_ckpt) { 2403 f2fs_stop_ckpt_thread(sbi); 2404 } 2405 restore_gc: 2406 if (need_restart_gc) { 2407 if (f2fs_start_gc_thread(sbi)) 2408 f2fs_warn(sbi, "background gc thread has stopped"); 2409 } else if (need_stop_gc) { 2410 f2fs_stop_gc_thread(sbi); 2411 } 2412 restore_opts: 2413 #ifdef CONFIG_QUOTA 2414 F2FS_OPTION(sbi).s_jquota_fmt = org_mount_opt.s_jquota_fmt; 2415 for (i = 0; i < MAXQUOTAS; i++) { 2416 kfree(F2FS_OPTION(sbi).s_qf_names[i]); 2417 F2FS_OPTION(sbi).s_qf_names[i] = org_mount_opt.s_qf_names[i]; 2418 } 2419 #endif 2420 sbi->mount_opt = org_mount_opt; 2421 sb->s_flags = old_sb_flags; 2422 return err; 2423 } 2424 2425 #ifdef CONFIG_QUOTA 2426 /* Read data from quotafile */ 2427 static ssize_t f2fs_quota_read(struct super_block *sb, int type, char *data, 2428 size_t len, loff_t off) 2429 { 2430 struct inode *inode = sb_dqopt(sb)->files[type]; 2431 struct address_space *mapping = inode->i_mapping; 2432 block_t blkidx = F2FS_BYTES_TO_BLK(off); 2433 int offset = off & (sb->s_blocksize - 1); 2434 int tocopy; 2435 size_t toread; 2436 loff_t i_size = i_size_read(inode); 2437 struct page *page; 2438 char *kaddr; 2439 2440 if (off > i_size) 2441 return 0; 2442 2443 if (off + len > i_size) 2444 len = i_size - off; 2445 toread = len; 2446 while (toread > 0) { 2447 tocopy = min_t(unsigned long, sb->s_blocksize - offset, toread); 2448 repeat: 2449 page = read_cache_page_gfp(mapping, blkidx, GFP_NOFS); 2450 if (IS_ERR(page)) { 2451 if (PTR_ERR(page) == -ENOMEM) { 2452 memalloc_retry_wait(GFP_NOFS); 2453 goto repeat; 2454 } 2455 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR); 2456 return PTR_ERR(page); 2457 } 2458 2459 lock_page(page); 2460 2461 if (unlikely(page->mapping != mapping)) { 2462 f2fs_put_page(page, 1); 2463 goto repeat; 2464 } 2465 if (unlikely(!PageUptodate(page))) { 2466 f2fs_put_page(page, 1); 2467 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR); 2468 return -EIO; 2469 } 2470 2471 kaddr = kmap_atomic(page); 2472 memcpy(data, kaddr + offset, tocopy); 2473 kunmap_atomic(kaddr); 2474 f2fs_put_page(page, 1); 2475 2476 offset = 0; 2477 toread -= tocopy; 2478 data += tocopy; 2479 blkidx++; 2480 } 2481 return len; 2482 } 2483 2484 /* Write to quotafile */ 2485 static ssize_t f2fs_quota_write(struct super_block *sb, int type, 2486 const char *data, size_t len, loff_t off) 2487 { 2488 struct inode *inode = sb_dqopt(sb)->files[type]; 2489 struct address_space *mapping = inode->i_mapping; 2490 const struct address_space_operations *a_ops = mapping->a_ops; 2491 int offset = off & (sb->s_blocksize - 1); 2492 size_t towrite = len; 2493 struct page *page; 2494 void *fsdata = NULL; 2495 char *kaddr; 2496 int err = 0; 2497 int tocopy; 2498 2499 while (towrite > 0) { 2500 tocopy = min_t(unsigned long, sb->s_blocksize - offset, 2501 towrite); 2502 retry: 2503 err = a_ops->write_begin(NULL, mapping, off, tocopy, 0, 2504 &page, &fsdata); 2505 if (unlikely(err)) { 2506 if (err == -ENOMEM) { 2507 congestion_wait(BLK_RW_ASYNC, 2508 DEFAULT_IO_TIMEOUT); 2509 goto retry; 2510 } 2511 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR); 2512 break; 2513 } 2514 2515 kaddr = kmap_atomic(page); 2516 memcpy(kaddr + offset, data, tocopy); 2517 kunmap_atomic(kaddr); 2518 flush_dcache_page(page); 2519 2520 a_ops->write_end(NULL, mapping, off, tocopy, tocopy, 2521 page, fsdata); 2522 offset = 0; 2523 towrite -= tocopy; 2524 off += tocopy; 2525 data += tocopy; 2526 cond_resched(); 2527 } 2528 2529 if (len == towrite) 2530 return err; 2531 inode->i_mtime = inode->i_ctime = current_time(inode); 2532 f2fs_mark_inode_dirty_sync(inode, false); 2533 return len - towrite; 2534 } 2535 2536 int f2fs_dquot_initialize(struct inode *inode) 2537 { 2538 if (time_to_inject(F2FS_I_SB(inode), FAULT_DQUOT_INIT)) { 2539 f2fs_show_injection_info(F2FS_I_SB(inode), FAULT_DQUOT_INIT); 2540 return -ESRCH; 2541 } 2542 2543 return dquot_initialize(inode); 2544 } 2545 2546 static struct dquot **f2fs_get_dquots(struct inode *inode) 2547 { 2548 return F2FS_I(inode)->i_dquot; 2549 } 2550 2551 static qsize_t *f2fs_get_reserved_space(struct inode *inode) 2552 { 2553 return &F2FS_I(inode)->i_reserved_quota; 2554 } 2555 2556 static int f2fs_quota_on_mount(struct f2fs_sb_info *sbi, int type) 2557 { 2558 if (is_set_ckpt_flags(sbi, CP_QUOTA_NEED_FSCK_FLAG)) { 2559 f2fs_err(sbi, "quota sysfile may be corrupted, skip loading it"); 2560 return 0; 2561 } 2562 2563 return dquot_quota_on_mount(sbi->sb, F2FS_OPTION(sbi).s_qf_names[type], 2564 F2FS_OPTION(sbi).s_jquota_fmt, type); 2565 } 2566 2567 int f2fs_enable_quota_files(struct f2fs_sb_info *sbi, bool rdonly) 2568 { 2569 int enabled = 0; 2570 int i, err; 2571 2572 if (f2fs_sb_has_quota_ino(sbi) && rdonly) { 2573 err = f2fs_enable_quotas(sbi->sb); 2574 if (err) { 2575 f2fs_err(sbi, "Cannot turn on quota_ino: %d", err); 2576 return 0; 2577 } 2578 return 1; 2579 } 2580 2581 for (i = 0; i < MAXQUOTAS; i++) { 2582 if (F2FS_OPTION(sbi).s_qf_names[i]) { 2583 err = f2fs_quota_on_mount(sbi, i); 2584 if (!err) { 2585 enabled = 1; 2586 continue; 2587 } 2588 f2fs_err(sbi, "Cannot turn on quotas: %d on %d", 2589 err, i); 2590 } 2591 } 2592 return enabled; 2593 } 2594 2595 static int f2fs_quota_enable(struct super_block *sb, int type, int format_id, 2596 unsigned int flags) 2597 { 2598 struct inode *qf_inode; 2599 unsigned long qf_inum; 2600 int err; 2601 2602 BUG_ON(!f2fs_sb_has_quota_ino(F2FS_SB(sb))); 2603 2604 qf_inum = f2fs_qf_ino(sb, type); 2605 if (!qf_inum) 2606 return -EPERM; 2607 2608 qf_inode = f2fs_iget(sb, qf_inum); 2609 if (IS_ERR(qf_inode)) { 2610 f2fs_err(F2FS_SB(sb), "Bad quota inode %u:%lu", type, qf_inum); 2611 return PTR_ERR(qf_inode); 2612 } 2613 2614 /* Don't account quota for quota files to avoid recursion */ 2615 qf_inode->i_flags |= S_NOQUOTA; 2616 err = dquot_load_quota_inode(qf_inode, type, format_id, flags); 2617 iput(qf_inode); 2618 return err; 2619 } 2620 2621 static int f2fs_enable_quotas(struct super_block *sb) 2622 { 2623 struct f2fs_sb_info *sbi = F2FS_SB(sb); 2624 int type, err = 0; 2625 unsigned long qf_inum; 2626 bool quota_mopt[MAXQUOTAS] = { 2627 test_opt(sbi, USRQUOTA), 2628 test_opt(sbi, GRPQUOTA), 2629 test_opt(sbi, PRJQUOTA), 2630 }; 2631 2632 if (is_set_ckpt_flags(F2FS_SB(sb), CP_QUOTA_NEED_FSCK_FLAG)) { 2633 f2fs_err(sbi, "quota file may be corrupted, skip loading it"); 2634 return 0; 2635 } 2636 2637 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE; 2638 2639 for (type = 0; type < MAXQUOTAS; type++) { 2640 qf_inum = f2fs_qf_ino(sb, type); 2641 if (qf_inum) { 2642 err = f2fs_quota_enable(sb, type, QFMT_VFS_V1, 2643 DQUOT_USAGE_ENABLED | 2644 (quota_mopt[type] ? DQUOT_LIMITS_ENABLED : 0)); 2645 if (err) { 2646 f2fs_err(sbi, "Failed to enable quota tracking (type=%d, err=%d). Please run fsck to fix.", 2647 type, err); 2648 for (type--; type >= 0; type--) 2649 dquot_quota_off(sb, type); 2650 set_sbi_flag(F2FS_SB(sb), 2651 SBI_QUOTA_NEED_REPAIR); 2652 return err; 2653 } 2654 } 2655 } 2656 return 0; 2657 } 2658 2659 static int f2fs_quota_sync_file(struct f2fs_sb_info *sbi, int type) 2660 { 2661 struct quota_info *dqopt = sb_dqopt(sbi->sb); 2662 struct address_space *mapping = dqopt->files[type]->i_mapping; 2663 int ret = 0; 2664 2665 ret = dquot_writeback_dquots(sbi->sb, type); 2666 if (ret) 2667 goto out; 2668 2669 ret = filemap_fdatawrite(mapping); 2670 if (ret) 2671 goto out; 2672 2673 /* if we are using journalled quota */ 2674 if (is_journalled_quota(sbi)) 2675 goto out; 2676 2677 ret = filemap_fdatawait(mapping); 2678 2679 truncate_inode_pages(&dqopt->files[type]->i_data, 0); 2680 out: 2681 if (ret) 2682 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR); 2683 return ret; 2684 } 2685 2686 int f2fs_quota_sync(struct super_block *sb, int type) 2687 { 2688 struct f2fs_sb_info *sbi = F2FS_SB(sb); 2689 struct quota_info *dqopt = sb_dqopt(sb); 2690 int cnt; 2691 int ret; 2692 2693 /* 2694 * Now when everything is written we can discard the pagecache so 2695 * that userspace sees the changes. 