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 if (time_to_inject(F2FS_SB(sb), FAULT_SLAB_ALLOC)) { 1349 f2fs_show_injection_info(F2FS_SB(sb), FAULT_SLAB_ALLOC); 1350 return NULL; 1351 } 1352 1353 fi = alloc_inode_sb(sb, f2fs_inode_cachep, GFP_F2FS_ZERO); 1354 if (!fi) 1355 return NULL; 1356 1357 init_once((void *) fi); 1358 1359 /* Initialize f2fs-specific inode info */ 1360 atomic_set(&fi->dirty_pages, 0); 1361 atomic_set(&fi->i_compr_blocks, 0); 1362 init_rwsem(&fi->i_sem); 1363 spin_lock_init(&fi->i_size_lock); 1364 INIT_LIST_HEAD(&fi->dirty_list); 1365 INIT_LIST_HEAD(&fi->gdirty_list); 1366 INIT_LIST_HEAD(&fi->inmem_ilist); 1367 INIT_LIST_HEAD(&fi->inmem_pages); 1368 mutex_init(&fi->inmem_lock); 1369 init_rwsem(&fi->i_gc_rwsem[READ]); 1370 init_rwsem(&fi->i_gc_rwsem[WRITE]); 1371 init_rwsem(&fi->i_xattr_sem); 1372 1373 /* Will be used by directory only */ 1374 fi->i_dir_level = F2FS_SB(sb)->dir_level; 1375 1376 return &fi->vfs_inode; 1377 } 1378 1379 static int f2fs_drop_inode(struct inode *inode) 1380 { 1381 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1382 int ret; 1383 1384 /* 1385 * during filesystem shutdown, if checkpoint is disabled, 1386 * drop useless meta/node dirty pages. 1387 */ 1388 if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED))) { 1389 if (inode->i_ino == F2FS_NODE_INO(sbi) || 1390 inode->i_ino == F2FS_META_INO(sbi)) { 1391 trace_f2fs_drop_inode(inode, 1); 1392 return 1; 1393 } 1394 } 1395 1396 /* 1397 * This is to avoid a deadlock condition like below. 1398 * writeback_single_inode(inode) 1399 * - f2fs_write_data_page 1400 * - f2fs_gc -> iput -> evict 1401 * - inode_wait_for_writeback(inode) 1402 */ 1403 if ((!inode_unhashed(inode) && inode->i_state & I_SYNC)) { 1404 if (!inode->i_nlink && !is_bad_inode(inode)) { 1405 /* to avoid evict_inode call simultaneously */ 1406 atomic_inc(&inode->i_count); 1407 spin_unlock(&inode->i_lock); 1408 1409 /* some remained atomic pages should discarded */ 1410 if (f2fs_is_atomic_file(inode)) 1411 f2fs_drop_inmem_pages(inode); 1412 1413 /* should remain fi->extent_tree for writepage */ 1414 f2fs_destroy_extent_node(inode); 1415 1416 sb_start_intwrite(inode->i_sb); 1417 f2fs_i_size_write(inode, 0); 1418 1419 f2fs_submit_merged_write_cond(F2FS_I_SB(inode), 1420 inode, NULL, 0, DATA); 1421 truncate_inode_pages_final(inode->i_mapping); 1422 1423 if (F2FS_HAS_BLOCKS(inode)) 1424 f2fs_truncate(inode); 1425 1426 sb_end_intwrite(inode->i_sb); 1427 1428 spin_lock(&inode->i_lock); 1429 atomic_dec(&inode->i_count); 1430 } 1431 trace_f2fs_drop_inode(inode, 0); 1432 return 0; 1433 } 1434 ret = generic_drop_inode(inode); 1435 if (!ret) 1436 ret = fscrypt_drop_inode(inode); 1437 trace_f2fs_drop_inode(inode, ret); 1438 return ret; 1439 } 1440 1441 int f2fs_inode_dirtied(struct inode *inode, bool sync) 1442 { 1443 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1444 int ret = 0; 1445 1446 spin_lock(&sbi->inode_lock[DIRTY_META]); 1447 if (is_inode_flag_set(inode, FI_DIRTY_INODE)) { 1448 ret = 1; 1449 } else { 1450 set_inode_flag(inode, FI_DIRTY_INODE); 1451 stat_inc_dirty_inode(sbi, DIRTY_META); 1452 } 1453 if (sync && list_empty(&F2FS_I(inode)->gdirty_list)) { 1454 list_add_tail(&F2FS_I(inode)->gdirty_list, 1455 &sbi->inode_list[DIRTY_META]); 1456 inc_page_count(sbi, F2FS_DIRTY_IMETA); 1457 } 1458 spin_unlock(&sbi->inode_lock[DIRTY_META]); 1459 return ret; 1460 } 1461 1462 void f2fs_inode_synced(struct inode *inode) 1463 { 1464 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1465 1466 spin_lock(&sbi->inode_lock[DIRTY_META]); 1467 if (!is_inode_flag_set(inode, FI_DIRTY_INODE)) { 1468 spin_unlock(&sbi->inode_lock[DIRTY_META]); 1469 return; 1470 } 1471 if (!list_empty(&F2FS_I(inode)->gdirty_list)) { 1472 list_del_init(&F2FS_I(inode)->gdirty_list); 1473 dec_page_count(sbi, F2FS_DIRTY_IMETA); 1474 } 1475 clear_inode_flag(inode, FI_DIRTY_INODE); 1476 clear_inode_flag(inode, FI_AUTO_RECOVER); 1477 stat_dec_dirty_inode(F2FS_I_SB(inode), DIRTY_META); 1478 spin_unlock(&sbi->inode_lock[DIRTY_META]); 1479 } 1480 1481 /* 1482 * f2fs_dirty_inode() is called from __mark_inode_dirty() 1483 * 1484 * We should call set_dirty_inode to write the dirty inode through write_inode. 1485 */ 1486 static void f2fs_dirty_inode(struct inode *inode, int flags) 1487 { 1488 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1489 1490 if (inode->i_ino == F2FS_NODE_INO(sbi) || 1491 inode->i_ino == F2FS_META_INO(sbi)) 1492 return; 1493 1494 if (is_inode_flag_set(inode, FI_AUTO_RECOVER)) 1495 clear_inode_flag(inode, FI_AUTO_RECOVER); 1496 1497 f2fs_inode_dirtied(inode, false); 1498 } 1499 1500 static void f2fs_free_inode(struct inode *inode) 1501 { 1502 fscrypt_free_inode(inode); 1503 kmem_cache_free(f2fs_inode_cachep, F2FS_I(inode)); 1504 } 1505 1506 static void destroy_percpu_info(struct f2fs_sb_info *sbi) 1507 { 1508 percpu_counter_destroy(&sbi->alloc_valid_block_count); 1509 percpu_counter_destroy(&sbi->total_valid_inode_count); 1510 } 1511 1512 static void destroy_device_list(struct f2fs_sb_info *sbi) 1513 { 1514 int i; 1515 1516 for (i = 0; i < sbi->s_ndevs; i++) { 1517 blkdev_put(FDEV(i).bdev, FMODE_EXCL); 1518 #ifdef CONFIG_BLK_DEV_ZONED 1519 kvfree(FDEV(i).blkz_seq); 1520 kfree(FDEV(i).zone_capacity_blocks); 1521 #endif 1522 } 1523 kvfree(sbi->devs); 1524 } 1525 1526 static void f2fs_put_super(struct super_block *sb) 1527 { 1528 struct f2fs_sb_info *sbi = F2FS_SB(sb); 1529 int i; 1530 bool dropped; 1531 1532 /* unregister procfs/sysfs entries in advance to avoid race case */ 1533 f2fs_unregister_sysfs(sbi); 1534 1535 f2fs_quota_off_umount(sb); 1536 1537 /* prevent remaining shrinker jobs */ 1538 mutex_lock(&sbi->umount_mutex); 1539 1540 /* 1541 * flush all issued checkpoints and stop checkpoint issue thread. 1542 * after then, all checkpoints should be done by each process context. 1543 */ 1544 f2fs_stop_ckpt_thread(sbi); 1545 1546 /* 1547 * We don't need to do checkpoint when superblock is clean. 1548 * But, the previous checkpoint was not done by umount, it needs to do 1549 * clean checkpoint again. 1550 */ 1551 if ((is_sbi_flag_set(sbi, SBI_IS_DIRTY) || 1552 !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG))) { 1553 struct cp_control cpc = { 1554 .reason = CP_UMOUNT, 1555 }; 1556 f2fs_write_checkpoint(sbi, &cpc); 1557 } 1558 1559 /* be sure to wait for any on-going discard commands */ 1560 dropped = f2fs_issue_discard_timeout(sbi); 1561 1562 if ((f2fs_hw_support_discard(sbi) || f2fs_hw_should_discard(sbi)) && 1563 !sbi->discard_blks && !dropped) { 1564 struct cp_control cpc = { 1565 .reason = CP_UMOUNT | CP_TRIMMED, 1566 }; 1567 f2fs_write_checkpoint(sbi, &cpc); 1568 } 1569 1570 /* 1571 * normally superblock is clean, so we need to release this. 1572 * In addition, EIO will skip do checkpoint, we need this as well. 1573 */ 1574 f2fs_release_ino_entry(sbi, true); 1575 1576 f2fs_leave_shrinker(sbi); 1577 mutex_unlock(&sbi->umount_mutex); 1578 1579 /* our cp_error case, we can wait for any writeback page */ 1580 f2fs_flush_merged_writes(sbi); 1581 1582 f2fs_wait_on_all_pages(sbi, F2FS_WB_CP_DATA); 1583 1584 f2fs_bug_on(sbi, sbi->fsync_node_num); 1585 1586 f2fs_destroy_compress_inode(sbi); 1587 1588 iput(sbi->node_inode); 1589 sbi->node_inode = NULL; 1590 1591 iput(sbi->meta_inode); 1592 sbi->meta_inode = NULL; 1593 1594 /* 1595 * iput() can update stat information, if f2fs_write_checkpoint() 1596 * above failed with error. 1597 */ 1598 f2fs_destroy_stats(sbi); 1599 1600 /* destroy f2fs internal modules */ 1601 f2fs_destroy_node_manager(sbi); 1602 f2fs_destroy_segment_manager(sbi); 1603 1604 f2fs_destroy_post_read_wq(sbi); 1605 1606 kvfree(sbi->ckpt); 1607 1608 sb->s_fs_info = NULL; 1609 if (sbi->s_chksum_driver) 1610 crypto_free_shash(sbi->s_chksum_driver); 1611 kfree(sbi->raw_super); 1612 1613 destroy_device_list(sbi); 1614 f2fs_destroy_page_array_cache(sbi); 1615 f2fs_destroy_xattr_caches(sbi); 1616 mempool_destroy(sbi->write_io_dummy); 1617 #ifdef CONFIG_QUOTA 1618 for (i = 0; i < MAXQUOTAS; i++) 1619 kfree(F2FS_OPTION(sbi).s_qf_names[i]); 1620 #endif 1621 fscrypt_free_dummy_policy(&F2FS_OPTION(sbi).dummy_enc_policy); 1622 destroy_percpu_info(sbi); 1623 f2fs_destroy_iostat(sbi); 1624 for (i = 0; i < NR_PAGE_TYPE; i++) 1625 kvfree(sbi->write_io[i]); 1626 #if IS_ENABLED(CONFIG_UNICODE) 1627 utf8_unload(sb->s_encoding); 1628 #endif 1629 kfree(sbi); 1630 } 1631 1632 int f2fs_sync_fs(struct super_block *sb, int sync) 1633 { 1634 struct f2fs_sb_info *sbi = F2FS_SB(sb); 1635 int err = 0; 1636 1637 if (unlikely(f2fs_cp_error(sbi))) 1638 return 0; 1639 if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED))) 1640 return 0; 1641 1642 trace_f2fs_sync_fs(sb, sync); 1643 1644 if (unlikely(is_sbi_flag_set(sbi, SBI_POR_DOING))) 1645 return -EAGAIN; 1646 1647 if (sync) 1648 err = f2fs_issue_checkpoint(sbi); 1649 1650 return err; 1651 } 1652 1653 static int f2fs_freeze(struct super_block *sb) 1654 { 1655 if (f2fs_readonly(sb)) 1656 return 0; 1657 1658 /* IO error happened before */ 1659 if (unlikely(f2fs_cp_error(F2FS_SB(sb)))) 1660 return -EIO; 1661 1662 /* must be clean, since sync_filesystem() was already called */ 1663 if (is_sbi_flag_set(F2FS_SB(sb), SBI_IS_DIRTY)) 1664 return -EINVAL; 1665 1666 /* ensure no checkpoint required */ 1667 if (!llist_empty(&F2FS_SB(sb)->cprc_info.issue_list)) 1668 return -EINVAL; 1669 return 0; 1670 } 1671 1672 static int f2fs_unfreeze(struct super_block *sb) 1673 { 1674 return 0; 1675 } 1676 1677 #ifdef CONFIG_QUOTA 1678 static int f2fs_statfs_project(struct super_block *sb, 1679 kprojid_t projid, struct kstatfs *buf) 1680 { 1681 struct kqid qid; 1682 struct dquot *dquot; 1683 u64 limit; 1684 u64 curblock; 1685 1686 qid = make_kqid_projid(projid); 1687 dquot = dqget(sb, qid); 1688 if (IS_ERR(dquot)) 1689 return PTR_ERR(dquot); 1690 spin_lock(&dquot->dq_dqb_lock); 1691 1692 limit = min_not_zero(dquot->dq_dqb.dqb_bsoftlimit, 1693 dquot->dq_dqb.dqb_bhardlimit); 1694 if (limit) 1695 limit >>= sb->s_blocksize_bits; 1696 1697 if (limit && buf->f_blocks > limit) { 1698 curblock = (dquot->dq_dqb.dqb_curspace + 1699 dquot->dq_dqb.dqb_rsvspace) >> sb->s_blocksize_bits; 1700 buf->f_blocks = limit; 1701 buf->f_bfree = buf->f_bavail = 1702 (buf->f_blocks > curblock) ? 1703 (buf->f_blocks - curblock) : 0; 1704 } 1705 1706 limit = min_not_zero(dquot->dq_dqb.dqb_isoftlimit, 1707 dquot->dq_dqb.dqb_ihardlimit); 1708 1709 if (limit && buf->f_files > limit) { 1710 buf->f_files = limit; 1711 buf->f_ffree = 1712 (buf->f_files > dquot->dq_dqb.dqb_curinodes) ? 