2696 */ 2697 for (cnt = 0; cnt < MAXQUOTAS; cnt++) { 2698 2699 if (type != -1 && cnt != type) 2700 continue; 2701 2702 if (!sb_has_quota_active(sb, type)) 2703 return 0; 2704 2705 inode_lock(dqopt->files[cnt]); 2706 2707 /* 2708 * do_quotactl 2709 * f2fs_quota_sync 2710 * down_read(quota_sem) 2711 * dquot_writeback_dquots() 2712 * f2fs_dquot_commit 2713 * block_operation 2714 * down_read(quota_sem) 2715 */ 2716 f2fs_lock_op(sbi); 2717 down_read(&sbi->quota_sem); 2718 2719 ret = f2fs_quota_sync_file(sbi, cnt); 2720 2721 up_read(&sbi->quota_sem); 2722 f2fs_unlock_op(sbi); 2723 2724 inode_unlock(dqopt->files[cnt]); 2725 2726 if (ret) 2727 break; 2728 } 2729 return ret; 2730 } 2731 2732 static int f2fs_quota_on(struct super_block *sb, int type, int format_id, 2733 const struct path *path) 2734 { 2735 struct inode *inode; 2736 int err; 2737 2738 /* if quota sysfile exists, deny enabling quota with specific file */ 2739 if (f2fs_sb_has_quota_ino(F2FS_SB(sb))) { 2740 f2fs_err(F2FS_SB(sb), "quota sysfile already exists"); 2741 return -EBUSY; 2742 } 2743 2744 err = f2fs_quota_sync(sb, type); 2745 if (err) 2746 return err; 2747 2748 err = dquot_quota_on(sb, type, format_id, path); 2749 if (err) 2750 return err; 2751 2752 inode = d_inode(path->dentry); 2753 2754 inode_lock(inode); 2755 F2FS_I(inode)->i_flags |= F2FS_NOATIME_FL | F2FS_IMMUTABLE_FL; 2756 f2fs_set_inode_flags(inode); 2757 inode_unlock(inode); 2758 f2fs_mark_inode_dirty_sync(inode, false); 2759 2760 return 0; 2761 } 2762 2763 static int __f2fs_quota_off(struct super_block *sb, int type) 2764 { 2765 struct inode *inode = sb_dqopt(sb)->files[type]; 2766 int err; 2767 2768 if (!inode || !igrab(inode)) 2769 return dquot_quota_off(sb, type); 2770 2771 err = f2fs_quota_sync(sb, type); 2772 if (err) 2773 goto out_put; 2774 2775 err = dquot_quota_off(sb, type); 2776 if (err || f2fs_sb_has_quota_ino(F2FS_SB(sb))) 2777 goto out_put; 2778 2779 inode_lock(inode); 2780 F2FS_I(inode)->i_flags &= ~(F2FS_NOATIME_FL | F2FS_IMMUTABLE_FL); 2781 f2fs_set_inode_flags(inode); 2782 inode_unlock(inode); 2783 f2fs_mark_inode_dirty_sync(inode, false); 2784 out_put: 2785 iput(inode); 2786 return err; 2787 } 2788 2789 static int f2fs_quota_off(struct super_block *sb, int type) 2790 { 2791 struct f2fs_sb_info *sbi = F2FS_SB(sb); 2792 int err; 2793 2794 err = __f2fs_quota_off(sb, type); 2795 2796 /* 2797 * quotactl can shutdown journalled quota, result in inconsistence 2798 * between quota record and fs data by following updates, tag the 2799 * flag to let fsck be aware of it. 2800 */ 2801 if (is_journalled_quota(sbi)) 2802 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR); 2803 return err; 2804 } 2805 2806 void f2fs_quota_off_umount(struct super_block *sb) 2807 { 2808 int type; 2809 int err; 2810 2811 for (type = 0; type < MAXQUOTAS; type++) { 2812 err = __f2fs_quota_off(sb, type); 2813 if (err) { 2814 int ret = dquot_quota_off(sb, type); 2815 2816 f2fs_err(F2FS_SB(sb), "Fail to turn off disk quota (type: %d, err: %d, ret:%d), Please run fsck to fix it.", 2817 type, err, ret); 2818 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR); 2819 } 2820 } 2821 /* 2822 * In case of checkpoint=disable, we must flush quota blocks. 2823 * This can cause NULL exception for node_inode in end_io, since 2824 * put_super already dropped it. 2825 */ 2826 sync_filesystem(sb); 2827 } 2828 2829 static void f2fs_truncate_quota_inode_pages(struct super_block *sb) 2830 { 2831 struct quota_info *dqopt = sb_dqopt(sb); 2832 int type; 2833 2834 for (type = 0; type < MAXQUOTAS; type++) { 2835 if (!dqopt->files[type]) 2836 continue; 2837 f2fs_inode_synced(dqopt->files[type]); 2838 } 2839 } 2840 2841 static int f2fs_dquot_commit(struct dquot *dquot) 2842 { 2843 struct f2fs_sb_info *sbi = F2FS_SB(dquot->dq_sb); 2844 int ret; 2845 2846 down_read_nested(&sbi->quota_sem, SINGLE_DEPTH_NESTING); 2847 ret = dquot_commit(dquot); 2848 if (ret < 0) 2849 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR); 2850 up_read(&sbi->quota_sem); 2851 return ret; 2852 } 2853 2854 static int f2fs_dquot_acquire(struct dquot *dquot) 2855 { 2856 struct f2fs_sb_info *sbi = F2FS_SB(dquot->dq_sb); 2857 int ret; 2858 2859 down_read(&sbi->quota_sem); 2860 ret = dquot_acquire(dquot); 2861 if (ret < 0) 2862 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR); 2863 up_read(&sbi->quota_sem); 2864 return ret; 2865 } 2866 2867 static int f2fs_dquot_release(struct dquot *dquot) 2868 { 2869 struct f2fs_sb_info *sbi = F2FS_SB(dquot->dq_sb); 2870 int ret = dquot_release(dquot); 2871 2872 if (ret < 0) 2873 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR); 2874 return ret; 2875 } 2876 2877 static int f2fs_dquot_mark_dquot_dirty(struct dquot *dquot) 2878 { 2879 struct super_block *sb = dquot->dq_sb; 2880 struct f2fs_sb_info *sbi = F2FS_SB(sb); 2881 int ret = dquot_mark_dquot_dirty(dquot); 2882 2883 /* if we are using journalled quota */ 2884 if (is_journalled_quota(sbi)) 2885 set_sbi_flag(sbi, SBI_QUOTA_NEED_FLUSH); 2886 2887 return ret; 2888 } 2889 2890 static int f2fs_dquot_commit_info(struct super_block *sb, int type) 2891 { 2892 struct f2fs_sb_info *sbi = F2FS_SB(sb); 2893 int ret = dquot_commit_info(sb, type); 2894 2895 if (ret < 0) 2896 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR); 2897 return ret; 2898 } 2899 2900 static int f2fs_get_projid(struct inode *inode, kprojid_t *projid) 2901 { 2902 *projid = F2FS_I(inode)->i_projid; 2903 return 0; 2904 } 2905 2906 static const struct dquot_operations f2fs_quota_operations = { 2907 .get_reserved_space = f2fs_get_reserved_space, 2908 .write_dquot = f2fs_dquot_commit, 2909 .acquire_dquot = f2fs_dquot_acquire, 2910 .release_dquot = f2fs_dquot_release, 2911 .mark_dirty = f2fs_dquot_mark_dquot_dirty, 2912 .write_info = f2fs_dquot_commit_info, 2913 .alloc_dquot = dquot_alloc, 2914 .destroy_dquot = dquot_destroy, 2915 .get_projid = f2fs_get_projid, 2916 .get_next_id = dquot_get_next_id, 2917 }; 2918 2919 static const struct quotactl_ops f2fs_quotactl_ops = { 2920 .quota_on = f2fs_quota_on, 2921 .quota_off = f2fs_quota_off, 2922 .quota_sync = f2fs_quota_sync, 2923 .get_state = dquot_get_state, 2924 .set_info = dquot_set_dqinfo, 2925 .get_dqblk = dquot_get_dqblk, 2926 .set_dqblk = dquot_set_dqblk, 2927 .get_nextdqblk = dquot_get_next_dqblk, 2928 }; 2929 #else 2930 int f2fs_dquot_initialize(struct inode *inode) 2931 { 2932 return 0; 2933 } 2934 2935 int f2fs_quota_sync(struct super_block *sb, int type) 2936 { 2937 return 0; 2938 } 2939 2940 void f2fs_quota_off_umount(struct super_block *sb) 2941 { 2942 } 2943 #endif 2944 2945 static const struct super_operations f2fs_sops = { 2946 .alloc_inode = f2fs_alloc_inode, 2947 .free_inode = f2fs_free_inode, 2948 .drop_inode = f2fs_drop_inode, 2949 .write_inode = f2fs_write_inode, 2950 .dirty_inode = f2fs_dirty_inode, 2951 .show_options = f2fs_show_options, 2952 #ifdef CONFIG_QUOTA 2953 .quota_read = f2fs_quota_read, 2954 .quota_write = f2fs_quota_write, 2955 .get_dquots = f2fs_get_dquots, 2956 #endif 2957 .evict_inode = f2fs_evict_inode, 2958 .put_super = f2fs_put_super, 2959 .sync_fs = f2fs_sync_fs, 2960 .freeze_fs = f2fs_freeze, 2961 .unfreeze_fs = f2fs_unfreeze, 2962 .statfs = f2fs_statfs, 2963 .remount_fs = f2fs_remount, 2964 }; 2965 2966 #ifdef CONFIG_FS_ENCRYPTION 2967 static int f2fs_get_context(struct inode *inode, void *ctx, size_t len) 2968 { 2969 return f2fs_getxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION, 2970 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT, 2971 ctx, len, NULL); 2972 } 2973 2974 static int f2fs_set_context(struct inode *inode, const void *ctx, size_t len, 2975 void *fs_data) 2976 { 2977 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 2978 2979 /* 2980 * Encrypting the root directory is not allowed because fsck 2981 * expects lost+found directory to exist and remain unencrypted 2982 * if LOST_FOUND feature is enabled. 2983 * 2984 */ 2985 if (f2fs_sb_has_lost_found(sbi) && 2986 inode->i_ino == F2FS_ROOT_INO(sbi)) 2987 return -EPERM; 2988 2989 return f2fs_setxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION, 2990 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT, 2991 ctx, len, fs_data, XATTR_CREATE); 2992 } 2993 2994 static const union fscrypt_policy *f2fs_get_dummy_policy(struct super_block *sb) 2995 { 2996 return F2FS_OPTION(F2FS_SB(sb)).dummy_enc_policy.policy; 2997 } 2998 2999 static bool f2fs_has_stable_inodes(struct super_block *sb) 3000 { 3001 return true; 3002 } 3003 3004 static void f2fs_get_ino_and_lblk_bits(struct super_block *sb, 3005 int *ino_bits_ret, int *lblk_bits_ret) 3006 { 3007 *ino_bits_ret = 8 * sizeof(nid_t); 3008 *lblk_bits_ret = 8 * sizeof(block_t); 3009 } 3010 3011 static int f2fs_get_num_devices(struct super_block *sb) 3012 { 3013 struct f2fs_sb_info *sbi = F2FS_SB(sb); 3014 3015 if (f2fs_is_multi_device(sbi)) 3016 return sbi->s_ndevs; 3017 return 1; 3018 } 3019 3020 static void f2fs_get_devices(struct super_block *sb, 3021 struct request_queue **devs) 3022 { 3023 struct f2fs_sb_info *sbi = F2FS_SB(sb); 3024 int i; 3025 3026 for (i = 0; i < sbi->s_ndevs; i++) 3027 devs[i] = bdev_get_queue(FDEV(i).bdev); 3028 } 3029 3030 static const struct fscrypt_operations f2fs_cryptops = { 3031 .key_prefix = "f2fs:", 3032 .get_context = f2fs_get_context, 3033 .set_context = f2fs_set_context, 3034 .get_dummy_policy = f2fs_get_dummy_policy, 3035 .empty_dir = f2fs_empty_dir, 3036 .has_stable_inodes = f2fs_has_stable_inodes, 3037 .get_ino_and_lblk_bits = f2fs_get_ino_and_lblk_bits, 3038 .get_num_devices = f2fs_get_num_devices, 3039 .get_devices = f2fs_get_devices, 3040 }; 3041 #endif 3042 3043 static struct inode *f2fs_nfs_get_inode(struct super_block *sb, 3044 u64 ino, u32 generation) 3045 { 3046 struct f2fs_sb_info *sbi = F2FS_SB(sb); 3047 struct inode *inode; 3048 3049 if (f2fs_check_nid_range(sbi, ino)) 3050 return ERR_PTR(-ESTALE); 3051 3052 /* 3053 * f2fs_iget isn't quite right if the inode is currently unallocated! 