1713 (buf->f_files - dquot->dq_dqb.dqb_curinodes) : 0; 1714 } 1715 1716 spin_unlock(&dquot->dq_dqb_lock); 1717 dqput(dquot); 1718 return 0; 1719 } 1720 #endif 1721 1722 static int f2fs_statfs(struct dentry *dentry, struct kstatfs *buf) 1723 { 1724 struct super_block *sb = dentry->d_sb; 1725 struct f2fs_sb_info *sbi = F2FS_SB(sb); 1726 u64 id = huge_encode_dev(sb->s_bdev->bd_dev); 1727 block_t total_count, user_block_count, start_count; 1728 u64 avail_node_count; 1729 1730 total_count = le64_to_cpu(sbi->raw_super->block_count); 1731 user_block_count = sbi->user_block_count; 1732 start_count = le32_to_cpu(sbi->raw_super->segment0_blkaddr); 1733 buf->f_type = F2FS_SUPER_MAGIC; 1734 buf->f_bsize = sbi->blocksize; 1735 1736 buf->f_blocks = total_count - start_count; 1737 buf->f_bfree = user_block_count - valid_user_blocks(sbi) - 1738 sbi->current_reserved_blocks; 1739 1740 spin_lock(&sbi->stat_lock); 1741 if (unlikely(buf->f_bfree <= sbi->unusable_block_count)) 1742 buf->f_bfree = 0; 1743 else 1744 buf->f_bfree -= sbi->unusable_block_count; 1745 spin_unlock(&sbi->stat_lock); 1746 1747 if (buf->f_bfree > F2FS_OPTION(sbi).root_reserved_blocks) 1748 buf->f_bavail = buf->f_bfree - 1749 F2FS_OPTION(sbi).root_reserved_blocks; 1750 else 1751 buf->f_bavail = 0; 1752 1753 avail_node_count = sbi->total_node_count - F2FS_RESERVED_NODE_NUM; 1754 1755 if (avail_node_count > user_block_count) { 1756 buf->f_files = user_block_count; 1757 buf->f_ffree = buf->f_bavail; 1758 } else { 1759 buf->f_files = avail_node_count; 1760 buf->f_ffree = min(avail_node_count - valid_node_count(sbi), 1761 buf->f_bavail); 1762 } 1763 1764 buf->f_namelen = F2FS_NAME_LEN; 1765 buf->f_fsid = u64_to_fsid(id); 1766 1767 #ifdef CONFIG_QUOTA 1768 if (is_inode_flag_set(dentry->d_inode, FI_PROJ_INHERIT) && 1769 sb_has_quota_limits_enabled(sb, PRJQUOTA)) { 1770 f2fs_statfs_project(sb, F2FS_I(dentry->d_inode)->i_projid, buf); 1771 } 1772 #endif 1773 return 0; 1774 } 1775 1776 static inline void f2fs_show_quota_options(struct seq_file *seq, 1777 struct super_block *sb) 1778 { 1779 #ifdef CONFIG_QUOTA 1780 struct f2fs_sb_info *sbi = F2FS_SB(sb); 1781 1782 if (F2FS_OPTION(sbi).s_jquota_fmt) { 1783 char *fmtname = ""; 1784 1785 switch (F2FS_OPTION(sbi).s_jquota_fmt) { 1786 case QFMT_VFS_OLD: 1787 fmtname = "vfsold"; 1788 break; 1789 case QFMT_VFS_V0: 1790 fmtname = "vfsv0"; 1791 break; 1792 case QFMT_VFS_V1: 1793 fmtname = "vfsv1"; 1794 break; 1795 } 1796 seq_printf(seq, ",jqfmt=%s", fmtname); 1797 } 1798 1799 if (F2FS_OPTION(sbi).s_qf_names[USRQUOTA]) 1800 seq_show_option(seq, "usrjquota", 1801 F2FS_OPTION(sbi).s_qf_names[USRQUOTA]); 1802 1803 if (F2FS_OPTION(sbi).s_qf_names[GRPQUOTA]) 1804 seq_show_option(seq, "grpjquota", 1805 F2FS_OPTION(sbi).s_qf_names[GRPQUOTA]); 1806 1807 if (F2FS_OPTION(sbi).s_qf_names[PRJQUOTA]) 1808 seq_show_option(seq, "prjjquota", 1809 F2FS_OPTION(sbi).s_qf_names[PRJQUOTA]); 1810 #endif 1811 } 1812 1813 #ifdef CONFIG_F2FS_FS_COMPRESSION 1814 static inline void f2fs_show_compress_options(struct seq_file *seq, 1815 struct super_block *sb) 1816 { 1817 struct f2fs_sb_info *sbi = F2FS_SB(sb); 1818 char *algtype = ""; 1819 int i; 1820 1821 if (!f2fs_sb_has_compression(sbi)) 1822 return; 1823 1824 switch (F2FS_OPTION(sbi).compress_algorithm) { 1825 case COMPRESS_LZO: 1826 algtype = "lzo"; 1827 break; 1828 case COMPRESS_LZ4: 1829 algtype = "lz4"; 1830 break; 1831 case COMPRESS_ZSTD: 1832 algtype = "zstd"; 1833 break; 1834 case COMPRESS_LZORLE: 1835 algtype = "lzo-rle"; 1836 break; 1837 } 1838 seq_printf(seq, ",compress_algorithm=%s", algtype); 1839 1840 if (F2FS_OPTION(sbi).compress_level) 1841 seq_printf(seq, ":%d", F2FS_OPTION(sbi).compress_level); 1842 1843 seq_printf(seq, ",compress_log_size=%u", 1844 F2FS_OPTION(sbi).compress_log_size); 1845 1846 for (i = 0; i < F2FS_OPTION(sbi).compress_ext_cnt; i++) { 1847 seq_printf(seq, ",compress_extension=%s", 1848 F2FS_OPTION(sbi).extensions[i]); 1849 } 1850 1851 for (i = 0; i < F2FS_OPTION(sbi).nocompress_ext_cnt; i++) { 1852 seq_printf(seq, ",nocompress_extension=%s", 1853 F2FS_OPTION(sbi).noextensions[i]); 1854 } 1855 1856 if (F2FS_OPTION(sbi).compress_chksum) 1857 seq_puts(seq, ",compress_chksum"); 1858 1859 if (F2FS_OPTION(sbi).compress_mode == COMPR_MODE_FS) 1860 seq_printf(seq, ",compress_mode=%s", "fs"); 1861 else if (F2FS_OPTION(sbi).compress_mode == COMPR_MODE_USER) 1862 seq_printf(seq, ",compress_mode=%s", "user"); 1863 1864 if (test_opt(sbi, COMPRESS_CACHE)) 1865 seq_puts(seq, ",compress_cache"); 1866 } 1867 #endif 1868 1869 static int f2fs_show_options(struct seq_file *seq, struct dentry *root) 1870 { 1871 struct f2fs_sb_info *sbi = F2FS_SB(root->d_sb); 1872 1873 if (F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_SYNC) 1874 seq_printf(seq, ",background_gc=%s", "sync"); 1875 else if (F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_ON) 1876 seq_printf(seq, ",background_gc=%s", "on"); 1877 else if (F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_OFF) 1878 seq_printf(seq, ",background_gc=%s", "off"); 1879 1880 if (test_opt(sbi, GC_MERGE)) 1881 seq_puts(seq, ",gc_merge"); 1882 1883 if (test_opt(sbi, DISABLE_ROLL_FORWARD)) 1884 seq_puts(seq, ",disable_roll_forward"); 1885 if (test_opt(sbi, NORECOVERY)) 1886 seq_puts(seq, ",norecovery"); 1887 if (test_opt(sbi, DISCARD)) 1888 seq_puts(seq, ",discard"); 1889 else 1890 seq_puts(seq, ",nodiscard"); 1891 if (test_opt(sbi, NOHEAP)) 1892 seq_puts(seq, ",no_heap"); 1893 else 1894 seq_puts(seq, ",heap"); 1895 #ifdef CONFIG_F2FS_FS_XATTR 1896 if (test_opt(sbi, XATTR_USER)) 1897 seq_puts(seq, ",user_xattr"); 1898 else 1899 seq_puts(seq, ",nouser_xattr"); 1900 if (test_opt(sbi, INLINE_XATTR)) 1901 seq_puts(seq, ",inline_xattr"); 1902 else 1903 seq_puts(seq, ",noinline_xattr"); 1904 if (test_opt(sbi, INLINE_XATTR_SIZE)) 1905 seq_printf(seq, ",inline_xattr_size=%u", 1906 F2FS_OPTION(sbi).inline_xattr_size); 1907 #endif 1908 #ifdef CONFIG_F2FS_FS_POSIX_ACL 1909 if (test_opt(sbi, POSIX_ACL)) 1910 seq_puts(seq, ",acl"); 1911 else 1912 seq_puts(seq, ",noacl"); 1913 #endif 1914 if (test_opt(sbi, DISABLE_EXT_IDENTIFY)) 1915 seq_puts(seq, ",disable_ext_identify"); 1916 if (test_opt(sbi, INLINE_DATA)) 1917 seq_puts(seq, ",inline_data"); 1918 else 1919 seq_puts(seq, ",noinline_data"); 1920 if (test_opt(sbi, INLINE_DENTRY)) 1921 seq_puts(seq, ",inline_dentry"); 1922 else 1923 seq_puts(seq, ",noinline_dentry"); 1924 if (!f2fs_readonly(sbi->sb) && test_opt(sbi, FLUSH_MERGE)) 1925 seq_puts(seq, ",flush_merge"); 1926 if (test_opt(sbi, NOBARRIER)) 1927 seq_puts(seq, ",nobarrier"); 1928 if (test_opt(sbi, FASTBOOT)) 1929 seq_puts(seq, ",fastboot"); 1930 if (test_opt(sbi, EXTENT_CACHE)) 1931 seq_puts(seq, ",extent_cache"); 1932 else 1933 seq_puts(seq, ",noextent_cache"); 1934 if (test_opt(sbi, DATA_FLUSH)) 1935 seq_puts(seq, ",data_flush"); 1936 1937 seq_puts(seq, ",mode="); 1938 if (F2FS_OPTION(sbi).fs_mode == FS_MODE_ADAPTIVE) 1939 seq_puts(seq, "adaptive"); 1940 else if (F2FS_OPTION(sbi).fs_mode == FS_MODE_LFS) 1941 seq_puts(seq, "lfs"); 1942 else if (F2FS_OPTION(sbi).fs_mode == FS_MODE_FRAGMENT_SEG) 1943 seq_puts(seq, "fragment:segment"); 1944 else if (F2FS_OPTION(sbi).fs_mode == FS_MODE_FRAGMENT_BLK) 1945 seq_puts(seq, "fragment:block"); 1946 seq_printf(seq, ",active_logs=%u", F2FS_OPTION(sbi).active_logs); 1947 if (test_opt(sbi, RESERVE_ROOT)) 1948 seq_printf(seq, ",reserve_root=%u,resuid=%u,resgid=%u", 1949 F2FS_OPTION(sbi).root_reserved_blocks, 1950 from_kuid_munged(&init_user_ns, 1951 F2FS_OPTION(sbi).s_resuid), 1952 from_kgid_munged(&init_user_ns, 1953 F2FS_OPTION(sbi).s_resgid)); 1954 if (F2FS_IO_SIZE_BITS(sbi)) 1955 seq_printf(seq, ",io_bits=%u", 1956 F2FS_OPTION(sbi).write_io_size_bits); 1957 #ifdef CONFIG_F2FS_FAULT_INJECTION 1958 if (test_opt(sbi, FAULT_INJECTION)) { 1959 seq_printf(seq, ",fault_injection=%u", 1960 F2FS_OPTION(sbi).fault_info.inject_rate); 1961 seq_printf(seq, ",fault_type=%u", 1962 F2FS_OPTION(sbi).fault_info.inject_type); 1963 } 1964 #endif 1965 #ifdef CONFIG_QUOTA 1966 if (test_opt(sbi, QUOTA)) 1967 seq_puts(seq, ",quota"); 1968 if (test_opt(sbi, USRQUOTA)) 1969 seq_puts(seq, ",usrquota"); 1970 if (test_opt(sbi, GRPQUOTA)) 1971 seq_puts(seq, ",grpquota"); 1972 if (test_opt(sbi, PRJQUOTA)) 1973 seq_puts(seq, ",prjquota"); 1974 #endif 1975 f2fs_show_quota_options(seq, sbi->sb); 1976 if (F2FS_OPTION(sbi).whint_mode == WHINT_MODE_USER) 1977 seq_printf(seq, ",whint_mode=%s", "user-based"); 1978 else if (F2FS_OPTION(sbi).whint_mode == WHINT_MODE_FS) 1979 seq_printf(seq, ",whint_mode=%s", "fs-based"); 1980 1981 fscrypt_show_test_dummy_encryption(seq, ',', sbi->sb); 1982 1983 if (sbi->sb->s_flags & SB_INLINECRYPT) 1984 seq_puts(seq, ",inlinecrypt"); 1985 1986 if (F2FS_OPTION(sbi).alloc_mode == ALLOC_MODE_DEFAULT) 1987 seq_printf(seq, ",alloc_mode=%s", "default"); 1988 else if (F2FS_OPTION(sbi).alloc_mode == ALLOC_MODE_REUSE) 1989 seq_printf(seq, ",alloc_mode=%s", "reuse"); 1990 1991 if (test_opt(sbi, DISABLE_CHECKPOINT)) 1992 seq_printf(seq, ",checkpoint=disable:%u", 1993 F2FS_OPTION(sbi).unusable_cap); 1994 if (test_opt(sbi, MERGE_CHECKPOINT)) 1995 seq_puts(seq, ",checkpoint_merge"); 1996 else 1997 seq_puts(seq, ",nocheckpoint_merge"); 1998 if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_POSIX) 1999 seq_printf(seq, ",fsync_mode=%s", "posix"); 2000 else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_STRICT) 2001 seq_printf(seq, ",fsync_mode=%s", "strict"); 2002 else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_NOBARRIER) 2003 seq_printf(seq, ",fsync_mode=%s", "nobarrier"); 2004 2005 #ifdef CONFIG_F2FS_FS_COMPRESSION 2006 f2fs_show_compress_options(seq, sbi->sb); 2007 #endif 2008 2009 if (test_opt(sbi, ATGC)) 2010 seq_puts(seq, ",atgc"); 2011 2012 if (F2FS_OPTION(sbi).discard_unit == DISCARD_UNIT_BLOCK) 2013 seq_printf(seq, ",discard_unit=%s", "block"); 2014 else if (F2FS_OPTION(sbi).discard_unit == DISCARD_UNIT_SEGMENT) 2015 seq_printf(seq, ",discard_unit=%s", "segment"); 2016 else if (F2FS_OPTION(sbi).discard_unit == DISCARD_UNIT_SECTION) 2017 seq_printf(seq, ",discard_unit=%s", "section"); 2018 2019 return 0; 2020 } 2021 2022 static void default_options(struct f2fs_sb_info *sbi) 2023 { 2024 /* init some FS parameters */ 2025 if (f2fs_sb_has_readonly(sbi)) 2026 F2FS_OPTION(sbi).active_logs = NR_CURSEG_RO_TYPE; 2027 else 2028 F2FS_OPTION(sbi).active_logs = NR_CURSEG_PERSIST_TYPE; 2029 2030 F2FS_OPTION(sbi).inline_xattr_size = DEFAULT_INLINE_XATTR_ADDRS; 2031 F2FS_OPTION(sbi).whint_mode = WHINT_MODE_OFF; 2032 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_DEFAULT; 2033 F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_POSIX; 2034 F2FS_OPTION(sbi).s_resuid = make_kuid(&init_user_ns, F2FS_DEF_RESUID); 2035 F2FS_OPTION(sbi).s_resgid = make_kgid(&init_user_ns, F2FS_DEF_RESGID); 2036 F2FS_OPTION(sbi).compress_algorithm = COMPRESS_LZ4; 2037 F2FS_OPTION(sbi).compress_log_size = MIN_COMPRESS_LOG_SIZE; 2038 F2FS_OPTION(sbi).compress_ext_cnt = 0; 2039 F2FS_OPTION(sbi).compress_mode = COMPR_MODE_FS; 2040 F2FS_OPTION(sbi).bggc_mode = BGGC_MODE_ON; 2041 2042 sbi->sb->s_flags &= ~SB_INLINECRYPT; 2043 2044 set_opt(sbi, INLINE_XATTR); 2045 set_opt(sbi, INLINE_DATA); 2046 set_opt(sbi, INLINE_DENTRY); 2047 set_opt(sbi, EXTENT_CACHE); 2048 set_opt(sbi, NOHEAP); 2049 clear_opt(sbi, DISABLE_CHECKPOINT); 2050 set_opt(sbi, MERGE_CHECKPOINT); 2051 F2FS_OPTION(sbi).unusable_cap = 0; 2052 sbi->sb->s_flags |= SB_LAZYTIME; 2053 set_opt(sbi, FLUSH_MERGE); 2054 if (f2fs_hw_support_discard(sbi) || f2fs_hw_should_discard(sbi)) 2055 set_opt(sbi, DISCARD); 2056 if (f2fs_sb_has_blkzoned(sbi)) { 2057 F2FS_OPTION(sbi).fs_mode = FS_MODE_LFS; 2058 F2FS_OPTION(sbi).discard_unit = DISCARD_UNIT_SECTION; 2059 } else { 2060 F2FS_OPTION(sbi).fs_mode = FS_MODE_ADAPTIVE; 2061 F2FS_OPTION(sbi).discard_unit = DISCARD_UNIT_BLOCK; 2062 } 2063 2064 #ifdef CONFIG_F2FS_FS_XATTR 2065 set_opt(sbi, XATTR_USER); 2066 #endif 2067 #ifdef CONFIG_F2FS_FS_POSIX_ACL 2068 set_opt(sbi, POSIX_ACL); 2069 #endif 2070 2071 f2fs_build_fault_attr(sbi, 0, 0); 2072 } 2073 2074 #ifdef CONFIG_QUOTA 2075 static int f2fs_enable_quotas(struct super_block *sb); 2076 #endif 2077 2078 static int f2fs_disable_checkpoint(struct f2fs_sb_info *sbi) 2079 { 2080 unsigned int s_flags = sbi->sb->s_flags; 2081 struct cp_control cpc; 2082 int err = 0; 2083 int ret; 2084 block_t unusable; 2085 2086 if (s_flags & SB_RDONLY) { 2087 f2fs_err(sbi, "checkpoint=disable on readonly fs"); 2088 return -EINVAL; 2089 } 2090 sbi->sb->s_flags |= SB_ACTIVE; 2091 2092 f2fs_update_time(sbi, DISABLE_TIME); 2093 2094 while (!f2fs_time_over(sbi, DISABLE_TIME)) { 2095 down_write(&sbi->gc_lock); 2096 err = f2fs_gc(sbi, true, false, false, NULL_SEGNO); 2097 if (err == -ENODATA) { 2098 err = 0; 2099 break; 2100 } 2101 if (err && err != -EAGAIN) 2102 break; 2103 } 2104 2105 ret = sync_filesystem(sbi->sb); 2106 if (ret || err) { 2107 err = ret ? ret : err; 2108 goto restore_flag; 2109 } 2110 2111 unusable = f2fs_get_unusable_blocks(sbi); 2112 if (f2fs_disable_cp_again(sbi, unusable)) { 2113 err = -EAGAIN; 2114 goto restore_flag; 2115 } 2116 2117 down_write(&sbi->gc_lock); 2118 cpc.reason = CP_PAUSE; 2119 set_sbi_flag(sbi, SBI_CP_DISABLED); 2120 err = f2fs_write_checkpoint(sbi, &cpc); 2121 if (err) 2122 goto out_unlock; 2123 2124 spin_lock(&sbi->stat_lock); 2125 sbi->unusable_block_count = unusable; 2126 spin_unlock(&sbi->stat_lock); 2127 2128 out_unlock: 2129 up_write(&sbi->gc_lock); 2130 restore_flag: 2131 sbi->sb->s_flags = s_flags; /* Restore SB_RDONLY status */ 2132 return err; 2133 } 2134 2135 static void f2fs_enable_checkpoint(struct f2fs_sb_info *sbi) 2136 { 2137 int retry = DEFAULT_RETRY_IO_COUNT; 2138 2139 /* we should flush all the data to keep data consistency */ 2140 do { 2141 sync_inodes_sb(sbi->sb); 2142 f2fs_io_schedule_timeout(DEFAULT_IO_TIMEOUT); 2143 } while (get_pages(sbi, F2FS_DIRTY_DATA) && retry--); 2144 2145 if (unlikely(retry < 0)) 2146 f2fs_warn(sbi, "checkpoint=enable has some unwritten data."); 2147 2148 down_write(&sbi->gc_lock); 2149 f2fs_dirty_to_prefree(sbi); 2150 2151 clear_sbi_flag(sbi, SBI_CP_DISABLED); 2152 set_sbi_flag(sbi, SBI_IS_DIRTY); 2153 up_write(&sbi->gc_lock); 2154 2155 f2fs_sync_fs(sbi->sb, 1); 2156 } 2157 2158 static int f2fs_remount(struct super_block *sb, int *flags, char *data) 2159 { 2160 struct f2fs_sb_info *sbi = F2FS_SB(sb); 2161 struct f2fs_mount_info org_mount_opt; 2162 unsigned long old_sb_flags; 2163 int err; 2164 bool need_restart_gc = false, need_stop_gc = false; 2165 bool need_restart_ckpt = false, need_stop_ckpt = false; 2166 bool need_restart_flush = false, need_stop_flush = false; 2167 bool need_restart_discard = false, need_stop_discard = false; 2168 bool no_extent_cache = !test_opt(sbi, EXTENT_CACHE); 2169 bool enable_checkpoint = !test_opt(sbi, DISABLE_CHECKPOINT); 2170 bool no_io_align = !F2FS_IO_ALIGNED(sbi); 2171 bool no_atgc = !test_opt(sbi, ATGC); 2172 bool no_discard = !test_opt(sbi, DISCARD); 2173 bool no_compress_cache = !test_opt(sbi, COMPRESS_CACHE); 2174 bool block_unit_discard = f2fs_block_unit_discard(sbi); 2175 struct discard_cmd_control *dcc; 2176 #ifdef CONFIG_QUOTA 2177 int i, j; 2178 #endif 2179 2180 /* 2181 * Save the old mount options in case we 2182 * need to restore them. 2183 */ 2184 org_mount_opt = sbi->mount_opt; 2185 old_sb_flags = sb->s_flags; 2186 2187 #ifdef CONFIG_QUOTA 2188 org_mount_opt.s_jquota_fmt = F2FS_OPTION(sbi).s_jquota_fmt; 2189 for (i = 0; i < MAXQUOTAS; i++) { 2190 if (F2FS_OPTION(sbi).s_qf_names[i]) { 2191 org_mount_opt.s_qf_names[i] = 2192 kstrdup(F2FS_OPTION(sbi).s_qf_names[i], 2193 GFP_KERNEL); 2194 if (!org_mount_opt.s_qf_names[i]) { 2195 for (j = 0; j < i; j++) 2196 kfree(org_mount_opt.s_qf_names[j]); 2197 return -ENOMEM; 2198 } 2199 } else { 2200 org_mount_opt.s_qf_names[i] = NULL; 2201 } 2202 } 2203 #endif 2204 2205 /* recover superblocks we couldn't write due to previous RO mount */ 2206 if (!(*flags & SB_RDONLY) && is_sbi_flag_set(sbi, SBI_NEED_SB_WRITE)) { 2207 err = f2fs_commit_super(sbi, false); 2208 f2fs_info(sbi, "Try to recover all the superblocks, ret: %d", 2209 err); 2210 if (!err) 2211 clear_sbi_flag(sbi, SBI_NEED_SB_WRITE); 2212 } 2213 2214 default_options(sbi); 2215 2216 /* parse mount options */ 2217 err = parse_options(sb, data, true); 2218 if (err) 2219 goto restore_opts; 2220 2221 /* 2222 * Previous and new state of filesystem is RO, 2223 * so skip checking GC and FLUSH_MERGE conditions. 2224 */ 2225 if (f2fs_readonly(sb) && (*flags & SB_RDONLY)) 2226 goto skip; 2227 2228 if (f2fs_sb_has_readonly(sbi) && !(*flags & SB_RDONLY)) { 2229 err = -EROFS; 2230 goto restore_opts; 2231 } 2232 2233 #ifdef CONFIG_QUOTA 2234 if (!f2fs_readonly(sb) && (*flags & SB_RDONLY)) { 2235 err = dquot_suspend(sb, -1); 2236 if (err < 0) 2237 goto restore_opts; 2238 } else if (f2fs_readonly(sb) && !(*flags & SB_RDONLY)) { 2239 /* dquot_resume needs RW */ 2240 sb->s_flags &= ~SB_RDONLY; 2241 if (sb_any_quota_suspended(sb)) { 2242 dquot_resume(sb, -1); 2243 } else if (f2fs_sb_has_quota_ino(sbi)) { 2244 err = f2fs_enable_quotas(sb); 2245 if (err) 2246 goto restore_opts; 2247 } 2248 } 2249 #endif 2250 /* disallow enable atgc dynamically */ 2251 if (no_atgc == !!test_opt(sbi, ATGC)) { 2252 err = -EINVAL; 2253 f2fs_warn(sbi, "switch atgc option is not allowed"); 2254 goto restore_opts; 2255 } 2256 2257 /* disallow enable/disable extent_cache dynamically */ 2258 if (no_extent_cache == !!test_opt(sbi, EXTENT_CACHE)) { 2259 err = -EINVAL; 2260 f2fs_warn(sbi, "switch extent_cache option is not allowed"); 2261 goto restore_opts; 2262 } 2263 2264 if (no_io_align == !!F2FS_IO_ALIGNED(sbi)) { 2265 err = -EINVAL; 2266 f2fs_warn(sbi, "switch io_bits option is not allowed"); 2267 goto restore_opts; 2268 } 2269 2270 if (no_compress_cache == !!test_opt(sbi, COMPRESS_CACHE)) { 2271 err = -EINVAL; 2272 f2fs_warn(sbi, "switch compress_cache option is not allowed"); 2273 goto restore_opts; 2274 } 2275 2276 if (block_unit_discard != f2fs_block_unit_discard(sbi)) { 2277 err = -EINVAL; 2278 f2fs_warn(sbi, "switch discard_unit option is not allowed"); 2279 goto restore_opts; 2280 } 2281 2282 if ((*flags & SB_RDONLY) && test_opt(sbi, DISABLE_CHECKPOINT)) { 2283 err = -EINVAL; 2284 f2fs_warn(sbi, "disabling checkpoint not compatible with read-only"); 2285 goto restore_opts; 2286 } 2287 2288 /* 2289 * We stop the GC thread if FS is mounted as RO 2290 * or if background_gc = off is passed in mount 2291 * option. Also sync the filesystem. 2292 */ 2293 if ((*flags & SB_RDONLY) || 2294 (F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_OFF && 2295 !test_opt(sbi, GC_MERGE))) { 2296 if (sbi->gc_thread) { 2297 f2fs_stop_gc_thread(sbi); 2298 need_restart_gc = true; 2299 } 2300 } else if (!sbi->gc_thread) { 2301 err = f2fs_start_gc_thread(sbi); 2302 if (err) 2303 goto restore_opts; 2304 need_stop_gc = true; 2305 } 2306 2307 if (*flags & SB_RDONLY || 2308 F2FS_OPTION(sbi).whint_mode != org_mount_opt.whint_mode) { 2309 sync_inodes_sb(sb); 2310 2311 set_sbi_flag(sbi, SBI_IS_DIRTY); 2312 set_sbi_flag(sbi, SBI_IS_CLOSE); 2313 f2fs_sync_fs(sb, 1); 2314 clear_sbi_flag(sbi, SBI_IS_CLOSE); 2315 } 2316 2317 if ((*flags & SB_RDONLY) || test_opt(sbi, DISABLE_CHECKPOINT) || 2318 !test_opt(sbi, MERGE_CHECKPOINT)) { 2319 f2fs_stop_ckpt_thread(sbi); 2320 need_restart_ckpt = true; 2321 } else { 2322 err = f2fs_start_ckpt_thread(sbi); 2323 if (err) { 2324 f2fs_err(sbi, 2325 "Failed to start F2FS issue_checkpoint_thread (%d)", 2326 err); 2327 goto restore_gc; 2328 } 2329 need_stop_ckpt = true; 2330 } 2331 2332 /* 2333 * We stop issue flush thread if FS is mounted as RO 2334 * or if flush_merge is not passed in mount option. 2335 */ 2336 if ((*flags & SB_RDONLY) || !test_opt(sbi, FLUSH_MERGE)) { 2337 clear_opt(sbi, FLUSH_MERGE); 2338 f2fs_destroy_flush_cmd_control(sbi, false); 2339 need_restart_flush = true; 2340 } else { 2341 err = f2fs_create_flush_cmd_control(sbi); 2342 if (err) 2343 goto restore_ckpt; 2344 need_stop_flush = true; 2345 } 2346 2347 if (no_discard == !!test_opt(sbi, DISCARD)) { 2348 if (test_opt(sbi, DISCARD)) { 2349 err = f2fs_start_discard_thread(sbi); 2350 if (err) 2351 goto restore_flush; 2352 need_stop_discard = true; 2353 } else { 2354 dcc = SM_I(sbi)->dcc_info; 2355 f2fs_stop_discard_thread(sbi); 2356 if (atomic_read(&dcc->discard_cmd_cnt)) 2357 f2fs_issue_discard_timeout(sbi); 2358 need_restart_discard = true; 2359 } 2360 } 2361 2362 if (enable_checkpoint == !!test_opt(sbi, DISABLE_CHECKPOINT)) { 2363 if (test_opt(sbi, DISABLE_CHECKPOINT)) { 2364 err = f2fs_disable_checkpoint(sbi); 2365 if (err) 2366 goto restore_discard; 2367 } else { 2368 f2fs_enable_checkpoint(sbi); 2369 } 2370 } 2371 2372 skip: 2373 #ifdef CONFIG_QUOTA 2374 /* Release old quota file names */ 2375 for (i = 0; i < MAXQUOTAS; i++) 2376 kfree(org_mount_opt.s_qf_names[i]); 2377 #endif 2378 /* Update the POSIXACL Flag */ 2379 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) | 2380 (test_opt(sbi, POSIX_ACL) ? SB_POSIXACL : 0); 2381 2382 limit_reserve_root(sbi); 2383 adjust_unusable_cap_perc(sbi); 2384 *flags = (*flags & ~SB_LAZYTIME) | (sb->s_flags & SB_LAZYTIME); 2385 return 0; 2386 restore_discard: 2387 if (need_restart_discard) { 2388 if (f2fs_start_discard_thread(sbi)) 2389 f2fs_warn(sbi, "discard has