3054 * However f2fs_iget currently does appropriate checks to handle stale 3055 * inodes so everything is OK. 3056 */ 3057 inode = f2fs_iget(sb, ino); 3058 if (IS_ERR(inode)) 3059 return ERR_CAST(inode); 3060 if (unlikely(generation && inode->i_generation != generation)) { 3061 /* we didn't find the right inode.. */ 3062 iput(inode); 3063 return ERR_PTR(-ESTALE); 3064 } 3065 return inode; 3066 } 3067 3068 static struct dentry *f2fs_fh_to_dentry(struct super_block *sb, struct fid *fid, 3069 int fh_len, int fh_type) 3070 { 3071 return generic_fh_to_dentry(sb, fid, fh_len, fh_type, 3072 f2fs_nfs_get_inode); 3073 } 3074 3075 static struct dentry *f2fs_fh_to_parent(struct super_block *sb, struct fid *fid, 3076 int fh_len, int fh_type) 3077 { 3078 return generic_fh_to_parent(sb, fid, fh_len, fh_type, 3079 f2fs_nfs_get_inode); 3080 } 3081 3082 static const struct export_operations f2fs_export_ops = { 3083 .fh_to_dentry = f2fs_fh_to_dentry, 3084 .fh_to_parent = f2fs_fh_to_parent, 3085 .get_parent = f2fs_get_parent, 3086 }; 3087 3088 loff_t max_file_blocks(struct inode *inode) 3089 { 3090 loff_t result = 0; 3091 loff_t leaf_count; 3092 3093 /* 3094 * note: previously, result is equal to (DEF_ADDRS_PER_INODE - 3095 * DEFAULT_INLINE_XATTR_ADDRS), but now f2fs try to reserve more 3096 * space in inode.i_addr, it will be more safe to reassign 3097 * result as zero. 3098 */ 3099 3100 if (inode && f2fs_compressed_file(inode)) 3101 leaf_count = ADDRS_PER_BLOCK(inode); 3102 else 3103 leaf_count = DEF_ADDRS_PER_BLOCK; 3104 3105 /* two direct node blocks */ 3106 result += (leaf_count * 2); 3107 3108 /* two indirect node blocks */ 3109 leaf_count *= NIDS_PER_BLOCK; 3110 result += (leaf_count * 2); 3111 3112 /* one double indirect node block */ 3113 leaf_count *= NIDS_PER_BLOCK; 3114 result += leaf_count; 3115 3116 return result; 3117 } 3118 3119 static int __f2fs_commit_super(struct buffer_head *bh, 3120 struct f2fs_super_block *super) 3121 { 3122 lock_buffer(bh); 3123 if (super) 3124 memcpy(bh->b_data + F2FS_SUPER_OFFSET, super, sizeof(*super)); 3125 set_buffer_dirty(bh); 3126 unlock_buffer(bh); 3127 3128 /* it's rare case, we can do fua all the time */ 3129 return __sync_dirty_buffer(bh, REQ_SYNC | REQ_PREFLUSH | REQ_FUA); 3130 } 3131 3132 static inline bool sanity_check_area_boundary(struct f2fs_sb_info *sbi, 3133 struct buffer_head *bh) 3134 { 3135 struct f2fs_super_block *raw_super = (struct f2fs_super_block *) 3136 (bh->b_data + F2FS_SUPER_OFFSET); 3137 struct super_block *sb = sbi->sb; 3138 u32 segment0_blkaddr = le32_to_cpu(raw_super->segment0_blkaddr); 3139 u32 cp_blkaddr = le32_to_cpu(raw_super->cp_blkaddr); 3140 u32 sit_blkaddr = le32_to_cpu(raw_super->sit_blkaddr); 3141 u32 nat_blkaddr = le32_to_cpu(raw_super->nat_blkaddr); 3142 u32 ssa_blkaddr = le32_to_cpu(raw_super->ssa_blkaddr); 3143 u32 main_blkaddr = le32_to_cpu(raw_super->main_blkaddr); 3144 u32 segment_count_ckpt = le32_to_cpu(raw_super->segment_count_ckpt); 3145 u32 segment_count_sit = le32_to_cpu(raw_super->segment_count_sit); 3146 u32 segment_count_nat = le32_to_cpu(raw_super->segment_count_nat); 3147 u32 segment_count_ssa = le32_to_cpu(raw_super->segment_count_ssa); 3148 u32 segment_count_main = le32_to_cpu(raw_super->segment_count_main); 3149 u32 segment_count = le32_to_cpu(raw_super->segment_count); 3150 u32 log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg); 3151 u64 main_end_blkaddr = main_blkaddr + 3152 (segment_count_main << log_blocks_per_seg); 3153 u64 seg_end_blkaddr = segment0_blkaddr + 3154 (segment_count << log_blocks_per_seg); 3155 3156 if (segment0_blkaddr != cp_blkaddr) { 3157 f2fs_info(sbi, "Mismatch start address, segment0(%u) cp_blkaddr(%u)", 3158 segment0_blkaddr, cp_blkaddr); 3159 return true; 3160 } 3161 3162 if (cp_blkaddr + (segment_count_ckpt << log_blocks_per_seg) != 3163 sit_blkaddr) { 3164 f2fs_info(sbi, "Wrong CP boundary, start(%u) end(%u) blocks(%u)", 3165 cp_blkaddr, sit_blkaddr, 3166 segment_count_ckpt << log_blocks_per_seg); 3167 return true; 3168 } 3169 3170 if (sit_blkaddr + (segment_count_sit << log_blocks_per_seg) != 3171 nat_blkaddr) { 3172 f2fs_info(sbi, "Wrong SIT boundary, start(%u) end(%u) blocks(%u)", 3173 sit_blkaddr, nat_blkaddr, 3174 segment_count_sit << log_blocks_per_seg); 3175 return true; 3176 } 3177 3178 if (nat_blkaddr + (segment_count_nat << log_blocks_per_seg) != 3179 ssa_blkaddr) { 3180 f2fs_info(sbi, "Wrong NAT boundary, start(%u) end(%u) blocks(%u)", 3181 nat_blkaddr, ssa_blkaddr, 3182 segment_count_nat << log_blocks_per_seg); 3183 return true; 3184 } 3185 3186 if (ssa_blkaddr + (segment_count_ssa << log_blocks_per_seg) != 3187 main_blkaddr) { 3188 f2fs_info(sbi, "Wrong SSA boundary, start(%u) end(%u) blocks(%u)", 3189 ssa_blkaddr, main_blkaddr, 3190 segment_count_ssa << log_blocks_per_seg); 3191 return true; 3192 } 3193 3194 if (main_end_blkaddr > seg_end_blkaddr) { 3195 f2fs_info(sbi, "Wrong MAIN_AREA boundary, start(%u) end(%llu) block(%u)", 3196 main_blkaddr, seg_end_blkaddr, 3197 segment_count_main << log_blocks_per_seg); 3198 return true; 3199 } else if (main_end_blkaddr < seg_end_blkaddr) { 3200 int err = 0; 3201 char *res; 3202 3203 /* fix in-memory information all the time */ 3204 raw_super->segment_count = cpu_to_le32((main_end_blkaddr - 3205 segment0_blkaddr) >> log_blocks_per_seg); 3206 3207 if (f2fs_readonly(sb) || bdev_read_only(sb->s_bdev)) { 3208 set_sbi_flag(sbi, SBI_NEED_SB_WRITE); 3209 res = "internally"; 3210 } else { 3211 err = __f2fs_commit_super(bh, NULL); 3212 res = err ? "failed" : "done"; 3213 } 3214 f2fs_info(sbi, "Fix alignment : %s, start(%u) end(%llu) block(%u)", 3215 res, main_blkaddr, seg_end_blkaddr, 3216 segment_count_main << log_blocks_per_seg); 3217 if (err) 3218 return true; 3219 } 3220 return false; 3221 } 3222 3223 static int sanity_check_raw_super(struct f2fs_sb_info *sbi, 3224 struct buffer_head *bh) 3225 { 3226 block_t segment_count, segs_per_sec, secs_per_zone, segment_count_main; 3227 block_t total_sections, blocks_per_seg; 3228 struct f2fs_super_block *raw_super = (struct f2fs_super_block *) 3229 (bh->b_data + F2FS_SUPER_OFFSET); 3230 size_t crc_offset = 0; 3231 __u32 crc = 0; 3232 3233 if (le32_to_cpu(raw_super->magic) != F2FS_SUPER_MAGIC) { 3234 f2fs_info(sbi, "Magic Mismatch, valid(0x%x) - read(0x%x)", 3235 F2FS_SUPER_MAGIC, le32_to_cpu(raw_super->magic)); 3236 return -EINVAL; 3237 } 3238 3239 /* Check checksum_offset and crc in superblock */ 3240 if (__F2FS_HAS_FEATURE(raw_super, F2FS_FEATURE_SB_CHKSUM)) { 3241 crc_offset = le32_to_cpu(raw_super->checksum_offset); 3242 if (crc_offset != 3243 offsetof(struct f2fs_super_block, crc)) { 3244 f2fs_info(sbi, "Invalid SB checksum offset: %zu", 3245 crc_offset); 3246 return -EFSCORRUPTED; 3247 } 3248 crc = le32_to_cpu(raw_super->crc); 3249 if (!f2fs_crc_valid(sbi, crc, raw_super, crc_offset)) { 3250 f2fs_info(sbi, "Invalid SB checksum value: %u", crc); 3251 return -EFSCORRUPTED; 3252 } 3253 } 3254 3255 /* Currently, support only 4KB block size */ 3256 if (le32_to_cpu(raw_super->log_blocksize) != F2FS_BLKSIZE_BITS) { 3257 f2fs_info(sbi, "Invalid log_blocksize (%u), supports only %u", 3258 le32_to_cpu(raw_super->log_blocksize), 3259 F2FS_BLKSIZE_BITS); 3260 return -EFSCORRUPTED; 3261 } 3262 3263 /* check log blocks per segment */ 3264 if (le32_to_cpu(raw_super->log_blocks_per_seg) != 9) { 3265 f2fs_info(sbi, "Invalid log blocks per segment (%u)", 3266 le32_to_cpu(raw_super->log_blocks_per_seg)); 3267 return -EFSCORRUPTED; 3268 } 3269 3270 /* Currently, support 512/1024/2048/4096 bytes sector size */ 3271 if (le32_to_cpu(raw_super->log_sectorsize) > 3272 F2FS_MAX_LOG_SECTOR_SIZE || 3273 le32_to_cpu(raw_super->log_sectorsize) < 3274 F2FS_MIN_LOG_SECTOR_SIZE) { 3275 f2fs_info(sbi, "Invalid log sectorsize (%u)", 3276 le32_to_cpu(raw_super->log_sectorsize)); 3277 return -EFSCORRUPTED; 3278 } 3279 if (le32_to_cpu(raw_super->log_sectors_per_block) + 3280 le32_to_cpu(raw_super->log_sectorsize) != 3281 F2FS_MAX_LOG_SECTOR_SIZE) { 3282 f2fs_info(sbi, "Invalid log sectors per block(%u) log sectorsize(%u)", 