been stopped"); 2390 } else if (need_stop_discard) { 2391 f2fs_stop_discard_thread(sbi); 2392 } 2393 restore_flush: 2394 if (need_restart_flush) { 2395 if (f2fs_create_flush_cmd_control(sbi)) 2396 f2fs_warn(sbi, "background flush thread has stopped"); 2397 } else if (need_stop_flush) { 2398 clear_opt(sbi, FLUSH_MERGE); 2399 f2fs_destroy_flush_cmd_control(sbi, false); 2400 } 2401 restore_ckpt: 2402 if (need_restart_ckpt) { 2403 if (f2fs_start_ckpt_thread(sbi)) 2404 f2fs_warn(sbi, "background ckpt thread has stopped"); 2405 } else if (need_stop_ckpt) { 2406 f2fs_stop_ckpt_thread(sbi); 2407 } 2408 restore_gc: 2409 if (need_restart_gc) { 2410 if (f2fs_start_gc_thread(sbi)) 2411 f2fs_warn(sbi, "background gc thread has stopped"); 2412 } else if (need_stop_gc) { 2413 f2fs_stop_gc_thread(sbi); 2414 } 2415 restore_opts: 2416 #ifdef CONFIG_QUOTA 2417 F2FS_OPTION(sbi).s_jquota_fmt = org_mount_opt.s_jquota_fmt; 2418 for (i = 0; i < MAXQUOTAS; i++) { 2419 kfree(F2FS_OPTION(sbi).s_qf_names[i]); 2420 F2FS_OPTION(sbi).s_qf_names[i] = org_mount_opt.s_qf_names[i]; 2421 } 2422 #endif 2423 sbi->mount_opt = org_mount_opt; 2424 sb->s_flags = old_sb_flags; 2425 return err; 2426 } 2427 2428 #ifdef CONFIG_QUOTA 2429 /* Read data from quotafile */ 2430 static ssize_t f2fs_quota_read(struct super_block *sb, int type, char *data, 2431 size_t len, loff_t off) 2432 { 2433 struct inode *inode = sb_dqopt(sb)->files[type]; 2434 struct address_space *mapping = inode->i_mapping; 2435 block_t blkidx = F2FS_BYTES_TO_BLK(off); 2436 int offset = off & (sb->s_blocksize - 1); 2437 int tocopy; 2438 size_t toread; 2439 loff_t i_size = i_size_read(inode); 2440 struct page *page; 2441 char *kaddr; 2442 2443 if (off > i_size) 2444 return 0; 2445 2446 if (off + len > i_size) 2447 len = i_size - off; 2448 toread = len; 2449 while (toread > 0) { 2450 tocopy = min_t(unsigned long, sb->s_blocksize - offset, toread); 2451 repeat: 2452 page = read_cache_page_gfp(mapping, blkidx, GFP_NOFS); 2453 if (IS_ERR(page)) { 2454 if (PTR_ERR(page) == -ENOMEM) { 2455 memalloc_retry_wait(GFP_NOFS); 2456 goto repeat; 2457 } 2458 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR); 2459 return PTR_ERR(page); 2460 } 2461 2462 lock_page(page); 2463 2464 if (unlikely(page->mapping != mapping)) { 2465 f2fs_put_page(page, 1); 2466 goto repeat; 2467 } 2468 if (unlikely(!PageUptodate(page))) { 2469 f2fs_put_page(page, 1); 2470 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR); 2471 return -EIO; 2472 } 2473 2474 kaddr = kmap_atomic(page); 2475 memcpy(data, kaddr + offset, tocopy); 2476 kunmap_atomic(kaddr); 2477 f2fs_put_page(page, 1); 2478 2479 offset = 0; 2480 toread -= tocopy; 2481 data += tocopy; 2482 blkidx++; 2483 } 2484 return len; 2485 } 2486 2487 /* Write to quotafile */ 2488 static ssize_t f2fs_quota_write(struct super_block *sb, int type, 2489 const char *data, size_t len, loff_t off) 2490 { 2491 struct inode *inode = sb_dqopt(sb)->files[type]; 2492 struct address_space *mapping = inode->i_mapping; 2493 const struct address_space_operations *a_ops = mapping->a_ops; 2494 int offset = off & (sb->s_blocksize - 1); 2495 size_t towrite = len; 2496 struct page *page; 2497 void *fsdata = NULL; 2498 char *kaddr; 2499 int err = 0; 2500 int tocopy; 2501 2502 while (towrite > 0) { 2503 tocopy = min_t(unsigned long, sb->s_blocksize - offset, 2504 towrite); 2505 retry: 2506 err = a_ops->write_begin(NULL, mapping, off, tocopy, 0, 2507 &page, &fsdata); 2508 if (unlikely(err)) { 2509 if (err == -ENOMEM) { 2510 f2fs_io_schedule_timeout(DEFAULT_IO_TIMEOUT); 2511 goto retry; 2512 } 2513 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR); 2514 break; 2515 } 2516 2517 kaddr = kmap_atomic(page); 2518 memcpy(kaddr + offset, data, tocopy); 2519 kunmap_atomic(kaddr); 2520 flush_dcache_page(page); 2521 2522 a_ops->write_end(NULL, mapping, off, tocopy, tocopy, 2523 page, fsdata); 2524 offset = 0; 2525 towrite -= tocopy; 2526 off += tocopy; 2527 data += tocopy; 2528 cond_resched(); 2529 } 2530 2531 if (len == towrite) 2532 return err; 2533 inode->i_mtime = inode->i_ctime = current_time(inode); 2534 f2fs_mark_inode_dirty_sync(inode, false); 2535 return len - towrite; 2536 } 2537 2538 int f2fs_dquot_initialize(struct inode *inode) 2539 { 2540 if (time_to_inject(F2FS_I_SB(inode), FAULT_DQUOT_INIT)) { 2541 f2fs_show_injection_info(F2FS_I_SB(inode), FAULT_DQUOT_INIT); 2542 return -ESRCH; 2543 } 2544 2545 return dquot_initialize(inode); 2546 } 2547 2548 static struct dquot **f2fs_get_dquots(struct inode *inode) 2549 { 2550 return F2FS_I(inode)->i_dquot; 2551 } 2552 2553 static qsize_t *f2fs_get_reserved_space(struct inode *inode) 2554 { 2555 return &F2FS_I(inode)->i_reserved_quota; 2556 } 2557 2558 static int f2fs_quota_on_mount(struct f2fs_sb_info *sbi, int type) 2559 { 2560 if (is_set_ckpt_flags(sbi, CP_QUOTA_NEED_FSCK_FLAG)) { 2561 f2fs_err(sbi, "quota sysfile may be corrupted, skip loading it"); 2562 return 0; 2563 } 2564 2565 return dquot_quota_on_mount(sbi->sb, F2FS_OPTION(sbi).s_qf_names[type], 2566 F2FS_OPTION(sbi).s_jquota_fmt, type); 2567 } 2568 2569 int f2fs_enable_quota_files(struct f2fs_sb_info *sbi, bool rdonly) 2570 { 2571 int enabled = 0; 2572 int i, err; 2573 2574 if (f2fs_sb_has_quota_ino(sbi) && rdonly) { 2575 err = f2fs_enable_quotas(sbi->sb); 2576 if (err) { 2577 f2fs_err(sbi, "Cannot turn on quota_ino: %d", err); 2578 return 0; 2579 } 2580 return 1; 2581 } 2582 2583 for (i = 0; i < MAXQUOTAS; i++) { 2584 if (F2FS_OPTION(sbi).s_qf_names[i]) { 2585 err = f2fs_quota_on_mount(sbi, i); 2586 if (!err) { 2587 enabled = 1; 2588 continue; 2589 } 2590 f2fs_err(sbi, "Cannot turn on quotas: %d on %d", 2591 err, i); 2592 } 2593 } 2594 return enabled; 2595 } 2596 2597 static int f2fs_quota_enable(struct super_block *sb, int type, int format_id, 2598 unsigned int flags) 2599 { 2600 struct inode *qf_inode; 2601 unsigned long qf_inum; 2602 int err; 2603 2604 BUG_ON(!f2fs_sb_has_quota_ino(F2FS_SB(sb))); 2605 2606 qf_inum = f2fs_qf_ino(sb, type); 2607 if (!qf_inum) 2608 return -EPERM; 2609 2610 qf_inode = f2fs_iget(sb, qf_inum); 2611 if (IS_ERR(qf_inode)) { 2612 f2fs_err(F2FS_SB(sb), "Bad quota inode %u:%lu", type, qf_inum); 2613 return PTR_ERR(qf_inode); 2614 } 2615 2616 /* Don't account quota for quota files to avoid recursion */ 2617 qf_inode->i_flags |= S_NOQUOTA; 2618 err = dquot_load_quota_inode(qf_inode, type, format_id, flags); 2619 iput(qf_inode); 2620 return err; 2621 } 2622 2623 static int f2fs_enable_quotas(struct super_block *sb) 2624 { 2625 struct f2fs_sb_info *sbi = F2FS_SB(sb); 2626 int type, err = 0; 2627 unsigned long qf_inum; 2628 bool quota_mopt[MAXQUOTAS] = { 2629 test_opt(sbi, USRQUOTA), 2630 test_opt(sbi, GRPQUOTA), 2631 test_opt(sbi, PRJQUOTA), 2632 }; 2633 2634 if (is_set_ckpt_flags(F2FS_SB(sb), CP_QUOTA_NEED_FSCK_FLAG)) { 2635 f2fs_err(sbi, "quota file may be corrupted, skip loading it"); 2636 return 0; 2637 } 2638 2639 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE; 2640 2641 for (type = 0; type < MAXQUOTAS; type++) { 2642 qf_inum = f2fs_qf_ino(sb, type); 2643 if (qf_inum) { 2644 err = f2fs_quota_enable(sb, type, QFMT_VFS_V1, 2645 DQUOT_USAGE_ENABLED | 2646 (quota_mopt[type] ? DQUOT_LIMITS_ENABLED : 0)); 2647 if (err) { 2648 f2fs_err(sbi, "Failed to enable quota tracking (type=%d, err=%d). Please run fsck to fix.", 2649 type, err); 2650 for (type--; type >= 0; type--) 2651 dquot_quota_off(sb, type); 2652 set_sbi_flag(F2FS_SB(sb), 2653 SBI_QUOTA_NEED_REPAIR); 2654 return err; 2655 } 2656 } 2657 } 2658 return 0; 2659 } 2660 2661 static int f2fs_quota_sync_file(struct f2fs_sb_info *sbi, int type) 2662 { 2663 struct quota_info *dqopt = sb_dqopt(sbi->sb); 2664 struct address_space *mapping = dqopt->files[type]->i_mapping; 2665 int ret = 0; 2666 2667 ret = dquot_writeback_dquots(sbi->sb, type); 2668 if (ret) 2669 goto out; 2670 2671 ret = filemap_fdatawrite(mapping); 2672 if (ret) 2673 goto out; 2674 2675 /* if we are using journalled quota */ 2676 if (is_journalled_quota(sbi)) 2677 goto out; 2678 2679 ret = filemap_fdatawait(mapping); 2680 2681 truncate_inode_pages(&dqopt->files[type]->i_data, 0); 2682 out: 2683 if (ret) 2684 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR); 2685 return ret; 2686 } 2687 2688 int f2fs_quota_sync(struct super_block *sb, int type) 2689 { 2690 struct f2fs_sb_info *sbi = F2FS_SB(sb); 2691 struct quota_info *dqopt = sb_dqopt(sb); 2692 int cnt; 2693 int ret; 2694 2695 /* 2696 * Now when everything is written we can discard the pagecache so 2697 * that userspace sees the changes. 2698 */ 2699 for (cnt = 0; cnt < MAXQUOTAS; cnt++) { 2700 2701 if (type != -1 && cnt != type) 2702 continue; 2703 2704 if (!sb_has_quota_active(sb, type)) 2705 return 0; 2706 2707 inode_lock(dqopt->files[cnt]); 2708 2709 /* 2710 * do_quotactl 2711 * f2fs_quota_sync 2712 * down_read(quota_sem) 2713 * dquot_writeback_dquots() 2714 * f2fs_dquot_commit 2715 * block_operation 2716 * down_read(quota_sem) 2717 */ 2718 f2fs_lock_op(sbi); 2719 down_read(&sbi->quota_sem); 2720 2721 ret = f2fs_quota_sync_file(sbi, cnt); 2722 2723 up_read(&sbi->quota_sem); 2724 f2fs_unlock_op(sbi); 2725 2726 inode_unlock(dqopt->files[cnt]); 2727 2728 if (ret) 2729 break; 2730 } 2731 return ret; 2732 } 2733 2734 static int f2fs_quota_on(struct super_block *sb, int type, int format_id, 2735 const struct path *path) 2736 { 2737 struct inode *inode; 2738 int err; 2739 2740 /* if quota sysfile exists, deny enabling quota with specific file */ 2741 if (f2fs_sb_has_quota_ino(F2FS_SB(sb))) { 2742 f2fs_err(F2FS_SB(sb), "quota sysfile already exists"); 2743 return -EBUSY; 2744 } 2745 2746 err = f2fs_quota_sync(sb, type); 2747 if (err) 2748 return err; 2749 2750 err = dquot_quota_on(sb, type, format_id, path); 2751 if (err) 2752 return err; 2753 2754 inode = d_inode(path->dentry); 2755 2756 inode_lock(inode); 2757 F2FS_I(inode)->i_flags |= F2FS_NOATIME_FL | F2FS_IMMUTABLE_FL; 2758 f2fs_set_inode_flags(inode); 2759 inode_unlock(inode); 2760 f2fs_mark_inode_dirty_sync(inode, false); 2761 2762 return 0; 2763 } 2764 2765 static int __f2fs_quota_off(struct super_block *sb, int type) 2766 { 2767 struct inode *inode = sb_dqopt(sb)->files[type]; 2768 int err; 2769 2770 if (!inode || !igrab(inode)) 2771 return dquot_quota_off(sb, type); 2772 2773 err = f2fs_quota_sync(sb, type); 2774 if (err) 2775 goto out_put; 2776 2777 err = dquot_quota_off(sb, type); 2778 if (err || f2fs_sb_has_quota_ino(F2FS_SB(sb))) 2779 goto out_put; 2780 2781 inode_lock(inode); 2782 F2FS_I(inode)->i_flags &= ~(F2FS_NOATIME_FL | F2FS_IMMUTABLE_FL); 2783 f2fs_set_inode_flags(inode); 2784 inode_unlock(inode); 2785 f2fs_mark_inode_dirty_sync(inode, false); 2786 out_put: 2787 iput(inode); 2788 return err; 2789 } 2790 2791 static int f2fs_quota_off(struct super_block *sb, int type) 2792 { 2793 struct f2fs_sb_info *sbi = F2FS_SB(sb); 2794 int err; 2795 2796 err = __f2fs_quota_off(sb, type); 2797 2798 /* 2799 * quotactl can shutdown journalled quota, result in inconsistence 2800 * between quota record and fs data by following updates, tag the 2801 * flag to let fsck be aware of it. 2802 */ 2803 if (is_journalled_quota(sbi)) 2804 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR); 2805 return err; 2806 } 2807 2808 void f2fs_quota_off_umount(struct super_block *sb) 2809 { 2810 int type; 2811 int err; 2812 2813 for (type = 0; type < MAXQUOTAS; type++) { 2814 err = __f2fs_quota_off(sb, type); 2815 if (err) { 2816 int ret = dquot_quota_off(sb, type); 2817 2818 f2fs_err(F2FS_SB(sb), "Fail to turn off disk quota (type: %d, err: %d, ret:%d), Please run fsck to fix it.", 2819 type, err, ret); 2820 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR); 2821 } 2822 } 2823 /* 2824 * In case of checkpoint=disable, we must flush quota blocks. 2825 * This can cause NULL exception for node_inode in end_io, since 2826 * put_super already dropped it. 2827 */ 2828 sync_filesystem(sb); 2829 } 2830 2831 static void f2fs_truncate_quota_inode_pages(struct super_block *sb) 2832 { 2833 struct quota_info *dqopt = sb_dqopt(sb); 2834 int type; 2835 2836 for (type = 0; type < MAXQUOTAS; type++) { 2837 if (!dqopt->files[type]) 2838 continue; 2839 f2fs_inode_synced(dqopt->files[type]); 2840 } 2841 } 2842 2843 static int f2fs_dquot_commit(struct dquot *dquot) 2844 { 2845 struct f2fs_sb_info *sbi = F2FS_SB(dquot->dq_sb); 2846 int ret; 2847 2848 down_read_nested(&sbi->quota_sem, SINGLE_DEPTH_NESTING); 2849 ret = dquot_commit(dquot); 2850 if (ret < 0) 2851 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR); 2852 up_read(&sbi->quota_sem); 2853 return ret; 2854 } 2855 2856 static int f2fs_dquot_acquire(struct dquot *dquot) 2857 { 2858 struct f2fs_sb_info *sbi = F2FS_SB(dquot->dq_sb); 2859 int ret; 2860 2861 down_read(&sbi->quota_sem); 2862 ret = dquot_acquire(dquot); 2863 if (ret < 0) 2864 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR); 2865 up_read(&sbi->quota_sem); 2866 return ret; 2867 } 2868 2869 static int f2fs_dquot_release(struct dquot *dquot) 2870 { 2871 struct f2fs_sb_info *sbi = F2FS_SB(dquot->dq_sb); 2872 int ret = dquot_release(dquot); 2873 2874 if (ret < 0) 2875 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR); 2876 return ret; 2877 } 2878 2879 static int f2fs_dquot_mark_dquot_dirty(struct dquot *dquot) 2880 { 2881 struct super_block *sb = dquot->dq_sb; 2882 struct f2fs_sb_info *sbi = F2FS_SB(sb); 2883 int ret = dquot_mark_dquot_dirty(dquot); 2884 2885 /* if we are using journalled quota */ 2886 if (is_journalled_quota(sbi)) 2887 set_sbi_flag(sbi, SBI_QUOTA_NEED_FLUSH); 2888 2889 return ret; 2890 } 2891 2892 static int f2fs_dquot_commit_info(struct super_block *sb, int type) 2893 { 2894 struct f2fs_sb_info *sbi = F2FS_SB(sb); 2895 int ret = dquot_commit_info(sb, type); 2896 2897 if (ret < 0) 2898 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR); 2899 return ret; 2900 } 2901 2902 static int f2fs_get_projid(struct inode *inode, kprojid_t *projid) 2903 { 2904 *projid = F2FS_I(inode)->i_projid; 2905 return 0; 2906 } 2907 2908 static const struct dquot_operations f2fs_quota_operations = { 2909 .get_reserved_space = f2fs_get_reserved_space, 2910 .write_dquot = f2fs_dquot_commit, 2911 .acquire_dquot = f2fs_dquot_acquire, 2912 .release_dquot = f2fs_dquot_release, 2913 .mark_dirty = f2fs_dquot_mark_dquot_dirty, 2914 .write_info = f2fs_dquot_commit_info, 2915 .alloc_dquot = dquot_alloc, 2916 .destroy_dquot = dquot_destroy, 2917 .get_projid = f2fs_get_projid, 2918 .get_next_id = dquot_get_next_id, 2919 }; 2920 2921 static const struct quotactl_ops f2fs_quotactl_ops = { 2922 .quota_on = f2fs_quota_on, 2923 .quota_off = f2fs_quota_off, 2924 .quota_sync = f2fs_quota_sync, 2925 .get_state = dquot_get_state, 2926 .set_info = dquot_set_dqinfo, 2927 .get_dqblk = dquot_get_dqblk, 2928 .set_dqblk = dquot_set_dqblk, 2929 .get_nextdqblk = dquot_get_next_dqblk, 2930 }; 2931 #else 2932 int f2fs_dquot_initialize(struct inode *inode) 2933 { 2934 return 0; 2935 } 2936 2937 int f2fs_quota_sync(struct super_block *sb, int type) 2938 { 2939 return 0; 2940 } 2941 2942 void f2fs_quota_off_umount(struct super_block *sb) 2943 { 2944 } 2945 #endif 2946 2947 static const struct super_operations f2fs_sops = { 2948 .alloc_inode = f2fs_alloc_inode, 2949 .free_inode = f2fs_free_inode, 2950 .drop_inode = f2fs_drop_inode, 2951 .write_inode = f2fs_write_inode, 2952 .dirty_inode = f2fs_dirty_inode, 2953 .show_options = f2fs_show_options, 2954 #ifdef CONFIG_QUOTA 2955 .quota_read = f2fs_quota_read, 2956 .quota_write = f2fs_quota_write, 2957 .get_dquots = f2fs_get_dquots, 2958 #endif 2959 .evict_inode = f2fs_evict_inode, 2960 .put_super = f2fs_put_super, 2961 .sync_fs = f2fs_sync_fs, 2962 .freeze_fs = f2fs_freeze, 2963 .unfreeze_fs = f2fs_unfreeze, 2964 .statfs = f2fs_statfs, 2965 .remount_fs = f2fs_remount, 2966 }; 2967 2968 #ifdef CONFIG_FS_ENCRYPTION 2969 static int f2fs_get_context(struct inode *inode, void *ctx, size_t len) 2970 { 2971 return f2fs_getxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION, 2972 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT, 2973 ctx, len, NULL); 2974 } 2975 2976 static int f2fs_set_context(struct inode *inode, const void *ctx, size_t len, 2977 void *fs_data) 2978 { 2979 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 2980 2981 /* 2982 * Encrypting the root directory is not allowed because fsck 2983 * expects lost+found directory to exist and remain unencrypted 2984 * if LOST_FOUND feature is enabled. 2985 * 2986 */ 2987 if (f2fs_sb_has_lost_found(sbi) && 2988 inode->i_ino == F2FS_ROOT_INO(sbi)) 2989 return -EPERM; 2990 2991 return f2fs_setxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION, 2992 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT, 2993 ctx, len, fs_data, XATTR_CREATE); 2994 } 2995 2996 static const union fscrypt_policy *f2fs_get_dummy_policy(struct super_block *sb) 2997 { 2998 return F2FS_OPTION(F2FS_SB(sb)).dummy_enc_policy.policy; 2999 } 3000 3001 static bool f2fs_has_stable_inodes(struct super_block *sb) 3002 { 3003 return true; 3004 } 3005 3006 static void f2fs_get_ino_and_lblk_bits(struct super_block *sb, 3007 int *ino_bits_ret, int *lblk_bits_ret) 3008 { 3009 *ino_bits_ret = 8 * sizeof(nid_t); 3010 *lblk_bits_ret = 8 * sizeof(block_t); 3011 } 3012 3013 static int f2fs_get_num_devices(struct super_block *sb) 3014 { 3015 struct f2fs_sb_info *sbi = F2FS_SB(sb); 3016 3017 if (f2fs_is_multi_device(sbi)) 3018 return sbi->s_ndevs; 3019 return 1; 3020 } 3021 3022 static void f2fs_get_devices(struct super_block *sb, 3023 struct request_queue **devs) 3024 { 3025 struct f2fs_sb_info *sbi = F2FS_SB(sb); 3026 int i; 3027 3028 for (i = 0; i < sbi->s_ndevs; i++) 3029 devs[i] = bdev_get_queue(FDEV(i).bdev); 3030 } 3031 3032 static const struct fscrypt_operations f2fs_cryptops = { 3033 .key_prefix = "f2fs:", 3034 .get_context = f2fs_get_context, 3035 .set_context = f2fs_set_context, 3036 .get_dummy_policy = f2fs_get_dummy_policy, 3037 .empty_dir = f2fs_empty_dir, 3038 .has_stable_inodes = f2fs_has_stable_inodes, 3039 .get_ino_and_lblk_bits = f2fs_get_ino_and_lblk_bits, 3040 .get_num_devices = f2fs_get_num_devices, 3041 .get_devices = f2fs_get_devices, 3042 }; 3043 #endif 3044 3045 static struct inode *f2fs_nfs_get_inode(struct super_block *sb, 3046 u64 ino, u32 generation) 3047 { 3048 struct f2fs_sb_info *sbi = F2FS_SB(sb); 3049 struct inode *inode; 3050 3051 if (f2fs_check_nid_range(sbi, ino)) 3052 return ERR_PTR(-ESTALE); 3053 3054 /* 3055 * f2fs_iget isn't quite right if the inode is currently unallocated! 3056 * However f2fs_iget currently does appropriate checks to handle stale 3057 * inodes so everything is OK. 3058 */ 3059 inode = f2fs_iget(sb, ino); 3060 if (IS_ERR(inode)) 3061 return ERR_CAST(inode); 3062 if (unlikely(generation && inode->i_generation != generation)) { 3063 /* we didn't find the right inode.. */ 3064 iput(inode); 3065 return ERR_PTR(-ESTALE); 3066 } 3067 return inode; 3068 } 3069 3070 static struct dentry *f2fs_fh_to_dentry(struct super_block *sb, struct fid *fid, 3071 int fh_len, int fh_type) 3072 { 3073 return generic_fh_to_dentry(sb, fid, fh_len, fh_type, 3074 f2fs_nfs_get_inode); 3075 } 3076 3077 static struct dentry *f2fs_fh_to_parent(struct super_block *sb, struct fid *fid, 3078 int fh_len, int fh_type) 3079 { 3080 return generic_fh_to_parent(sb, fid, fh_len, fh_type, 3081 f2fs_nfs_get_inode); 3082 } 3083 3084 static const struct export_operations f2fs_export_ops = { 3085 .fh_to_dentry = f2fs_fh_to_dentry, 3086 .fh_to_parent = f2fs_fh_to_parent, 3087 .get_parent = f2fs_get_parent, 3088 }; 3089 3090 loff_t max_file_blocks(struct inode *inode) 3091 { 3092 loff_t result = 0; 3093 loff_t leaf_count; 3094 3095 /* 3096 * note: previously, result is equal to (DEF_ADDRS_PER_INODE - 3097 * DEFAULT_INLINE_XATTR_ADDRS), but now f2fs try to reserve more 3098 * space in inode.i_addr, it will be more safe to reassign 3099 * result as zero. 3100 */ 3101 3102 if (inode && f2fs_compressed_file(inode)) 3103 leaf_count = ADDRS_PER_BLOCK(inode); 3104 else 3105 leaf_count = DEF_ADDRS_PER_BLOCK; 3106 3107 /* two direct node blocks */ 3108 result += (leaf_count * 2); 3109 3110 /* two indirect node blocks */ 3111 leaf_count *= NIDS_PER_BLOCK; 3112 result += (leaf_count * 2); 3113 3114 /* one double indirect node block */ 3115 leaf_count *= NIDS_PER_BLOCK; 3116 result += leaf_count; 3117 3118 return result; 3119 } 3120 3121 static int __f2fs_commit_super(struct buffer_head *bh, 3122 struct f2fs_super_block *super) 3123 { 3124 lock_buffer(bh); 3125 if (super) 3126 memcpy(bh->b_data + F2FS_SUPER_OFFSET, super, sizeof(*super)); 3127 set_buffer_dirty(bh); 3128 unlock_buffer(bh); 3129 3130 /* it's rare case, we can do fua all the time */ 3131 return __sync_dirty_buffer(bh, REQ_SYNC | REQ_PREFLUSH | REQ_FUA); 3132 } 3133 3134 static inline bool sanity_check_area_boundary(struct f2fs_sb_info *sbi, 3135 struct buffer_head *bh) 3136 { 3137 struct f2fs_super_block *raw_super = (struct f2fs_super_block *) 3138 (bh->b_data + F2FS_SUPER_OFFSET); 3139 struct super_block *sb = sbi->sb; 3140 u32 segment0_blkaddr = le32_to_cpu(raw_super->segment0_blkaddr); 3141 u32 cp_blkaddr = le32_to_cpu(raw_super->cp_blkaddr); 3142 u32 sit_blkaddr = le32_to_cpu(raw_super->sit_blkaddr); 3143 u32 