3283 le32_to_cpu(raw_super->log_sectors_per_block), 3284 le32_to_cpu(raw_super->log_sectorsize)); 3285 return -EFSCORRUPTED; 3286 } 3287 3288 segment_count = le32_to_cpu(raw_super->segment_count); 3289 segment_count_main = le32_to_cpu(raw_super->segment_count_main); 3290 segs_per_sec = le32_to_cpu(raw_super->segs_per_sec); 3291 secs_per_zone = le32_to_cpu(raw_super->secs_per_zone); 3292 total_sections = le32_to_cpu(raw_super->section_count); 3293 3294 /* blocks_per_seg should be 512, given the above check */ 3295 blocks_per_seg = 1 << le32_to_cpu(raw_super->log_blocks_per_seg); 3296 3297 if (segment_count > F2FS_MAX_SEGMENT || 3298 segment_count < F2FS_MIN_SEGMENTS) { 3299 f2fs_info(sbi, "Invalid segment count (%u)", segment_count); 3300 return -EFSCORRUPTED; 3301 } 3302 3303 if (total_sections > segment_count_main || total_sections < 1 || 3304 segs_per_sec > segment_count || !segs_per_sec) { 3305 f2fs_info(sbi, "Invalid segment/section count (%u, %u x %u)", 3306 segment_count, total_sections, segs_per_sec); 3307 return -EFSCORRUPTED; 3308 } 3309 3310 if (segment_count_main != total_sections * segs_per_sec) { 3311 f2fs_info(sbi, "Invalid segment/section count (%u != %u * %u)", 3312 segment_count_main, total_sections, segs_per_sec); 3313 return -EFSCORRUPTED; 3314 } 3315 3316 if ((segment_count / segs_per_sec) < total_sections) { 3317 f2fs_info(sbi, "Small segment_count (%u < %u * %u)", 3318 segment_count, segs_per_sec, total_sections); 3319 return -EFSCORRUPTED; 3320 } 3321 3322 if (segment_count > (le64_to_cpu(raw_super->block_count) >> 9)) { 3323 f2fs_info(sbi, "Wrong segment_count / block_count (%u > %llu)", 3324 segment_count, le64_to_cpu(raw_super->block_count)); 3325 return -EFSCORRUPTED; 3326 } 3327 3328 if (RDEV(0).path[0]) { 3329 block_t dev_seg_count = le32_to_cpu(RDEV(0).total_segments); 3330 int i = 1; 3331 3332 while (i < MAX_DEVICES && RDEV(i).path[0]) { 3333 dev_seg_count += le32_to_cpu(RDEV(i).total_segments); 3334 i++; 3335 } 3336 if (segment_count != dev_seg_count) { 3337 f2fs_info(sbi, "Segment count (%u) mismatch with total segments from devices (%u)", 3338 segment_count, dev_seg_count); 3339 return -EFSCORRUPTED; 3340 } 3341 } else { 3342 if (__F2FS_HAS_FEATURE(raw_super, F2FS_FEATURE_BLKZONED) && 3343 !bdev_is_zoned(sbi->sb->s_bdev)) { 3344 f2fs_info(sbi, "Zoned block device path is missing"); 3345 return -EFSCORRUPTED; 3346 } 3347 } 3348 3349 if (secs_per_zone > total_sections || !secs_per_zone) { 3350 f2fs_info(sbi, "Wrong secs_per_zone / total_sections (%u, %u)", 3351 secs_per_zone, total_sections); 3352 return -EFSCORRUPTED; 3353 } 3354 if (le32_to_cpu(raw_super->extension_count) > F2FS_MAX_EXTENSION || 3355 raw_super->hot_ext_count > F2FS_MAX_EXTENSION || 3356 (le32_to_cpu(raw_super->extension_count) + 3357 raw_super->hot_ext_count) > F2FS_MAX_EXTENSION) { 3358 f2fs_info(sbi, "Corrupted extension count (%u + %u > %u)", 3359 le32_to_cpu(raw_super->extension_count), 3360 raw_super->hot_ext_count, 3361 F2FS_MAX_EXTENSION); 3362 return -EFSCORRUPTED; 3363 } 3364 3365 if (le32_to_cpu(raw_super->cp_payload) >= 3366 (blocks_per_seg - F2FS_CP_PACKS - 3367 NR_CURSEG_PERSIST_TYPE)) { 3368 f2fs_info(sbi, "Insane cp_payload (%u >= %u)", 3369 le32_to_cpu(raw_super->cp_payload), 3370 blocks_per_seg - F2FS_CP_PACKS - 3371 NR_CURSEG_PERSIST_TYPE); 3372 return -EFSCORRUPTED; 3373 } 3374 3375 /* check reserved ino info */ 3376 if (le32_to_cpu(raw_super->node_ino) != 1 || 3377 le32_to_cpu(raw_super->meta_ino) != 2 || 3378 le32_to_cpu(raw_super->root_ino) != 3) { 3379 f2fs_info(sbi, "Invalid Fs Meta Ino: node(%u) meta(%u) root(%u)", 3380 le32_to_cpu(raw_super->node_ino), 3381 le32_to_cpu(raw_super->meta_ino), 3382 le32_to_cpu(raw_super->root_ino)); 3383 return -EFSCORRUPTED; 3384 } 3385 3386 /* check CP/SIT/NAT/SSA/MAIN_AREA area boundary */ 3387 if (sanity_check_area_boundary(sbi, bh)) 3388 return -EFSCORRUPTED; 3389 3390 return 0; 3391 } 3392 3393 int f2fs_sanity_check_ckpt(struct f2fs_sb_info *sbi) 3394 { 3395 unsigned int total, fsmeta; 3396 struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi); 3397 struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi); 3398 unsigned int ovp_segments, reserved_segments; 3399 unsigned int main_segs, blocks_per_seg; 3400 unsigned int sit_segs, nat_segs; 3401 unsigned int sit_bitmap_size, nat_bitmap_size; 3402 unsigned int log_blocks_per_seg; 3403 unsigned int segment_count_main; 3404 unsigned int cp_pack_start_sum, cp_payload; 3405 block_t user_block_count, valid_user_blocks; 3406 block_t avail_node_count, valid_node_count; 3407 unsigned int nat_blocks, nat_bits_bytes, nat_bits_blocks; 3408 int i, j; 3409 3410 total = le32_to_cpu(raw_super->segment_count); 3411 fsmeta = le32_to_cpu(raw_super->segment_count_ckpt); 3412 sit_segs = le32_to_cpu(raw_super->segment_count_sit); 3413 fsmeta += sit_segs; 3414 nat_segs = le32_to_cpu(raw_super->segment_count_nat); 3415 fsmeta += nat_segs; 3416 fsmeta += le32_to_cpu(ckpt->rsvd_segment_count); 3417 fsmeta += le32_to_cpu(raw_super->segment_count_ssa); 3418 3419 if (unlikely(fsmeta >= total)) 3420 return 1; 3421 3422 ovp_segments = le32_to_cpu(ckpt->overprov_segment_count); 3423 reserved_segments = le32_to_cpu(ckpt->rsvd_segment_count); 3424 3425 if (!f2fs_sb_has_readonly(sbi) && 3426 unlikely(fsmeta < F2FS_MIN_META_SEGMENTS || 3427 ovp_segments == 0 || reserved_segments == 0)) { 3428 f2fs_err(sbi, "Wrong layout: check mkfs.f2fs version"); 3429 return 1; 3430 } 3431 user_block_count = le64_to_cpu(ckpt->user_block_count); 3432 segment_count_main = le32_to_cpu(raw_super->segment_count_main) + 3433 (f2fs_sb_has_readonly(sbi) ? 1 : 0); 3434 log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg); 3435 if (!user_block_count || user_block_count >= 3436 segment_count_main << log_blocks_per_seg) { 3437 f2fs_err(sbi, "Wrong user_block_count: %u", 3438 user_block_count); 3439 return 1; 3440 } 3441 3442 valid_user_blocks = le64_to_cpu(ckpt->valid_block_count); 3443 if (valid_user_blocks > user_block_count) { 3444 f2fs_err(sbi, "Wrong valid_user_blocks: %u, user_block_count: %u", 3445 valid_user_blocks, user_block_count); 3446 return 1; 3447 } 3448 3449 valid_node_count = le32_to_cpu(ckpt->valid_node_count); 3450 avail_node_count = sbi->total_node_count - F2FS_RESERVED_NODE_NUM; 3451 if (valid_node_count > avail_node_count) { 3452 f2fs_err(sbi, "Wrong valid_node_count: %u, avail_node_count: %u", 3453 valid_node_count, avail_node_count); 3454 return 1; 3455 } 3456 3457 main_segs = le32_to_cpu(raw_super->segment_count_main); 3458 blocks_per_seg = sbi->blocks_per_seg; 3459 3460 for (i = 0; i < NR_CURSEG_NODE_TYPE; i++) { 3461 if (le32_to_cpu(ckpt->cur_node_segno[i]) >= main_segs || 3462 le16_to_cpu(ckpt->cur_node_blkoff[i]) >= blocks_per_seg) 3463 return 1; 3464 3465 if (f2fs_sb_has_readonly(sbi)) 3466 goto check_data; 3467 3468 for (j = i + 1; j < NR_CURSEG_NODE_TYPE; j++) { 3469 if (le32_to_cpu(ckpt->cur_node_segno[i]) == 3470 le32_to_cpu(ckpt->cur_node_segno[j])) { 3471 f2fs_err(sbi, "Node segment (%u, %u) has the same segno: %u", 3472 i, j, 3473 le32_to_cpu(ckpt->cur_node_segno[i])); 3474 return 1; 3475 } 3476 } 3477 } 3478 check_data: 3479 for (i = 0; i < NR_CURSEG_DATA_TYPE; i++) { 3480 if (le32_to_cpu(ckpt->cur_data_segno[i]) >= main_segs || 3481 le16_to_cpu(ckpt->cur_data_blkoff[i]) >= blocks_per_seg) 3482 return 1; 3483 3484 if (f2fs_sb_has_readonly(sbi)) 3485 goto skip_cross; 3486 3487 for (j = i + 1; j < NR_CURSEG_DATA_TYPE; j++) { 3488 if (le32_to_cpu(ckpt->cur_data_segno[i]) == 3489 le32_to_cpu(ckpt->cur_data_segno[j])) { 3490 f2fs_err(sbi, "Data segment (%u, %u) has the same segno: %u", 3491 i, j, 3492 le32_to_cpu(ckpt->cur_data_segno[i])); 3493 return 1; 3494 } 3495 } 3496 } 3497 for (i = 0; i < NR_CURSEG_NODE_TYPE; i++) { 3498 for (j = 0; j < NR_CURSEG_DATA_TYPE; j++) { 3499 if (le32_to_cpu(ckpt->cur_node_segno[i]) == 3500 le32_to_cpu(ckpt->cur_data_segno[j])) { 3501 f2fs_err(sbi, "Node segment (%u) and Data segment (%u) has the same segno: %u", 3502 i, j, 3503 le32_to_cpu(ckpt->cur_node_segno[i])); 3504 return 1; 3505 } 3506 } 3507 } 3508 skip_cross: 3509 sit_bitmap_size = le32_to_cpu(ckpt->sit_ver_bitmap_bytesize); 3510 nat_bitmap_size = le32_to_cpu(ckpt->nat_ver_bitmap_bytesize); 3511 3512 if (sit_bitmap_size != ((sit_segs / 2) << log_blocks_per_seg) / 8 || 3513 nat_bitmap_size != ((nat_segs / 2) << log_blocks_per_seg) / 8) { 3514 f2fs_err(sbi, "Wrong bitmap size: sit: %u, nat:%u", 3515 sit_bitmap_size, nat_bitmap_size); 3516 return 1; 3517 } 3518 3519 cp_pack_start_sum = __start_sum_addr(sbi); 3520 cp_payload = __cp_payload(sbi); 3521 if (cp_pack_start_sum < cp_payload + 1 || 3522 cp_pack_start_sum > blocks_per_seg - 1 - 3523 NR_CURSEG_PERSIST_TYPE) { 3524 f2fs_err(sbi, "Wrong cp_pack_start_sum: %u", 3525 cp_pack_start_sum); 3526 return 1; 3527 } 3528 3529 if (__is_set_ckpt_flags(ckpt, CP_LARGE_NAT_BITMAP_FLAG) && 3530 le32_to_cpu(ckpt->checksum_offset) != CP_MIN_CHKSUM_OFFSET) { 3531 f2fs_warn(sbi, "using deprecated layout of large_nat_bitmap, " 3532 "please run fsck v1.13.0 or higher to repair, chksum_offset: %u, " 3533 "fixed with patch: \"f2fs-tools: relocate chksum_offset for large_nat_bitmap feature\"", 3534 le32_to_cpu(ckpt->checksum_offset)); 3535 return 1; 3536 } 3537 3538 nat_blocks = nat_segs << log_blocks_per_seg; 3539 nat_bits_bytes = nat_blocks / BITS_PER_BYTE; 3540 nat_bits_blocks = F2FS_BLK_ALIGN((nat_bits_bytes << 1) + 8); 3541 if (__is_set_ckpt_flags(ckpt, CP_NAT_BITS_FLAG) && 3542 (cp_payload + F2FS_CP_PACKS + 3543 NR_CURSEG_PERSIST_TYPE + nat_bits_blocks >= blocks_per_seg)) { 3544 f2fs_warn(sbi, "Insane cp_payload: %u, nat_bits_blocks: %u)", 3545 cp_payload, nat_bits_blocks); 3546 return 1; 3547 } 3548 3549 if (unlikely(f2fs_cp_error(sbi))) { 3550 f2fs_err(sbi, "A bug case: need to run fsck"); 3551 return 1; 3552 } 3553 return 0; 3554 } 3555 3556 static void init_sb_info(struct f2fs_sb_info *sbi) 3557 { 3558 struct f2fs_super_block *raw_super = sbi->raw_super; 3559 int i; 3560 3561 sbi->log_sectors_per_block = 3562 le32_to_cpu(raw_super->log_sectors_per_block); 3563 sbi->log_blocksize = le32_to_cpu(raw_super->log_blocksize); 3564 sbi->blocksize = 1 << sbi->log_blocksize; 3565 sbi->log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg); 3566 sbi->blocks_per_seg = 1 << sbi->log_blocks_per_seg; 3567 sbi->segs_per_sec = le32_to_cpu(raw_super->segs_per_sec); 3568 sbi->secs_per_zone = le32_to_cpu(raw_super->secs_per_zone); 3569 sbi->total_sections = le32_to_cpu(raw_super->section_count); 3570 sbi->total_node_count = 3571 (le32_to_cpu(raw_super->segment_count_nat) / 2) 3572 * sbi->blocks_per_seg * NAT_ENTRY_PER_BLOCK; 3573 F2FS_ROOT_INO(sbi) = le32_to_cpu(raw_super->root_ino); 3574 F2FS_NODE_INO(sbi) = le32_to_cpu(raw_super->node_ino); 3575 F2FS_META_INO(sbi) = le32_to_cpu(raw_super->meta_ino); 3576 sbi->cur_victim_sec = NULL_SECNO; 3577 sbi->next_victim_seg[BG_GC] = NULL_SEGNO; 3578 sbi->next_victim_seg[FG_GC] = NULL_SEGNO; 3579 sbi->max_victim_search = DEF_MAX_VICTIM_SEARCH; 3580 sbi->migration_granularity = sbi->segs_per_sec; 3581 sbi->seq_file_ra_mul = MIN_RA_MUL; 3582 sbi->max_fragment_chunk = DEF_FRAGMENT_SIZE; 3583 sbi->max_fragment_hole = DEF_FRAGMENT_SIZE; 3584 spin_lock_init(&sbi->gc_urgent_high_lock); 3585 3586 sbi->dir_level = DEF_DIR_LEVEL; 3587 sbi->interval_time[CP_TIME] = DEF_CP_INTERVAL; 3588 sbi->interval_time[REQ_TIME] = DEF_IDLE_INTERVAL; 3589 sbi->interval_time[DISCARD_TIME] = DEF_IDLE_INTERVAL; 3590 sbi->interval_time[GC_TIME] = DEF_IDLE_INTERVAL; 3591 sbi->interval_time[DISABLE_TIME] = DEF_DISABLE_INTERVAL; 3592 sbi->interval_time[UMOUNT_DISCARD_TIMEOUT] = 3593 DEF_UMOUNT_DISCARD_TIMEOUT; 3594 clear_sbi_flag(sbi, SBI_NEED_FSCK); 3595 3596 for (i = 0; i < NR_COUNT_TYPE; i++) 3597 atomic_set(&sbi->nr_pages[i], 0); 3598 3599 for (i = 0; i < META; i++) 3600 atomic_set(&sbi->wb_sync_req[i], 0); 3601 3602 INIT_LIST_HEAD(&sbi->s_list); 3603 mutex_init(&sbi->umount_mutex); 3604 init_rwsem(&sbi->io_order_lock); 3605 spin_lock_init(&sbi->cp_lock); 3606 3607 sbi->dirty_device = 0; 3608 spin_lock_init(&sbi->dev_lock); 3609 3610 init_rwsem(&sbi->sb_lock); 3611 init_rwsem(&sbi->pin_sem); 3612 } 3613 3614 static int init_percpu_info(struct f2fs_sb_info *sbi) 3615 { 3616 int err; 3617 3618 err = percpu_counter_init(&sbi->alloc_valid_block_count, 0, GFP_KERNEL); 3619 if (err) 3620 return err; 3621 3622 err = percpu_counter_init(&sbi->total_valid_inode_count, 0, 3623 GFP_KERNEL); 3624 if (err) 3625 percpu_counter_destroy(&sbi->alloc_valid_block_count); 3626 3627 return err; 3628 } 3629 3630 #ifdef CONFIG_BLK_DEV_ZONED 3631 3632 struct f2fs_report_zones_args { 3633 struct f2fs_dev_info *dev; 3634 bool zone_cap_mismatch; 3635 }; 3636 3637 static int f2fs_report_zone_cb(struct blk_zone *zone, unsigned int idx, 3638 void *data) 3639 { 3640 struct f2fs_report_zones_args *rz_args = data; 3641 3642 if (zone->type == BLK_ZONE_TYPE_CONVENTIONAL) 3643 return 0; 3644 3645 set_bit(idx, rz_args->dev->blkz_seq); 3646 rz_args->dev->zone_capacity_blocks[idx] = zone->capacity >> 3647 F2FS_LOG_SECTORS_PER_BLOCK; 3648 if (zone->len != zone->capacity && !rz_args->zone_cap_mismatch) 3649 rz_args->zone_cap_mismatch = true; 3650 3651 return 0; 3652 } 3653 3654 static int init_blkz_info(struct f2fs_sb_info *sbi, int devi) 3655 { 3656 struct block_device *bdev = FDEV(devi).bdev; 3657 sector_t nr_sectors = bdev_nr_sectors(bdev); 3658 struct f2fs_report_zones_args rep_zone_arg; 3659 int ret; 3660 3661 if (!f2fs_sb_has_blkzoned(sbi)) 3662 return 0; 3663 3664 if (sbi->blocks_per_blkz && sbi->blocks_per_blkz != 3665 SECTOR_TO_BLOCK(bdev_zone_sectors(bdev))) 3666 return -EINVAL; 3667 sbi->blocks_per_blkz = SECTOR_TO_BLOCK(bdev_zone_sectors(bdev)); 3668 if (sbi->log_blocks_per_blkz && sbi->log_blocks_per_blkz != 3669 __ilog2_u32(sbi->blocks_per_blkz)) 3670 return -EINVAL; 3671 sbi->log_blocks_per_blkz = __ilog2_u32(sbi->blocks_per_blkz); 3672 FDEV(devi).nr_blkz = SECTOR_TO_BLOCK(nr_sectors) >> 3673 sbi->log_blocks_per_blkz; 3674 if (nr_sectors & (bdev_zone_sectors(bdev) - 1)) 3675 FDEV(devi).nr_blkz++; 3676 3677 FDEV(devi).blkz_seq = f2fs_kvzalloc(sbi, 3678 BITS_TO_LONGS(FDEV(devi).nr_blkz) 3679 * sizeof(unsigned long), 3680 GFP_KERNEL); 3681 if (!FDEV(devi).blkz_seq) 3682 return -ENOMEM; 3683 3684 /* Get block zones type and zone-capacity */ 3685 FDEV(devi).zone_capacity_blocks = f2fs_kzalloc(sbi, 3686 FDEV(devi).nr_blkz * sizeof(block_t), 3687 GFP_KERNEL); 3688 if (!FDEV(devi).zone_capacity_blocks) 3689 return -ENOMEM; 3690 3691 rep_zone_arg.dev = &FDEV(devi); 3692 rep_zone_arg.zone_cap_mismatch = false; 3693 3694 ret = blkdev_report_zones(bdev, 0, BLK_ALL_ZONES, f2fs_report_zone_cb, 3695 &rep_zone_arg); 3696 if (ret < 0) 3697 return ret; 3698 3699 if (!rep_zone_arg.zone_cap_mismatch) { 3700 kfree(FDEV(devi).zone_capacity_blocks); 3701 FDEV(devi).zone_capacity_blocks = NULL; 3702 } 3703 3704 return 0; 3705 } 3706 #endif 3707 3708 /* 3709 * Read f2fs raw super block. 3710 * Because we have two copies of super block, so read both of them 3711 * to get the first valid one. If any one of them is broken, we pass 3712 * them recovery flag back to the caller. 3713 */ 3714 static int read_raw_super_block(struct f2fs_sb_info *sbi, 3715 struct f2fs_super_block **raw_super, 3716 int *valid_super_block, int *recovery) 3717 { 3718 struct super_block *sb = sbi->sb; 3719 int block; 3720 struct buffer_head *bh; 3721 struct f2fs_super_block *super; 3722 int err = 0; 3723 3724 super = kzalloc(sizeof(struct f2fs_super_block), GFP_KERNEL); 3725 if (!super) 3726 return -ENOMEM; 3727 3728 for (block = 0; block < 2; block++) { 3729 bh = sb_bread(sb, block); 3730 if (!bh) { 3731 f2fs_err(sbi, "Unable to read %dth superblock", 3732 block + 1); 3733 err = -EIO; 3734 *recovery = 1; 3735 continue; 3736 } 3737 3738 /* sanity checking of raw super */ 3739 err = sanity_check_raw_super(sbi, bh); 3740 if (err) { 3741 f2fs_err(sbi, "Can't find valid F2FS filesystem in %dth superblock", 3742 block + 1); 3743 brelse(bh); 3744 *recovery = 1; 3745 continue; 3746 } 3747 3748 if (!*raw_super) { 3749 memcpy(super, bh->b_data + F2FS_SUPER_OFFSET, 3750 sizeof(*super)); 3751 *valid_super_block = block; 3752 *raw_super = super; 3753 } 3754 brelse(bh); 3755 } 3756 3757 /* No valid superblock */ 3758 if (!*raw_super) 3759 kfree(super); 3760 else 3761 err = 0; 3762 3763 return err; 3764 } 3765 3766 int f2fs_commit_super(struct f2fs_sb_info *sbi, bool recover) 3767 { 3768 struct buffer_head *bh; 3769 __u32 crc = 0; 3770 int err; 3771 3772 if ((recover && f2fs_readonly(sbi->sb)) || 3773 bdev_read_only(sbi->sb->s_bdev)) { 3774 set_sbi_flag(sbi, SBI_NEED_SB_WRITE); 3775 return -EROFS; 3776 } 3777 3778 /* we should update superblock crc here */ 3779 if (!recover && f2fs_sb_has_sb_chksum(sbi)) { 3780 crc = f2fs_crc32(sbi, F2FS_RAW_SUPER(sbi), 3781 offsetof(struct f2fs_super_block, crc)); 3782 F2FS_RAW_SUPER(sbi)->crc = cpu_to_le32(crc); 3783 } 3784 3785 /* write back-up superblock first */ 3786 bh = sb_bread(sbi->sb, sbi->valid_super_block ? 