nat_blkaddr = le32_to_cpu(raw_super->nat_blkaddr); 3144 u32 ssa_blkaddr = le32_to_cpu(raw_super->ssa_blkaddr); 3145 u32 main_blkaddr = le32_to_cpu(raw_super->main_blkaddr); 3146 u32 segment_count_ckpt = le32_to_cpu(raw_super->segment_count_ckpt); 3147 u32 segment_count_sit = le32_to_cpu(raw_super->segment_count_sit); 3148 u32 segment_count_nat = le32_to_cpu(raw_super->segment_count_nat); 3149 u32 segment_count_ssa = le32_to_cpu(raw_super->segment_count_ssa); 3150 u32 segment_count_main = le32_to_cpu(raw_super->segment_count_main); 3151 u32 segment_count = le32_to_cpu(raw_super->segment_count); 3152 u32 log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg); 3153 u64 main_end_blkaddr = main_blkaddr + 3154 (segment_count_main << log_blocks_per_seg); 3155 u64 seg_end_blkaddr = segment0_blkaddr + 3156 (segment_count << log_blocks_per_seg); 3157 3158 if (segment0_blkaddr != cp_blkaddr) { 3159 f2fs_info(sbi, "Mismatch start address, segment0(%u) cp_blkaddr(%u)", 3160 segment0_blkaddr, cp_blkaddr); 3161 return true; 3162 } 3163 3164 if (cp_blkaddr + (segment_count_ckpt << log_blocks_per_seg) != 3165 sit_blkaddr) { 3166 f2fs_info(sbi, "Wrong CP boundary, start(%u) end(%u) blocks(%u)", 3167 cp_blkaddr, sit_blkaddr, 3168 segment_count_ckpt << log_blocks_per_seg); 3169 return true; 3170 } 3171 3172 if (sit_blkaddr + (segment_count_sit << log_blocks_per_seg) != 3173 nat_blkaddr) { 3174 f2fs_info(sbi, "Wrong SIT boundary, start(%u) end(%u) blocks(%u)", 3175 sit_blkaddr, nat_blkaddr, 3176 segment_count_sit << log_blocks_per_seg); 3177 return true; 3178 } 3179 3180 if (nat_blkaddr + (segment_count_nat << log_blocks_per_seg) != 3181 ssa_blkaddr) { 3182 f2fs_info(sbi, "Wrong NAT boundary, start(%u) end(%u) blocks(%u)", 3183 nat_blkaddr, ssa_blkaddr, 3184 segment_count_nat << log_blocks_per_seg); 3185 return true; 3186 } 3187 3188 if (ssa_blkaddr + (segment_count_ssa << log_blocks_per_seg) != 3189 main_blkaddr) { 3190 f2fs_info(sbi, "Wrong SSA boundary, start(%u) end(%u) blocks(%u)", 3191 ssa_blkaddr, main_blkaddr, 3192 segment_count_ssa << log_blocks_per_seg); 3193 return true; 3194 } 3195 3196 if (main_end_blkaddr > seg_end_blkaddr) { 3197 f2fs_info(sbi, "Wrong MAIN_AREA boundary, start(%u) end(%llu) block(%u)", 3198 main_blkaddr, seg_end_blkaddr, 3199 segment_count_main << log_blocks_per_seg); 3200 return true; 3201 } else if (main_end_blkaddr < seg_end_blkaddr) { 3202 int err = 0; 3203 char *res; 3204 3205 /* fix in-memory information all the time */ 3206 raw_super->segment_count = cpu_to_le32((main_end_blkaddr - 3207 segment0_blkaddr) >> log_blocks_per_seg); 3208 3209 if (f2fs_readonly(sb) || bdev_read_only(sb->s_bdev)) { 3210 set_sbi_flag(sbi, SBI_NEED_SB_WRITE); 3211 res = "internally"; 3212 } else { 3213 err = __f2fs_commit_super(bh, NULL); 3214 res = err ? "failed" : "done"; 3215 } 3216 f2fs_info(sbi, "Fix alignment : %s, start(%u) end(%llu) block(%u)", 3217 res, main_blkaddr, seg_end_blkaddr, 3218 segment_count_main << log_blocks_per_seg); 3219 if (err) 3220 return true; 3221 } 3222 return false; 3223 } 3224 3225 static int sanity_check_raw_super(struct f2fs_sb_info *sbi, 3226 struct buffer_head *bh) 3227 { 3228 block_t segment_count, segs_per_sec, secs_per_zone, segment_count_main; 3229 block_t total_sections, blocks_per_seg; 3230 struct f2fs_super_block *raw_super = (struct f2fs_super_block *) 3231 (bh->b_data + F2FS_SUPER_OFFSET); 3232 size_t crc_offset = 0; 3233 __u32 crc = 0; 3234 3235 if (le32_to_cpu(raw_super->magic) != F2FS_SUPER_MAGIC) { 3236 f2fs_info(sbi, "Magic Mismatch, valid(0x%x) - read(0x%x)", 3237 F2FS_SUPER_MAGIC, le32_to_cpu(raw_super->magic)); 3238 return -EINVAL; 3239 } 3240 3241 /* Check checksum_offset and crc in superblock */ 3242 if (__F2FS_HAS_FEATURE(raw_super, F2FS_FEATURE_SB_CHKSUM)) { 3243 crc_offset = le32_to_cpu(raw_super->checksum_offset); 3244 if (crc_offset != 3245 offsetof(struct f2fs_super_block, crc)) { 3246 f2fs_info(sbi, "Invalid SB checksum offset: %zu", 3247 crc_offset); 3248 return -EFSCORRUPTED; 3249 } 3250 crc = le32_to_cpu(raw_super->crc); 3251 if (!f2fs_crc_valid(sbi, crc, raw_super, crc_offset)) { 3252 f2fs_info(sbi, "Invalid SB checksum value: %u", crc); 3253 return -EFSCORRUPTED; 3254 } 3255 } 3256 3257 /* Currently, support only 4KB block size */ 3258 if (le32_to_cpu(raw_super->log_blocksize) != F2FS_BLKSIZE_BITS) { 3259 f2fs_info(sbi, "Invalid log_blocksize (%u), supports only %u", 3260 le32_to_cpu(raw_super->log_blocksize), 3261 F2FS_BLKSIZE_BITS); 3262 return -EFSCORRUPTED; 3263 } 3264 3265 /* check log blocks per segment */ 3266 if (le32_to_cpu(raw_super->log_blocks_per_seg) != 9) { 3267 f2fs_info(sbi, "Invalid log blocks per segment (%u)", 3268 le32_to_cpu(raw_super->log_blocks_per_seg)); 3269 return -EFSCORRUPTED; 3270 } 3271 3272 /* Currently, support 512/1024/2048/4096 bytes sector size */ 3273 if (le32_to_cpu(raw_super->log_sectorsize) > 3274 F2FS_MAX_LOG_SECTOR_SIZE || 3275 le32_to_cpu(raw_super->log_sectorsize) < 3276 F2FS_MIN_LOG_SECTOR_SIZE) { 3277 f2fs_info(sbi, "Invalid log sectorsize (%u)", 3278 le32_to_cpu(raw_super->log_sectorsize)); 3279 return -EFSCORRUPTED; 3280 } 3281 if (le32_to_cpu(raw_super->log_sectors_per_block) + 3282 le32_to_cpu(raw_super->log_sectorsize) != 3283 F2FS_MAX_LOG_SECTOR_SIZE) { 3284 f2fs_info(sbi, "Invalid log sectors per block(%u) log sectorsize(%u)", 3285 le32_to_cpu(raw_super->log_sectors_per_block), 3286 le32_to_cpu(raw_super->log_sectorsize)); 3287 return -EFSCORRUPTED; 3288 } 3289 3290 segment_count = le32_to_cpu(raw_super->segment_count); 3291 segment_count_main = le32_to_cpu(raw_super->segment_count_main); 3292 segs_per_sec = le32_to_cpu(raw_super->segs_per_sec); 3293 secs_per_zone = le32_to_cpu(raw_super->secs_per_zone); 3294 total_sections = le32_to_cpu(raw_super->section_count); 3295 3296 /* blocks_per_seg should be 512, given the above check */ 3297 blocks_per_seg = 1 << le32_to_cpu(raw_super->log_blocks_per_seg); 3298 3299 if (segment_count > F2FS_MAX_SEGMENT || 3300 segment_count < F2FS_MIN_SEGMENTS) { 3301 f2fs_info(sbi, "Invalid segment count (%u)", segment_count); 3302 return -EFSCORRUPTED; 3303 } 3304 3305 if (total_sections > segment_count_main || total_sections < 1 || 3306 segs_per_sec > segment_count || !segs_per_sec) { 3307 f2fs_info(sbi, "Invalid segment/section count (%u, %u x %u)", 3308 segment_count, total_sections, segs_per_sec); 3309 return -EFSCORRUPTED; 3310 } 3311 3312 if (segment_count_main != total_sections * segs_per_sec) { 3313 f2fs_info(sbi, "Invalid segment/section count (%u != %u * %u)", 3314 segment_count_main, total_sections, segs_per_sec); 3315 return -EFSCORRUPTED; 3316 } 3317 3318 if ((segment_count / segs_per_sec) < total_sections) { 3319 f2fs_info(sbi, "Small segment_count (%u < %u * %u)", 3320 segment_count, segs_per_sec, total_sections); 3321 return -EFSCORRUPTED; 3322 } 3323 3324 if (segment_count > (le64_to_cpu(raw_super->block_count) >> 9)) { 3325 f2fs_info(sbi, "Wrong segment_count / block_count (%u > %llu)", 3326 segment_count, le64_to_cpu(raw_super->block_count)); 3327 return -EFSCORRUPTED; 3328 } 3329 3330 if (RDEV(0).path[0]) { 3331 block_t dev_seg_count = le32_to_cpu(RDEV(0).total_segments); 3332 int i = 1; 3333 3334 while (i < MAX_DEVICES && RDEV(i).path[0]) { 3335 dev_seg_count += le32_to_cpu(RDEV(i).total_segments); 3336 i++; 3337 } 3338 if (segment_count != dev_seg_count) { 3339 f2fs_info(sbi, "Segment count (%u) mismatch with total segments from devices (%u)", 3340 segment_count, dev_seg_count); 3341 return -EFSCORRUPTED; 3342 } 3343 } else { 3344 if (__F2FS_HAS_FEATURE(raw_super, F2FS_FEATURE_BLKZONED) && 3345 !bdev_is_zoned(sbi->sb->s_bdev)) { 3346 f2fs_info(sbi, "Zoned block device path is missing"); 3347 return -EFSCORRUPTED; 3348 } 3349 } 3350 3351 if (secs_per_zone > total_sections || !secs_per_zone) { 3352 f2fs_info(sbi, "Wrong secs_per_zone / total_sections (%u, %u)", 3353 secs_per_zone, total_sections); 3354 return -EFSCORRUPTED; 3355 } 3356 if (le32_to_cpu(raw_super->extension_count) > F2FS_MAX_EXTENSION || 3357 raw_super->hot_ext_count > F2FS_MAX_EXTENSION || 3358 (le32_to_cpu(raw_super->extension_count) + 3359 raw_super->hot_ext_count) > F2FS_MAX_EXTENSION) { 3360 f2fs_info(sbi, "Corrupted extension count (%u + %u > %u)", 3361 le32_to_cpu(raw_super->extension_count), 3362 raw_super->hot_ext_count, 3363 F2FS_MAX_EXTENSION); 3364 return -EFSCORRUPTED; 3365 } 3366 3367 if (le32_to_cpu(raw_super->cp_payload) >= 3368 (blocks_per_seg - F2FS_CP_PACKS - 3369 NR_CURSEG_PERSIST_TYPE)) { 3370 f2fs_info(sbi, "Insane cp_payload (%u >= %u)", 3371 le32_to_cpu(raw_super->cp_payload), 3372 blocks_per_seg - F2FS_CP_PACKS - 3373 NR_CURSEG_PERSIST_TYPE); 3374 return -EFSCORRUPTED; 3375 } 3376 3377 /* check reserved ino info */ 3378 if (le32_to_cpu(raw_super->node_ino) != 1 || 3379 le32_to_cpu(raw_super->meta_ino) != 2 || 3380 le32_to_cpu(raw_super->root_ino) != 3) { 3381 f2fs_info(sbi, "Invalid Fs Meta Ino: node(%u) meta(%u) root(%u)", 3382 le32_to_cpu(raw_super->node_ino), 3383 le32_to_cpu(raw_super->meta_ino), 3384 le32_to_cpu(raw_super->root_ino)); 3385 return -EFSCORRUPTED; 3386 } 3387 3388 /* check CP/SIT/NAT/SSA/MAIN_AREA area boundary */ 3389 if (sanity_check_area_boundary(sbi, bh)) 3390 return -EFSCORRUPTED; 3391 3392 return 0; 3393 } 3394 3395 int f2fs_sanity_check_ckpt(struct f2fs_sb_info *sbi) 3396 { 3397 unsigned int total, fsmeta; 3398 struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi); 3399 struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi); 3400 unsigned int ovp_segments, reserved_segments; 3401 unsigned int main_segs, blocks_per_seg; 3402 unsigned int sit_segs, nat_segs; 3403 unsigned int sit_bitmap_size, nat_bitmap_size; 3404 unsigned int log_blocks_per_seg; 3405 unsigned int segment_count_main; 3406 unsigned int cp_pack_start_sum, cp_payload; 3407 block_t user_block_count, valid_user_blocks; 3408 block_t avail_node_count, valid_node_count; 3409 unsigned int nat_blocks, nat_bits_bytes, nat_bits_blocks; 3410 int i, j; 3411 3412 total = le32_to_cpu(raw_super->segment_count); 3413 fsmeta = le32_to_cpu(raw_super->segment_count_ckpt); 3414 sit_segs = le32_to_cpu(raw_super->segment_count_sit); 3415 fsmeta += sit_segs; 3416 nat_segs = le32_to_cpu(raw_super->segment_count_nat); 3417 fsmeta += nat_segs; 3418 fsmeta += le32_to_cpu(ckpt->rsvd_segment_count); 3419 fsmeta += le32_to_cpu(raw_super->segment_count_ssa); 3420 3421 if (unlikely(fsmeta >= total)) 3422 return 1; 3423 3424 ovp_segments = le32_to_cpu(ckpt->overprov_segment_count); 3425 reserved_segments = le32_to_cpu(ckpt->rsvd_segment_count); 3426 3427 if (!f2fs_sb_has_readonly(sbi) && 3428 unlikely(fsmeta < F2FS_MIN_META_SEGMENTS || 3429 ovp_segments == 0 || reserved_segments == 0)) { 3430 f2fs_err(sbi, "Wrong layout: check mkfs.f2fs version"); 3431 return 1; 3432 } 3433 user_block_count = le64_to_cpu(ckpt->user_block_count); 3434 segment_count_main = le32_to_cpu(raw_super->segment_count_main) + 3435 (f2fs_sb_has_readonly(sbi) ? 1 : 0); 3436 log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg); 3437 if (!user_block_count || user_block_count >= 3438 segment_count_main << log_blocks_per_seg) { 3439 f2fs_err(sbi, "Wrong user_block_count: %u", 3440 user_block_count); 3441 return 1; 3442 } 3443 3444 valid_user_blocks = le64_to_cpu(ckpt->valid_block_count); 3445 if (valid_user_blocks > user_block_count) { 3446 f2fs_err(sbi, "Wrong valid_user_blocks: %u, user_block_count: %u", 3447 valid_user_blocks, user_block_count); 3448 return 1; 3449 } 3450 3451 valid_node_count = le32_to_cpu(ckpt->valid_node_count); 3452 avail_node_count = sbi->total_node_count - F2FS_RESERVED_NODE_NUM; 3453 if (valid_node_count > avail_node_count) { 3454 f2fs_err(sbi, "Wrong valid_node_count: %u, avail_node_count: %u", 3455 valid_node_count, avail_node_count); 3456 return 1; 3457 } 3458 3459 main_segs = le32_to_cpu(raw_super->segment_count_main); 3460 blocks_per_seg = sbi->blocks_per_seg; 3461 3462 for (i = 0; i < NR_CURSEG_NODE_TYPE; i++) { 3463 if (le32_to_cpu(ckpt->cur_node_segno[i]) >= main_segs || 3464 le16_to_cpu(ckpt->cur_node_blkoff[i]) >= blocks_per_seg) 3465 return 1; 3466 3467 if (f2fs_sb_has_readonly(sbi)) 3468 goto check_data; 3469 3470 for (j = i + 1; j < NR_CURSEG_NODE_TYPE; j++) { 3471 if (le32_to_cpu(ckpt->cur_node_segno[i]) == 3472 le32_to_cpu(ckpt->cur_node_segno[j])) { 3473 f2fs_err(sbi, "Node segment (%u, %u) has the same segno: %u", 3474 i, j, 3475 le32_to_cpu(ckpt->cur_node_segno[i])); 3476 return 1; 3477 } 3478 } 3479 } 3480 check_data: 3481 for (i = 0; i < NR_CURSEG_DATA_TYPE; i++) { 3482 if (le32_to_cpu(ckpt->cur_data_segno[i]) >= main_segs || 3483 le16_to_cpu(ckpt->cur_data_blkoff[i]) >= blocks_per_seg) 3484 return 1; 3485 3486 if (f2fs_sb_has_readonly(sbi)) 3487 goto skip_cross; 3488 3489 for (j = i + 1; j < NR_CURSEG_DATA_TYPE; j++) { 3490 if (le32_to_cpu(ckpt->cur_data_segno[i]) == 3491 le32_to_cpu(ckpt->cur_data_segno[j])) { 3492 f2fs_err(sbi, "Data segment (%u, %u) has the same segno: %u", 3493 i, j, 3494 le32_to_cpu(ckpt->cur_data_segno[i])); 3495 return 1; 3496 } 3497 } 3498 } 3499 for (i = 0; i < NR_CURSEG_NODE_TYPE; i++) { 3500 for (j = 0; j < NR_CURSEG_DATA_TYPE; j++) { 3501 if (le32_to_cpu(ckpt->cur_node_segno[i]) == 3502 le32_to_cpu(ckpt->cur_data_segno[j])) { 3503 f2fs_err(sbi, "Node segment (%u) and Data segment (%u) has the same segno: %u", 3504 i, j, 3505 le32_to_cpu(ckpt->cur_node_segno[i])); 3506 return 1; 3507 } 3508 } 3509 } 3510 skip_cross: 3511 sit_bitmap_size = le32_to_cpu(ckpt->sit_ver_bitmap_bytesize); 3512 nat_bitmap_size = le32_to_cpu(ckpt->nat_ver_bitmap_bytesize); 3513 3514 if (sit_bitmap_size != ((sit_segs / 2) << log_blocks_per_seg) / 8 || 3515 nat_bitmap_size != ((nat_segs / 2) << log_blocks_per_seg) / 8) { 3516 f2fs_err(sbi, "Wrong bitmap size: sit: %u, nat:%u", 3517 sit_bitmap_size, nat_bitmap_size); 3518 return 1; 3519 } 3520 3521 cp_pack_start_sum = __start_sum_addr(sbi); 3522 cp_payload = __cp_payload(sbi); 3523 if (cp_pack_start_sum < cp_payload + 1 || 3524 cp_pack_start_sum > blocks_per_seg - 1 - 3525 NR_CURSEG_PERSIST_TYPE) { 3526 f2fs_err(sbi, "Wrong cp_pack_start_sum: %u", 3527 cp_pack_start_sum); 3528 return 1; 3529 } 3530 3531 if (__is_set_ckpt_flags(ckpt, CP_LARGE_NAT_BITMAP_FLAG) && 3532 le32_to_cpu(ckpt->checksum_offset) != CP_MIN_CHKSUM_OFFSET) { 3533 f2fs_warn(sbi, "using deprecated layout of large_nat_bitmap, " 3534 "please run fsck v1.13.0 or higher to repair, chksum_offset: %u, " 3535 "fixed with patch: \"f2fs-tools: relocate chksum_offset for large_nat_bitmap feature\"", 3536 le32_to_cpu(ckpt->checksum_offset)); 3537 return 1; 3538 } 3539 3540 nat_blocks = nat_segs << log_blocks_per_seg; 3541 nat_bits_bytes = nat_blocks / BITS_PER_BYTE; 3542 nat_bits_blocks = F2FS_BLK_ALIGN((nat_bits_bytes << 1) + 8); 3543 if (__is_set_ckpt_flags(ckpt, CP_NAT_BITS_FLAG) && 3544 (cp_payload + F2FS_CP_PACKS + 3545 NR_CURSEG_PERSIST_TYPE + nat_bits_blocks >= blocks_per_seg)) { 3546 f2fs_warn(sbi, "Insane cp_payload: %u, nat_bits_blocks: %u)", 3547 cp_payload, nat_bits_blocks); 3548 return 1; 3549 } 3550 3551 if (unlikely(f2fs_cp_error(sbi))) { 3552 f2fs_err(sbi, "A bug case: need to run fsck"); 3553 return 1; 3554 } 3555 return 0; 3556 } 3557 3558 static void init_sb_info(struct f2fs_sb_info *sbi) 3559 { 3560 struct f2fs_super_block *raw_super = sbi->raw_super; 3561 int i; 3562 3563 sbi->log_sectors_per_block = 3564 le32_to_cpu(raw_super->log_sectors_per_block); 3565 sbi->log_blocksize = le32_to_cpu(raw_super->log_blocksize); 3566 sbi->blocksize = 1 << sbi->log_blocksize; 3567 sbi->log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg); 3568 sbi->blocks_per_seg = 1 << sbi->log_blocks_per_seg; 3569 sbi->segs_per_sec = le32_to_cpu(raw_super->segs_per_sec); 3570 sbi->secs_per_zone = le32_to_cpu(raw_super->secs_per_zone); 3571 sbi->total_sections = le32_to_cpu(raw_super->section_count); 3572 sbi->total_node_count = 3573 (le32_to_cpu(raw_super->segment_count_nat) / 2) 3574 * sbi->blocks_per_seg * NAT_ENTRY_PER_BLOCK; 3575 F2FS_ROOT_INO(sbi) = le32_to_cpu(raw_super->root_ino); 3576 F2FS_NODE_INO(sbi) = le32_to_cpu(raw_super->node_ino); 3577 F2FS_META_INO(sbi) = le32_to_cpu(raw_super->meta_ino); 3578 sbi->cur_victim_sec = NULL_SECNO; 3579 sbi->next_victim_seg[BG_GC] = NULL_SEGNO; 3580 sbi->next_victim_seg[FG_GC] = NULL_SEGNO; 3581 sbi->max_victim_search = DEF_MAX_VICTIM_SEARCH; 3582 sbi->migration_granularity = sbi->segs_per_sec; 3583 sbi->seq_file_ra_mul = MIN_RA_MUL; 3584 sbi->max_fragment_chunk = DEF_FRAGMENT_SIZE; 3585 sbi->max_fragment_hole = DEF_FRAGMENT_SIZE; 3586 spin_lock_init(&sbi->gc_urgent_high_lock); 3587 3588 sbi->dir_level = DEF_DIR_LEVEL; 3589 sbi->interval_time[CP_TIME] = DEF_CP_INTERVAL; 3590 sbi->interval_time[REQ_TIME] = DEF_IDLE_INTERVAL; 3591 sbi->interval_time[DISCARD_TIME] = DEF_IDLE_INTERVAL; 3592 sbi->interval_time[GC_TIME] = DEF_IDLE_INTERVAL; 3593 sbi->interval_time[DISABLE_TIME] = DEF_DISABLE_INTERVAL; 3594 sbi->interval_time[UMOUNT_DISCARD_TIMEOUT] = 3595 DEF_UMOUNT_DISCARD_TIMEOUT; 3596 clear_sbi_flag(sbi, SBI_NEED_FSCK); 3597 3598 for (i = 0; i < NR_COUNT_TYPE; i++) 3599 atomic_set(&sbi->nr_pages[i], 0); 3600 3601 for (i = 0; i < META; i++) 3602 atomic_set(&sbi->wb_sync_req[i], 0); 3603 3604 INIT_LIST_HEAD(&sbi->s_list); 3605 mutex_init(&sbi->umount_mutex); 3606 init_rwsem(&sbi->io_order_lock); 3607 spin_lock_init(&sbi->cp_lock); 3608 3609 sbi->dirty_device = 0; 3610 spin_lock_init(&sbi->dev_lock); 3611 3612 init_rwsem(&sbi->sb_lock); 3613 init_rwsem(&sbi->pin_sem); 3614 } 3615 3616 static int init_percpu_info(struct f2fs_sb_info *sbi) 3617 { 3618 int err; 3619 3620 err = percpu_counter_init(&sbi->alloc_valid_block_count, 0, GFP_KERNEL); 3621 if (err) 3622 return err; 3623 3624 err = percpu_counter_init(&sbi->total_valid_inode_count, 0, 3625 GFP_KERNEL); 3626 if (err) 3627 percpu_counter_destroy(&sbi->alloc_valid_block_count); 3628 3629 return err; 3630 } 3631 3632 #ifdef CONFIG_BLK_DEV_ZONED 3633 3634 struct f2fs_report_zones_args { 3635 struct f2fs_dev_info *dev; 3636 bool zone_cap_mismatch; 3637 }; 3638 3639 static int f2fs_report_zone_cb(struct blk_zone *zone, unsigned int idx, 3640 void *data) 3641 { 3642 struct f2fs_report_zones_args *rz_args = data; 3643 3644 if (zone->type == BLK_ZONE_TYPE_CONVENTIONAL) 3645 return 0; 3646 3647 set_bit(idx, rz_args->dev->blkz_seq); 3648 rz_args->dev->zone_capacity_blocks[idx] = zone->capacity >> 3649 F2FS_LOG_SECTORS_PER_BLOCK; 3650 if (zone->len != zone->capacity && !rz_args->zone_cap_mismatch) 3651 rz_args->zone_cap_mismatch = true; 3652 3653 return 0; 3654 } 3655 3656 static int init_blkz_info(struct f2fs_sb_info *sbi, int devi) 3657 { 3658 struct block_device *bdev = FDEV(devi).bdev; 3659 sector_t nr_sectors = bdev_nr_sectors(bdev); 3660 struct f2fs_report_zones_args rep_zone_arg; 3661 int ret; 3662 3663 if (!f2fs_sb_has_blkzoned(sbi)) 3664 return 0; 3665 3666 if (sbi->blocks_per_blkz && sbi->blocks_per_blkz != 3667 SECTOR_TO_BLOCK(bdev_zone_sectors(bdev))) 3668 return -EINVAL; 3669 sbi->blocks_per_blkz = SECTOR_TO_BLOCK(bdev_zone_sectors(bdev)); 3670 if (sbi->log_blocks_per_blkz && sbi->log_blocks_per_blkz != 3671 __ilog2_u32(sbi->blocks_per_blkz)) 3672 return -EINVAL; 3673 sbi->log_blocks_per_blkz = __ilog2_u32(sbi->blocks_per_blkz); 3674 FDEV(devi).nr_blkz = SECTOR_TO_BLOCK(nr_sectors) >> 3675 sbi->log_blocks_per_blkz; 3676 if (nr_sectors & (bdev_zone_sectors(bdev) - 1)) 3677 FDEV(devi).nr_blkz++; 3678 3679 FDEV(devi).blkz_seq = f2fs_kvzalloc(sbi, 3680 BITS_TO_LONGS(FDEV(devi).nr_blkz) 3681 * sizeof(unsigned long), 3682 GFP_KERNEL); 3683 if (!FDEV(devi).blkz_seq) 3684 return -ENOMEM; 3685 3686 /* Get block zones type and zone-capacity */ 3687 FDEV(devi).zone_capacity_blocks = f2fs_kzalloc(sbi, 3688 FDEV(devi).nr_blkz * sizeof(block_t), 3689 GFP_KERNEL); 3690 if (!FDEV(devi).zone_capacity_blocks) 3691 return -ENOMEM; 3692 3693 rep_zone_arg.dev = &FDEV(devi); 3694 rep_zone_arg.zone_cap_mismatch = false; 3695 3696 ret = blkdev_report_zones(bdev, 0, BLK_ALL_ZONES, f2fs_report_zone_cb, 3697 &rep_zone_arg); 3698 if (ret < 0) 3699 return ret; 3700 3701 if (!rep_zone_arg.zone_cap_mismatch) { 3702 kfree(FDEV(devi).zone_capacity_blocks); 3703 FDEV(devi).zone_capacity_blocks = NULL; 3704 } 3705 3706 return 0; 3707 } 3708 #endif 3709 3710 /* 3711 * Read f2fs raw super block. 3712 * Because we have two copies of super block, so read both of them 3713 * to get the first valid one. If any one of them is broken, we pass 3714 * them recovery flag back to the caller. 3715 */ 3716 static int read_raw_super_block(struct f2fs_sb_info *sbi, 3717 struct f2fs_super_block **raw_super, 3718 int *valid_super_block, int *recovery) 3719 { 3720 struct super_block *sb = sbi->sb; 3721 int block; 3722 struct buffer_head *bh; 3723 struct f2fs_super_block *super; 3724 int err = 0; 3725 3726 super = kzalloc(sizeof(struct f2fs_super_block), GFP_KERNEL); 3727 if (!super) 3728 return -ENOMEM; 3729 3730 for (block = 0; block < 2; block++) { 3731 bh = sb_bread(sb, block); 3732 if (!bh) { 3733 f2fs_err(sbi, "Unable to read %dth superblock", 3734 block + 1); 3735 err = -EIO; 3736 *recovery = 1; 3737 continue; 3738 } 3739 3740 /* sanity checking of raw super */ 3741 err = sanity_check_raw_super(sbi, bh); 3742 if (err) { 3743 f2fs_err(sbi, "Can't find valid F2FS filesystem in %dth superblock", 3744 block + 1); 3745 brelse(bh); 3746 *recovery = 1; 3747 continue; 3748 } 3749 3750 if (!