0 : 1); 3787 if (!bh) 3788 return -EIO; 3789 err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi)); 3790 brelse(bh); 3791 3792 /* if we are in recovery path, skip writing valid superblock */ 3793 if (recover || err) 3794 return err; 3795 3796 /* write current valid superblock */ 3797 bh = sb_bread(sbi->sb, sbi->valid_super_block); 3798 if (!bh) 3799 return -EIO; 3800 err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi)); 3801 brelse(bh); 3802 return err; 3803 } 3804 3805 static int f2fs_scan_devices(struct f2fs_sb_info *sbi) 3806 { 3807 struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi); 3808 unsigned int max_devices = MAX_DEVICES; 3809 unsigned int logical_blksize; 3810 int i; 3811 3812 /* Initialize single device information */ 3813 if (!RDEV(0).path[0]) { 3814 if (!bdev_is_zoned(sbi->sb->s_bdev)) 3815 return 0; 3816 max_devices = 1; 3817 } 3818 3819 /* 3820 * Initialize multiple devices information, or single 3821 * zoned block device information. 3822 */ 3823 sbi->devs = f2fs_kzalloc(sbi, 3824 array_size(max_devices, 3825 sizeof(struct f2fs_dev_info)), 3826 GFP_KERNEL); 3827 if (!sbi->devs) 3828 return -ENOMEM; 3829 3830 logical_blksize = bdev_logical_block_size(sbi->sb->s_bdev); 3831 sbi->aligned_blksize = true; 3832 3833 for (i = 0; i < max_devices; i++) { 3834 3835 if (i > 0 && !RDEV(i).path[0]) 3836 break; 3837 3838 if (max_devices == 1) { 3839 /* Single zoned block device mount */ 3840 FDEV(0).bdev = 3841 blkdev_get_by_dev(sbi->sb->s_bdev->bd_dev, 3842 sbi->sb->s_mode, sbi->sb->s_type); 3843 } else { 3844 /* Multi-device mount */ 3845 memcpy(FDEV(i).path, RDEV(i).path, MAX_PATH_LEN); 3846 FDEV(i).total_segments = 3847 le32_to_cpu(RDEV(i).total_segments); 3848 if (i == 0) { 3849 FDEV(i).start_blk = 0; 3850 FDEV(i).end_blk = FDEV(i).start_blk + 3851 (FDEV(i).total_segments << 3852 sbi->log_blocks_per_seg) - 1 + 3853 le32_to_cpu(raw_super->segment0_blkaddr); 3854 } else { 3855 FDEV(i).start_blk = FDEV(i - 1).end_blk + 1; 3856 FDEV(i).end_blk = FDEV(i).start_blk + 3857 (FDEV(i).total_segments << 3858 sbi->log_blocks_per_seg) - 1; 3859 } 3860 FDEV(i).bdev = blkdev_get_by_path(FDEV(i).path, 3861 sbi->sb->s_mode, sbi->sb->s_type); 3862 } 3863 if (IS_ERR(FDEV(i).bdev)) 3864 return PTR_ERR(FDEV(i).bdev); 3865 3866 /* to release errored devices */ 3867 sbi->s_ndevs = i + 1; 3868 3869 if (logical_blksize != bdev_logical_block_size(FDEV(i).bdev)) 3870 sbi->aligned_blksize = false; 3871 3872 #ifdef CONFIG_BLK_DEV_ZONED 3873 if (bdev_zoned_model(FDEV(i).bdev) == BLK_ZONED_HM && 3874 !f2fs_sb_has_blkzoned(sbi)) { 3875 f2fs_err(sbi, "Zoned block device feature not enabled"); 3876 return -EINVAL; 3877 } 3878 if (bdev_zoned_model(FDEV(i).bdev) != BLK_ZONED_NONE) { 3879 if (init_blkz_info(sbi, i)) { 3880 f2fs_err(sbi, "Failed to initialize F2FS blkzone information"); 3881 return -EINVAL; 3882 } 3883 if (max_devices == 1) 3884 break; 3885 f2fs_info(sbi, "Mount Device [%2d]: %20s, %8u, %8x - %8x (zone: %s)", 3886 i, FDEV(i).path, 3887 FDEV(i).total_segments, 3888 FDEV(i).start_blk, FDEV(i).end_blk, 3889 bdev_zoned_model(FDEV(i).bdev) == BLK_ZONED_HA ? 3890 "Host-aware" : "Host-managed"); 3891 continue; 3892 } 3893 #endif 3894 f2fs_info(sbi, "Mount Device [%2d]: %20s, %8u, %8x - %8x", 3895 i, FDEV(i).path, 3896 FDEV(i).total_segments, 3897 FDEV(i).start_blk, FDEV(i).end_blk); 3898 } 3899 f2fs_info(sbi, 3900 "IO Block Size: %8d KB", F2FS_IO_SIZE_KB(sbi)); 3901 return 0; 3902 } 3903 3904 static int f2fs_setup_casefold(struct f2fs_sb_info *sbi) 3905 { 3906 #if IS_ENABLED(CONFIG_UNICODE) 3907 if (f2fs_sb_has_casefold(sbi) && !sbi->sb->s_encoding) { 3908 const struct f2fs_sb_encodings *encoding_info; 3909 struct unicode_map *encoding; 3910 __u16 encoding_flags; 3911 3912 encoding_info = f2fs_sb_read_encoding(sbi->raw_super); 3913 if (!encoding_info) { 3914 f2fs_err(sbi, 3915 "Encoding requested by superblock is unknown"); 3916 return -EINVAL; 3917 } 3918 3919 encoding_flags = le16_to_cpu(sbi->raw_super->s_encoding_flags); 3920 encoding = utf8_load(encoding_info->version); 3921 if (IS_ERR(encoding)) { 3922 f2fs_err(sbi, 3923 "can't mount with superblock charset: %s-%u.%u.%u " 3924 "not supported by the kernel. flags: 0x%x.", 3925 encoding_info->name, 3926 unicode_major(encoding_info->version), 3927 unicode_minor(encoding_info->version), 3928 unicode_rev(encoding_info->version), 3929 encoding_flags); 3930 return PTR_ERR(encoding); 3931 } 3932 f2fs_info(sbi, "Using encoding defined by superblock: " 3933 "%s-%u.%u.%u with flags 0x%hx", encoding_info->name, 3934 unicode_major(encoding_info->version), 3935 unicode_minor(encoding_info->version), 3936 unicode_rev(encoding_info->version), 3937 encoding_flags); 3938 3939 sbi->sb->s_encoding = encoding; 3940 sbi->sb->s_encoding_flags = encoding_flags; 3941 } 3942 #else 3943 if (f2fs_sb_has_casefold(sbi)) { 3944 f2fs_err(sbi, "Filesystem with casefold feature cannot be mounted without CONFIG_UNICODE"); 3945 return -EINVAL; 3946 } 3947 #endif 3948 return 0; 3949 } 3950 3951 static void f2fs_tuning_parameters(struct f2fs_sb_info *sbi) 3952 { 3953 struct f2fs_sm_info *sm_i = SM_I(sbi); 3954 3955 /* adjust parameters according to the volume size */ 3956 if (sm_i->main_segments <= SMALL_VOLUME_SEGMENTS) { 3957 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_REUSE; 3958 if (f2fs_block_unit_discard(sbi)) 3959 sm_i->dcc_info->discard_granularity = 1; 3960 sm_i->ipu_policy = 1 << F2FS_IPU_FORCE; 3961 } 3962 3963 sbi->readdir_ra = 1; 3964 } 3965 3966 static int f2fs_fill_super(struct super_block *sb, void *data, int silent) 3967 { 3968 struct f2fs_sb_info *sbi; 3969 struct f2fs_super_block *raw_super; 3970 struct inode *root; 3971 int err; 3972 bool skip_recovery = false, need_fsck = false; 3973 char *options = NULL; 3974 int recovery, i, valid_super_block; 3975 struct curseg_info *seg_i; 3976 int retry_cnt = 1; 3977 3978 try_onemore: 3979 err = -EINVAL; 3980 raw_super = NULL; 3981 valid_super_block = -1; 3982 recovery = 0; 3983 3984 /* allocate memory for f2fs-specific super block info */ 3985 sbi = kzalloc(sizeof(struct f2fs_sb_info), GFP_KERNEL); 3986 if (!sbi) 3987 return -ENOMEM; 3988 3989 sbi->sb = sb; 3990 3991 /* Load the checksum driver */ 3992 sbi->s_chksum_driver = crypto_alloc_shash("crc32", 0, 0); 3993 if (IS_ERR(sbi->s_chksum_driver)) { 3994 f2fs_err(sbi, "Cannot load crc32 driver."); 3995 err = PTR_ERR(sbi->s_chksum_driver); 3996 sbi->s_chksum_driver = NULL; 3997 goto free_sbi; 3998 } 3999 4000 /* set a block size */ 4001 if (unlikely(!sb_set_blocksize(sb, F2FS_BLKSIZE))) { 4002 f2fs_err(sbi, "unable to set blocksize"); 4003 goto free_sbi; 4004 } 4005 4006 err = read_raw_super_block(sbi, &raw_super, &valid_super_block, 4007 &recovery); 4008 if (err) 4009 goto free_sbi; 4010 4011 sb->s_fs_info = sbi; 4012 sbi->raw_super = raw_super; 4013 4014 /* precompute checksum seed for metadata */ 4015 if (f2fs_sb_has_inode_chksum(sbi)) 4016 sbi->s_chksum_seed = f2fs_chksum(sbi, ~0, raw_super->uuid, 4017 sizeof(raw_super->uuid)); 4018 4019 default_options(sbi); 4020 /* parse mount options */ 4021 options = kstrdup((const char *)data, GFP_KERNEL); 4022 if (data && !options) { 4023 err = -ENOMEM; 4024 goto free_sb_buf; 4025 } 4026 4027 err = parse_options(sb, options, false); 4028 if (err) 4029 goto free_options; 4030 4031 sb->s_maxbytes = max_file_blocks(NULL) << 4032 le32_to_cpu(raw_super->log_blocksize); 4033 sb->s_max_links = F2FS_LINK_MAX; 4034 4035 err = f2fs_setup_casefold(sbi); 4036 if (err) 4037 goto free_options; 4038 4039 #ifdef CONFIG_QUOTA 4040 sb->dq_op = &f2fs_quota_operations; 4041 sb->s_qcop = &f2fs_quotactl_ops; 4042 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ; 4043 4044 if (f2fs_sb_has_quota_ino(sbi)) { 4045 for (i = 0; i < MAXQUOTAS; i++) { 4046 if (f2fs_qf_ino(sbi->sb, i)) 4047 sbi->nquota_files++; 4048 } 4049 } 4050 #endif 4051 4052 sb->s_op = &f2fs_sops; 4053 #ifdef CONFIG_FS_ENCRYPTION 4054 sb->s_cop = &f2fs_cryptops; 4055 #endif 4056 #ifdef CONFIG_FS_VERITY 4057 sb->s_vop = &f2fs_verityops; 4058 #endif 4059 sb->s_xattr = f2fs_xattr_handlers; 4060 sb->s_export_op = &f2fs_export_ops; 4061 sb->s_magic = F2FS_SUPER_MAGIC; 4062 sb->s_time_gran = 1; 4063 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) | 4064 (test_opt(sbi, POSIX_ACL) ? SB_POSIXACL : 0); 4065 memcpy(&sb->s_uuid, raw_super->uuid, sizeof(raw_super->uuid)); 4066 sb->s_iflags |= SB_I_CGROUPWB; 4067 4068 /* init f2fs-specific super block info */ 4069 sbi->valid_super_block = valid_super_block; 4070 init_rwsem(&sbi->gc_lock); 4071 mutex_init(&sbi->writepages); 4072 init_rwsem(&sbi->cp_global_sem); 4073 init_rwsem(&sbi->node_write); 4074 init_rwsem(&sbi->node_change); 4075 4076 /* disallow all the data/node/meta page writes */ 4077 set_sbi_flag(sbi, SBI_POR_DOING); 4078 spin_lock_init(&sbi->stat_lock); 4079 4080 for (i = 0; i < NR_PAGE_TYPE; i++) { 4081 int n = (i == META) ? 1 : NR_TEMP_TYPE; 4082 int j; 4083 4084 sbi->write_io[i] = 4085 f2fs_kmalloc(sbi, 4086 array_size(n, 4087 sizeof(struct f2fs_bio_info)), 4088 GFP_KERNEL); 4089 if (!sbi->write_io[i]) { 4090 err = -ENOMEM; 4091 goto free_bio_info; 4092 } 4093 4094 for (j = HOT; j < n; j++) { 4095 init_rwsem(&sbi->write_io[i][j].io_rwsem); 4096 sbi->write_io[i][j].sbi = sbi; 4097 sbi->write_io[i][j].bio = NULL; 4098 spin_lock_init(&sbi->write_io[i][j].io_lock); 4099 INIT_LIST_HEAD(&sbi->write_io[i][j].io_list); 4100 INIT_LIST_HEAD(&sbi->write_io[i][j].bio_list); 4101 init_rwsem(&sbi->write_io[i][j].bio_list_lock); 4102 } 4103 } 4104 4105 init_rwsem(&sbi->cp_rwsem); 4106 init_rwsem(&sbi->quota_sem); 4107 init_waitqueue_head(&sbi->cp_wait); 4108 init_sb_info(sbi); 4109 4110 err = f2fs_init_iostat(sbi); 4111 if (err) 4112 goto free_bio_info; 4113 4114 err = init_percpu_info(sbi); 4115 if (err) 4116 goto free_iostat; 4117 4118 if (F2FS_IO_ALIGNED(sbi)) { 4119 sbi->write_io_dummy = 4120 mempool_create_page_pool(2 * (F2FS_IO_SIZE(sbi) - 1), 0); 4121 if (!sbi->write_io_dummy) { 4122 err = -ENOMEM; 4123 goto free_percpu; 4124 } 4125 } 4126 4127 /* init per sbi slab cache */ 4128 err = f2fs_init_xattr_caches(sbi); 4129 if (err) 4130 goto free_io_dummy; 4131 err = f2fs_init_page_array_cache(sbi); 4132 if (err) 4133 goto free_xattr_cache; 4134 4135 /* get an inode for meta space */ 4136 sbi->meta_inode = f2fs_iget(sb, F2FS_META_INO(sbi)); 4137 if (IS_ERR(sbi->meta_inode)) { 4138 f2fs_err(sbi, "Failed to read F2FS meta data inode"); 4139 err = PTR_ERR(sbi->meta_inode); 4140 goto free_page_array_cache; 4141 } 4142 4143 err = f2fs_get_valid_checkpoint(sbi); 4144 if (err) { 4145 f2fs_err(sbi, "Failed to get valid F2FS checkpoint"); 4146 goto free_meta_inode; 4147 } 4148 4149 if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_QUOTA_NEED_FSCK_FLAG)) 4150 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR); 4151 if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_DISABLED_QUICK_FLAG)) { 4152 set_sbi_flag(sbi, SBI_CP_DISABLED_QUICK); 4153 sbi->interval_time[DISABLE_TIME] = DEF_DISABLE_QUICK_INTERVAL; 4154 } 4155 4156 if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_FSCK_FLAG)) 4157 set_sbi_flag(sbi, SBI_NEED_FSCK); 4158 4159 /* Initialize device list */ 4160 err = f2fs_scan_devices(sbi); 4161 if (err) { 4162 f2fs_err(sbi, "Failed to find devices"); 4163 goto free_devices; 4164 } 4165 4166 err = f2fs_init_post_read_wq(sbi); 4167 if (err) { 4168 f2fs_err(sbi, "Failed to initialize post read workqueue"); 4169 goto free_devices; 4170 } 4171 4172 sbi->total_valid_node_count = 4173 le32_to_cpu(sbi->ckpt->valid_node_count); 4174 percpu_counter_set(&sbi->total_valid_inode_count, 4175 le32_to_cpu(sbi->ckpt->valid_inode_count)); 4176 sbi->user_block_count = le64_to_cpu(sbi->ckpt->user_block_count); 4177 sbi->total_valid_block_count = 4178 le64_to_cpu(sbi->ckpt->valid_block_count); 4179 sbi->last_valid_block_count = sbi->total_valid_block_count; 4180 sbi->reserved_blocks = 0; 4181 sbi->current_reserved_blocks = 0; 4182 limit_reserve_root(sbi); 4183 adjust_unusable_cap_perc(sbi); 4184 4185 for (i = 0; i < NR_INODE_TYPE; i++) { 4186 INIT_LIST_HEAD(&sbi->inode_list[i]); 4187 spin_lock_init(&sbi->inode_lock[i]); 4188 } 4189 mutex_init(&sbi->flush_lock); 4190 4191 f2fs_init_extent_cache_info(sbi); 4192 4193 f2fs_init_ino_entry_info(sbi); 4194 4195 f2fs_init_fsync_node_info(sbi); 4196 4197 /* setup checkpoint request control and start checkpoint issue thread */ 4198 f2fs_init_ckpt_req_control(sbi); 4199 if (!f2fs_readonly(sb) && !test_opt(sbi, DISABLE_CHECKPOINT) && 4200 test_opt(sbi, MERGE_CHECKPOINT)) { 4201 err = f2fs_start_ckpt_thread(sbi); 4202 if (err) { 4203 f2fs_err(sbi, 4204 "Failed to start F2FS issue_checkpoint_thread (%d)", 4205 err); 4206 goto stop_ckpt_thread; 4207 } 4208 } 4209 4210 /* setup f2fs internal modules */ 4211 err = f2fs_build_segment_manager(sbi); 4212 if (err) { 4213 f2fs_err(sbi, "Failed to initialize F2FS segment manager (%d)", 4214 err); 4215 goto free_sm; 4216 } 4217 err = f2fs_build_node_manager(sbi); 4218 if (err) { 4219 f2fs_err(sbi, "Failed to initialize F2FS node manager (%d)", 4220 err); 4221 goto free_nm; 4222 } 4223 4224 err = adjust_reserved_segment(sbi); 4225 if (err) 4226 goto free_nm; 4227 4228 /* For write statistics */ 4229 sbi->sectors_written_start = f2fs_get_sectors_written(sbi); 4230 4231 /* Read accumulated write IO statistics if exists */ 4232 seg_i = CURSEG_I(sbi, CURSEG_HOT_NODE); 4233 if (__exist_node_summaries(sbi)) 4234 sbi->kbytes_written = 4235 le64_to_cpu(seg_i->journal->info.kbytes_written); 4236 4237 f2fs_build_gc_manager(sbi); 4238 4239 err = f2fs_build_stats(sbi); 4240 if (err) 4241 goto free_nm; 4242 4243 /* get an inode for node space */ 4244 sbi->node_inode = f2fs_iget(sb, F2FS_NODE_INO(sbi)); 4245 if (IS_ERR(sbi->node_inode)) { 4246 f2fs_err(sbi, "Failed to read node inode"); 4247 err = PTR_ERR(sbi->node_inode); 4248 goto free_stats; 4249 } 4250 4251 /* read root inode and dentry */ 4252 root = f2fs_iget(sb, F2FS_ROOT_INO(sbi)); 4253 if (IS_ERR(root)) { 4254 f2fs_err(sbi, "Failed to read root inode"); 4255 err = PTR_ERR(root); 4256 goto free_node_inode; 4257 } 4258 if (!S_ISDIR(root->i_mode) || !root->i_blocks || 4259 !root->i_size || !root->i_nlink) { 4260 iput(root); 4261 err = -EINVAL; 4262 goto free_node_inode; 4263 } 4264 4265 sb->s_root = d_make_root(root); /* allocate root dentry */ 4266 if (!sb->s_root) { 4267 err = -ENOMEM; 4268 goto free_node_inode; 4269 } 4270 4271 err = f2fs_init_compress_inode(sbi); 4272 if (err) 4273 goto free_root_inode; 4274 4275 err = f2fs_register_sysfs(sbi); 4276 if (err) 4277 goto free_compress_inode; 4278 4279 #ifdef CONFIG_QUOTA 4280 /* Enable quota usage during mount */ 4281 if (f2fs_sb_has_quota_ino(sbi) && !f2fs_readonly(sb)) { 4282 err = f2fs_enable_quotas(sb); 4283 if (err) 4284 f2fs_err(sbi, "Cannot turn on quotas: error %d", err); 4285 } 4286 #endif 4287 /* if there are any orphan inodes, free them */ 4288 err = f2fs_recover_orphan_inodes(sbi); 4289 if (err) 4290 goto free_meta; 4291 4292 if (unlikely(is_set_ckpt_flags(sbi, CP_DISABLED_FLAG))) 4293 goto reset_checkpoint; 4294 4295 /* recover fsynced data */ 4296 if (!test_opt(sbi, DISABLE_ROLL_FORWARD) && 4297 !test_opt(sbi, NORECOVERY)) { 4298 /* 4299 * mount should be failed, when device has readonly mode, and 4300 * previous checkpoint was not done by clean system shutdown. 4301 */ 4302 if (f2fs_hw_is_readonly(sbi)) { 4303 if (!is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG)) { 4304 err = f2fs_recover_fsync_data(sbi, true); 4305 if (err > 0) { 4306 err = -EROFS; 4307 f2fs_err(sbi, "Need to recover fsync data, but " 4308 "write access unavailable, please try " 4309 "mount w/ disable_roll_forward or norecovery"); 4310 } 4311 if (err < 0) 4312 goto free_meta; 4313 } 4314 f2fs_info(sbi, "write access unavailable, skipping recovery"); 4315 goto reset_checkpoint; 4316 } 4317 4318 if (need_fsck) 4319 set_sbi_flag(sbi, SBI_NEED_FSCK); 4320 4321 if (skip_recovery) 4322 goto reset_checkpoint; 4323 4324 err = f2fs_recover_fsync_data(sbi, false); 4325 if (err < 0) { 4326 if (err != -ENOMEM) 4327 skip_recovery = true; 4328 need_fsck = true; 4329 f2fs_err(sbi, "Cannot recover all fsync data errno=%d", 4330 err); 4331 goto free_meta; 4332 } 4333 } else { 4334 err = f2fs_recover_fsync_data(sbi, true); 4335 4336 if (!f2fs_readonly(sb) && err > 0) { 4337 err = -EINVAL; 4338 f2fs_err(sbi, "Need to recover fsync data"); 4339 goto free_meta; 4340 } 4341 } 4342 4343 /* 4344 * If the f2fs is not readonly and fsync data recovery succeeds, 4345 * check zoned block devices' write pointer consistency. 4346 */ 4347 if (!err && !f2fs_readonly(sb) && f2fs_sb_has_blkzoned(sbi)) { 4348 err = f2fs_check_write_pointer(sbi); 4349 if (err) 4350 goto free_meta; 4351 } 4352 4353 reset_checkpoint: 4354 f2fs_init_inmem_curseg(sbi); 4355 4356 /* f2fs_recover_fsync_data() cleared this already */ 4357 clear_sbi_flag(sbi, SBI_POR_DOING); 4358 4359 if (test_opt(sbi, DISABLE_CHECKPOINT)) { 4360 err = f2fs_disable_checkpoint(sbi); 4361 if (err) 4362 goto sync_free_meta; 4363 } else if (is_set_ckpt_flags(sbi, CP_DISABLED_FLAG)) { 4364 f2fs_enable_checkpoint(sbi); 4365 } 4366 4367 /* 4368 * If filesystem is not mounted as read-only then 4369 * do start the gc_thread. 4370 */ 4371 if ((F2FS_OPTION(sbi).bggc_mode != BGGC_MODE_OFF || 4372 test_opt(sbi, GC_MERGE)) && !f2fs_readonly(sb)) { 4373 /* After POR, we can run background GC thread.*/ 4374 err = f2fs_start_gc_thread(sbi); 4375 if (err) 4376 goto sync_free_meta; 4377 } 4378 kvfree(options); 4379 4380 /* recover broken superblock */ 4381 if (recovery) { 4382 err = f2fs_commit_super(sbi, true); 4383 f2fs_info(sbi, "Try to recover %dth superblock, ret: %d", 4384 sbi->valid_super_block ? 1 : 2, err); 4385 } 4386 4387 f2fs_join_shrinker(sbi); 4388 4389 f2fs_tuning_parameters(sbi); 4390 4391 f2fs_notice(sbi, "Mounted with checkpoint version = %llx", 4392 cur_cp_version(F2FS_CKPT(sbi))); 4393 f2fs_update_time(sbi, CP_TIME); 4394 f2fs_update_time(sbi, REQ_TIME); 4395 clear_sbi_flag(sbi, SBI_CP_DISABLED_QUICK); 4396 return 0; 4397 4398 sync_free_meta: 4399 /* safe to flush all the data */ 4400 sync_filesystem(sbi->sb); 4401 retry_cnt = 0; 4402 4403 free_meta: 4404 #ifdef CONFIG_QUOTA 4405 f2fs_truncate_quota_inode_pages(sb); 4406 if (f2fs_sb_has_quota_ino(sbi) && !f2fs_readonly(sb)) 4407 f2fs_quota_off_umount(sbi->sb); 4408 #endif 4409 /* 4410 * Some dirty meta pages can be produced by f2fs_recover_orphan_inodes() 4411 * failed by EIO. Then, iput(node_inode) can trigger balance_fs_bg() 4412 * followed by f2fs_write_checkpoint() through f2fs_write_node_pages(), which 4413 * falls into an infinite loop in f2fs_sync_meta_pages(). 4414 */ 4415 truncate_inode_pages_final(META_MAPPING(sbi)); 4416 /* evict some inodes being cached by GC */ 4417 evict_inodes(sb); 4418 f2fs_unregister_sysfs(sbi); 4419 free_compress_inode: 4420 f2fs_destroy_compress_inode(sbi); 4421 free_root_inode: 4422 dput(sb->s_root); 4423 sb->s_root = NULL; 4424 free_node_inode: 4425 f2fs_release_ino_entry(sbi, true); 4426 truncate_inode_pages_final(NODE_MAPPING(sbi)); 4427 iput(sbi->node_inode); 4428 sbi->node_inode = NULL; 4429 free_stats: 4430 f2fs_destroy_stats(sbi); 4431 free_nm: 4432 /* stop discard thread before destroying node manager */ 4433 f2fs_stop_discard_thread(sbi); 4434 f2fs_destroy_node_manager(sbi); 4435 free_sm: 4436 f2fs_destroy_segment_manager(sbi); 4437 f2fs_destroy_post_read_wq(sbi); 4438 stop_ckpt_thread: 4439 f2fs_stop_ckpt_thread(sbi); 4440 free_devices: 4441 destroy_device_list(sbi); 4442 kvfree(sbi->ckpt); 4443 free_meta_inode: 4444 make_bad_inode(sbi->meta_inode); 4445 iput(sbi->meta_inode); 4446 sbi->meta_inode = NULL; 4447 free_page_array_cache: 4448 f2fs_destroy_page_array_cache(sbi); 4449 free_xattr_cache: 4450 f2fs_destroy_xattr_caches(sbi); 4451 free_io_dummy: 4452 mempool_destroy(sbi->write_io_dummy); 4453 free_percpu: 4454 destroy_percpu_info(sbi); 4455 free_iostat: 4456 f2fs_destroy_iostat(sbi); 4457 free_bio_info: 4458 for (i = 0; i < NR_PAGE_TYPE; i++) 4459 kvfree(sbi->write_io[i]); 4460 4461 #if IS_ENABLED(CONFIG_UNICODE) 4462 utf8_unload(sb->s_encoding); 4463 sb->s_encoding = NULL; 4464 #endif 4465 free_options: 4466 #ifdef CONFIG_QUOTA 4467 for (i = 0; i < MAXQUOTAS; i++) 4468 kfree(F2FS_OPTION(sbi).s_qf_names[i]); 4469 #endif 4470 fscrypt_free_dummy_policy(&F2FS_OPTION(sbi).dummy_enc_policy); 4471 kvfree(options); 4472 free_sb_buf: 4473 kfree(raw_super); 4474 free_sbi: 4475 if (sbi->s_chksum_driver) 4476 crypto_free_shash(sbi->s_chksum_driver); 4477 kfree(sbi); 4478 4479 /* give only one another chance */ 4480 if (retry_cnt > 0 && skip_recovery) { 4481 retry_cnt--; 4482 shrink_dcache_sb(sb); 4483 goto try_onemore; 4484 } 4485 return err; 4486 } 4487 4488 static struct dentry *f2fs_mount(struct file_system_type *fs_type, int flags, 4489 const char *dev_name, void *data) 4490 { 4491 return mount_bdev(fs_type, flags, dev_name, data, f2fs_fill_super); 4492 } 4493 4494 static void kill_f2fs_super(struct super_block *sb) 4495 { 4496 if (sb->s_root) { 4497 struct f2fs_sb_info *sbi = F2FS_SB(sb); 4498 4499 set_sbi_flag(sbi, SBI_IS_CLOSE); 4500 f2fs_stop_gc_thread(sbi); 4501 f2fs_stop_discard_thread(sbi); 4502 4503 #ifdef CONFIG_F2FS_FS_COMPRESSION 4504 /* 4505 * latter evict_inode() can bypass checking and invalidating 4506 * compress inode cache. 4507 */ 4508 if (test_opt(sbi, COMPRESS_CACHE)) 4509 truncate_inode_pages_final(COMPRESS_MAPPING(sbi)); 4510 #endif 4511 4512 if (is_sbi_flag_set(sbi, SBI_IS_DIRTY) || 4513 !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG)) { 4514 struct cp_control cpc = { 4515 .reason = CP_UMOUNT, 4516 }; 4517 f2fs_write_checkpoint(sbi, &cpc); 4518 } 4519 4520 if (is_sbi_flag_set(sbi, SBI_IS_RECOVERED) && f2fs_readonly(sb)) 4521 sb->s_flags &= ~SB_RDONLY; 4522 } 4523 kill_block_super(sb); 4524 } 4525 4526 static struct file_system_type f2fs_fs_type = { 4527 .owner = THIS_MODULE, 4528 .name = "f2fs", 4529 .mount = f2fs_mount, 4530 .kill_sb = kill_f2fs_super, 4531 .fs_flags = FS_REQUIRES_DEV, 4532 }; 4533 MODULE_ALIAS_FS("f2fs"); 4534 4535 static int __init init_inodecache(void) 4536 { 4537 f2fs_inode_cachep = kmem_cache_create("f2fs_inode_cache", 4538 sizeof(struct f2fs_inode_info), 0, 4539 SLAB_RECLAIM_ACCOUNT|SLAB_ACCOUNT, NULL); 4540 if (!f2fs_inode_cachep) 4541 return -ENOMEM; 4542 return 0; 4543 } 4544 4545 static void destroy_inodecache(void) 4546 { 4547 /* 4548 * Make sure all delayed rcu free inodes are flushed before we 4549 * destroy cache. 4550 */ 4551 rcu_barrier(); 4552 kmem_cache_destroy(f2fs_inode_cachep); 4553 } 4554 4555 static int __init init_f2fs_fs(void) 4556 { 4557 int err; 4558 4559 if (PAGE_SIZE != F2FS_BLKSIZE) { 4560 printk("F2FS not supported on PAGE_SIZE(%lu) != %d\n", 4561 PAGE_SIZE, F2FS_BLKSIZE); 4562 return -EINVAL; 4563 } 4564 4565 err = init_inodecache(); 4566 if (err) 4567 goto fail; 4568 err = f2fs_create_node_manager_caches(); 4569 if (err) 4570 goto free_inodecache; 4571 err = f2fs_create_segment_manager_caches(); 4572 if (err) 4573 goto free_node_manager_caches; 4574 err = f2fs_create_checkpoint_caches(); 4575 if (err) 4576 goto free_segment_manager_caches; 4577 err = f2fs_create_recovery_cache(); 4578 if (err) 4579 goto free_checkpoint_caches; 4580 err = f2fs_create_extent_cache(); 4581 if (err) 4582 goto free_recovery_cache; 4583 err = f2fs_create_garbage_collection_cache(); 4584 if (err) 4585 goto free_extent_cache; 4586 err = f2fs_init_sysfs(); 4587 if (err) 4588 goto free_garbage_collection_cache; 4589 err = register_shrinker(&f2fs_shrinker_info); 4590 if (err) 4591 goto free_sysfs; 4592 err = register_filesystem(&f2fs_fs_type); 4593 if (err) 4594 goto free_shrinker; 4595 f2fs_create_root_stats(); 4596 err = f2fs_init_post_read_processing(); 4597 if (err) 4598 goto free_root_stats; 4599 err = f2fs_init_iostat_processing(); 4600 if (err) 4601 goto free_post_read; 4602 err = f2fs_init_bio_entry_cache(); 4603 if (err) 4604 goto free_iostat; 4605 err = f2fs_init_bioset(); 4606 if (err) 4607 goto free_bio_enrty_cache; 4608 err = f2fs_init_compress_mempool(); 4609 if (err) 4610 goto free_bioset; 4611 err = f2fs_init_compress_cache(); 4612 if (err) 4613 goto free_compress_mempool; 4614 err = f2fs_create_casefold_cache(); 4615 if (err) 4616 goto free_compress_cache; 4617 return 0; 4618 free_compress_cache: 4619 f2fs_destroy_compress_cache(); 4620 free_compress_mempool: 4621 f2fs_destroy_compress_mempool(); 4622 free_bioset: 4623 f2fs_destroy_bioset(); 4624 free_bio_enrty_cache: 4625 f2fs_destroy_bio_entry_cache(); 4626 free_iostat: 4627 f2fs_destroy_iostat_processing(); 4628 free_post_read: 4629 f2fs_destroy_post_read_processing(); 4630 free_root_stats: 4631 f2fs_destroy_root_stats(); 4632 unregister_filesystem(&f2fs_fs_type); 4633 free_shrinker: 4634 unregister_shrinker(&f2fs_shrinker_info); 4635 free_sysfs: 4636 f2fs_exit_sysfs(); 4637 free_garbage_collection_cache: 4638 f2fs_destroy_garbage_collection_cache(); 4639 free_extent_cache: 4640 f2fs_destroy_extent_cache(); 4641 free_recovery_cache: 4642 f2fs_destroy_recovery_cache(); 4643 free_checkpoint_caches: 4644 f2fs_destroy_checkpoint_caches(); 4645 free_segment_manager_caches: 4646 f2fs_destroy_segment_manager_caches(); 4647 free_node_manager_caches: 4648 f2fs_destroy_node_manager_caches(); 4649 free_inodecache: 4650 destroy_inodecache(); 4651 fail: 4652 return err; 4653 } 4654 4655 static void __exit exit_f2fs_fs(void) 4656 { 4657 f2fs_destroy_casefold_cache(); 4658 f2fs_destroy_compress_cache(); 4659 f2fs_destroy_compress_mempool(); 4660 f2fs_destroy_bioset(); 4661 f2fs_destroy_bio_entry_cache(); 4662 f2fs_destroy_iostat_processing(); 4663 f2fs_destroy_post_read_processing(); 4664 f2fs_destroy_root_stats(); 4665 unregister_filesystem(&f2fs_fs_type); 4666 unregister_shrinker(&f2fs_shrinker_info); 4667 f2fs_exit_sysfs(); 4668 f2fs_destroy_garbage_collection_cache(); 4669 f2fs_destroy_extent_cache(); 4670 f2fs_destroy_recovery_cache(); 4671 f2fs_destroy_checkpoint_caches(); 4672 f2fs_destroy_segment_manager_caches(); 4673 f2fs_destroy_node_manager_caches(); 4674 destroy_inodecache(); 4675 } 4676 4677 module_init(init_f2fs_fs) 4678 module_exit(exit_f2fs_fs) 4679 4680 MODULE_AUTHOR("Samsung Electronics's Praesto Team"); 4681 MODULE_DESCRIPTION("Flash Friendly File System"); 4682 MODULE_LICENSE("GPL"); 4683 MODULE_SOFTDEP("pre: crc32"); 4684 4685