*raw_super) { 3751 memcpy(super, bh->b_data + F2FS_SUPER_OFFSET, 3752 sizeof(*super)); 3753 *valid_super_block = block; 3754 *raw_super = super; 3755 } 3756 brelse(bh); 3757 } 3758 3759 /* No valid superblock */ 3760 if (!*raw_super) 3761 kfree(super); 3762 else 3763 err = 0; 3764 3765 return err; 3766 } 3767 3768 int f2fs_commit_super(struct f2fs_sb_info *sbi, bool recover) 3769 { 3770 struct buffer_head *bh; 3771 __u32 crc = 0; 3772 int err; 3773 3774 if ((recover && f2fs_readonly(sbi->sb)) || 3775 bdev_read_only(sbi->sb->s_bdev)) { 3776 set_sbi_flag(sbi, SBI_NEED_SB_WRITE); 3777 return -EROFS; 3778 } 3779 3780 /* we should update superblock crc here */ 3781 if (!recover && f2fs_sb_has_sb_chksum(sbi)) { 3782 crc = f2fs_crc32(sbi, F2FS_RAW_SUPER(sbi), 3783 offsetof(struct f2fs_super_block, crc)); 3784 F2FS_RAW_SUPER(sbi)->crc = cpu_to_le32(crc); 3785 } 3786 3787 /* write back-up superblock first */ 3788 bh = sb_bread(sbi->sb, sbi->valid_super_block ? 0 : 1); 3789 if (!bh) 3790 return -EIO; 3791 err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi)); 3792 brelse(bh); 3793 3794 /* if we are in recovery path, skip writing valid superblock */ 3795 if (recover || err) 3796 return err; 3797 3798 /* write current valid superblock */ 3799 bh = sb_bread(sbi->sb, sbi->valid_super_block); 3800 if (!bh) 3801 return -EIO; 3802 err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi)); 3803 brelse(bh); 3804 return err; 3805 } 3806 3807 static int f2fs_scan_devices(struct f2fs_sb_info *sbi) 3808 { 3809 struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi); 3810 unsigned int max_devices = MAX_DEVICES; 3811 unsigned int logical_blksize; 3812 int i; 3813 3814 /* Initialize single device information */ 3815 if (!RDEV(0).path[0]) { 3816 if (!bdev_is_zoned(sbi->sb->s_bdev)) 3817 return 0; 3818 max_devices = 1; 3819 } 3820 3821 /* 3822 * Initialize multiple devices information, or single 3823 * zoned block device information. 3824 */ 3825 sbi->devs = f2fs_kzalloc(sbi, 3826 array_size(max_devices, 3827 sizeof(struct f2fs_dev_info)), 3828 GFP_KERNEL); 3829 if (!sbi->devs) 3830 return -ENOMEM; 3831 3832 logical_blksize = bdev_logical_block_size(sbi->sb->s_bdev); 3833 sbi->aligned_blksize = true; 3834 3835 for (i = 0; i < max_devices; i++) { 3836 3837 if (i > 0 && !RDEV(i).path[0]) 3838 break; 3839 3840 if (max_devices == 1) { 3841 /* Single zoned block device mount */ 3842 FDEV(0).bdev = 3843 blkdev_get_by_dev(sbi->sb->s_bdev->bd_dev, 3844 sbi->sb->s_mode, sbi->sb->s_type); 3845 } else { 3846 /* Multi-device mount */ 3847 memcpy(FDEV(i).path, RDEV(i).path, MAX_PATH_LEN); 3848 FDEV(i).total_segments = 3849 le32_to_cpu(RDEV(i).total_segments); 3850 if (i == 0) { 3851 FDEV(i).start_blk = 0; 3852 FDEV(i).end_blk = FDEV(i).start_blk + 3853 (FDEV(i).total_segments << 3854 sbi->log_blocks_per_seg) - 1 + 3855 le32_to_cpu(raw_super->segment0_blkaddr); 3856 } else { 3857 FDEV(i).start_blk = FDEV(i - 1).end_blk + 1; 3858 FDEV(i).end_blk = FDEV(i).start_blk + 3859 (FDEV(i).total_segments << 3860 sbi->log_blocks_per_seg) - 1; 3861 } 3862 FDEV(i).bdev = blkdev_get_by_path(FDEV(i).path, 3863 sbi->sb->s_mode, sbi->sb->s_type); 3864 } 3865 if (IS_ERR(FDEV(i).bdev)) 3866 return PTR_ERR(FDEV(i).bdev); 3867 3868 /* to release errored devices */ 3869 sbi->s_ndevs = i + 1; 3870 3871 if (logical_blksize != bdev_logical_block_size(FDEV(i).bdev)) 3872 sbi->aligned_blksize = false; 3873 3874 #ifdef CONFIG_BLK_DEV_ZONED 3875 if (bdev_zoned_model(FDEV(i).bdev) == BLK_ZONED_HM && 3876 !f2fs_sb_has_blkzoned(sbi)) { 3877 f2fs_err(sbi, "Zoned block device feature not enabled"); 3878 return -EINVAL; 3879 } 3880 if (bdev_zoned_model(FDEV(i).bdev) != BLK_ZONED_NONE) { 3881 if (init_blkz_info(sbi, i)) { 3882 f2fs_err(sbi, "Failed to initialize F2FS blkzone information"); 3883 return -EINVAL; 3884 } 3885 if (max_devices == 1) 3886 break; 3887 f2fs_info(sbi, "Mount Device [%2d]: %20s, %8u, %8x - %8x (zone: %s)", 3888 i, FDEV(i).path, 3889 FDEV(i).total_segments, 3890 FDEV(i).start_blk, FDEV(i).end_blk, 3891 bdev_zoned_model(FDEV(i).bdev) == BLK_ZONED_HA ? 3892 "Host-aware" : "Host-managed"); 3893 continue; 3894 } 3895 #endif 3896 f2fs_info(sbi, "Mount Device [%2d]: %20s, %8u, %8x - %8x", 3897 i, FDEV(i).path, 3898 FDEV(i).total_segments, 3899 FDEV(i).start_blk, FDEV(i).end_blk); 3900 } 3901 f2fs_info(sbi, 3902 "IO Block Size: %8d KB", F2FS_IO_SIZE_KB(sbi)); 3903 return 0; 3904 } 3905 3906 static int f2fs_setup_casefold(struct f2fs_sb_info *sbi) 3907 { 3908 #if IS_ENABLED(CONFIG_UNICODE) 3909 if (f2fs_sb_has_casefold(sbi) && !sbi->sb->s_encoding) { 3910 const struct f2fs_sb_encodings *encoding_info; 3911 struct unicode_map *encoding; 3912 __u16 encoding_flags; 3913 3914 encoding_info = f2fs_sb_read_encoding(sbi->raw_super); 3915 if (!encoding_info) { 3916 f2fs_err(sbi, 3917 "Encoding requested by superblock is unknown"); 3918 return -EINVAL; 3919 } 3920 3921 encoding_flags = le16_to_cpu(sbi->raw_super->s_encoding_flags); 3922 encoding = utf8_load(encoding_info->version); 3923 if (IS_ERR(encoding)) { 3924 f2fs_err(sbi, 3925 "can't mount with superblock charset: %s-%u.%u.%u " 3926 "not supported by the kernel. flags: 0x%x.", 3927 encoding_info->name, 3928 unicode_major(encoding_info->version), 3929 unicode_minor(encoding_info->version), 3930 unicode_rev(encoding_info->version), 3931 encoding_flags); 3932 return PTR_ERR(encoding); 3933 } 3934 f2fs_info(sbi, "Using encoding defined by superblock: " 3935 "%s-%u.%u.%u with flags 0x%hx", encoding_info->name, 3936 unicode_major(encoding_info->version), 3937 unicode_minor(encoding_info->version), 3938 unicode_rev(encoding_info->version), 3939 encoding_flags); 3940 3941 sbi->sb->s_encoding = encoding; 3942 sbi->sb->s_encoding_flags = encoding_flags; 3943 } 3944 #else 3945 if (f2fs_sb_has_casefold(sbi)) { 3946 f2fs_err(sbi, "Filesystem with casefold feature cannot be mounted without CONFIG_UNICODE"); 3947 return -EINVAL; 3948 } 3949 #endif 3950 return 0; 3951 } 3952 3953 static void f2fs_tuning_parameters(struct f2fs_sb_info *sbi) 3954 { 3955 struct f2fs_sm_info *sm_i = SM_I(sbi); 3956 3957 /* adjust parameters according to the volume size */ 3958 if (sm_i->main_segments <= SMALL_VOLUME_SEGMENTS) { 3959 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_REUSE; 3960 if (f2fs_block_unit_discard(sbi)) 3961 sm_i->dcc_info->discard_granularity = 1; 3962 sm_i->ipu_policy = 1 << F2FS_IPU_FORCE; 3963 } 3964 3965 sbi->readdir_ra = 1; 3966 } 3967 3968 static int f2fs_fill_super(struct super_block *sb, void *data, int silent) 3969 { 3970 struct f2fs_sb_info *sbi; 3971 struct f2fs_super_block *raw_super; 3972 struct inode *root; 3973 int err; 3974 bool skip_recovery = false, need_fsck = false; 3975 char *options = NULL; 3976 int recovery, i, valid_super_block; 3977 struct curseg_info *seg_i; 3978 int retry_cnt = 1; 3979 3980 try_onemore: 3981 err = -EINVAL; 3982 raw_super = NULL; 3983 valid_super_block = -1; 3984 recovery = 0; 3985 3986 /* allocate memory for f2fs-specific super block info */ 3987 sbi = kzalloc(sizeof(struct f2fs_sb_info), GFP_KERNEL); 3988 if (!sbi) 3989 return -ENOMEM; 3990 3991 sbi->sb = sb; 3992 3993 /* Load the checksum driver */ 3994 sbi->s_chksum_driver = crypto_alloc_shash("crc32", 0, 0); 3995 if (IS_ERR(sbi->s_chksum_driver)) { 3996 f2fs_err(sbi, "Cannot load crc32 driver."); 3997 err = PTR_ERR(sbi->s_chksum_driver); 3998 sbi->s_chksum_driver = NULL; 3999 goto free_sbi; 4000 } 4001 4002 /* set a block size */ 4003 if (unlikely(!sb_set_blocksize(sb, F2FS_BLKSIZE))) { 4004 f2fs_err(sbi, "unable to set blocksize"); 4005 goto free_sbi; 4006 } 4007 4008 err = read_raw_super_block(sbi, &raw_super, &valid_super_block, 4009 &recovery); 4010 if (err) 4011 goto free_sbi; 4012 4013 sb->s_fs_info = sbi; 4014 sbi->raw_super = raw_super; 4015 4016 /* precompute checksum seed for metadata */ 4017 if (f2fs_sb_has_inode_chksum(sbi)) 4018 sbi->s_chksum_seed = f2fs_chksum(sbi, ~0, raw_super->uuid, 4019 sizeof(raw_super->uuid)); 4020 4021 default_options(sbi); 4022 /* parse mount options */ 4023 options = kstrdup((const char *)data, GFP_KERNEL); 4024 if (data && !options) { 4025 err = -ENOMEM; 4026 goto free_sb_buf; 4027 } 4028 4029 err = parse_options(sb, options, false); 4030 if (err) 4031 goto free_options; 4032 4033 sb->s_maxbytes = max_file_blocks(NULL) << 4034 le32_to_cpu(raw_super->log_blocksize); 4035 sb->s_max_links = F2FS_LINK_MAX; 4036 4037 err = f2fs_setup_casefold(sbi); 4038 if (err) 4039 goto free_options; 4040 4041 #ifdef CONFIG_QUOTA 4042 sb->dq_op = &f2fs_quota_operations; 4043 sb->s_qcop = &f2fs_quotactl_ops; 4044 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ; 4045 4046 if (f2fs_sb_has_quota_ino(sbi)) { 4047 for (i = 0; i < MAXQUOTAS; i++) { 4048 if (f2fs_qf_ino(sbi->sb, i)) 4049 sbi->nquota_files++; 4050 } 4051 } 4052 #endif 4053 4054 sb->s_op = &f2fs_sops; 4055 #ifdef CONFIG_FS_ENCRYPTION 4056 sb->s_cop = &f2fs_cryptops; 4057 #endif 4058 #ifdef CONFIG_FS_VERITY 4059 sb->s_vop = &f2fs_verityops; 4060 #endif 4061 sb->s_xattr = f2fs_xattr_handlers; 4062 sb->s_export_op = &f2fs_export_ops; 4063 sb->s_magic = F2FS_SUPER_MAGIC; 4064 sb->s_time_gran = 1; 4065 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) | 4066 (test_opt(sbi, POSIX_ACL) ? SB_POSIXACL : 0); 4067 memcpy(&sb->s_uuid, raw_super->uuid, sizeof(raw_super->uuid)); 4068 sb->s_iflags |= SB_I_CGROUPWB; 4069 4070 /* init f2fs-specific super block info */ 4071 sbi->valid_super_block = valid_super_block; 4072 init_rwsem(&sbi->gc_lock); 4073 mutex_init(&sbi->writepages); 4074 init_rwsem(&sbi->cp_global_sem); 4075 init_rwsem(&sbi->node_write); 4076 init_rwsem(&sbi->node_change); 4077 4078 /* disallow all the data/node/meta page writes */ 4079 set_sbi_flag(sbi, SBI_POR_DOING); 4080 spin_lock_init(&sbi->stat_lock); 4081 4082 for (i = 0; i < NR_PAGE_TYPE; i++) { 4083 int n = (i == META) ? 1 : NR_TEMP_TYPE; 4084 int j; 4085 4086 sbi->write_io[i] = 4087 f2fs_kmalloc(sbi, 4088 array_size(n, 4089 sizeof(struct f2fs_bio_info)), 4090 GFP_KERNEL); 4091 if (!sbi->write_io[i]) { 4092 err = -ENOMEM; 4093 goto free_bio_info; 4094 } 4095 4096 for (j = HOT; j < n; j++) { 4097 init_rwsem(&sbi->write_io[i][j].io_rwsem); 4098 sbi->write_io[i][j].sbi = sbi; 4099 sbi->write_io[i][j].bio = NULL; 4100 spin_lock_init(&sbi->write_io[i][j].io_lock); 4101 INIT_LIST_HEAD(&sbi->write_io[i][j].io_list); 4102 INIT_LIST_HEAD(&sbi->write_io[i][j].bio_list); 4103 init_rwsem(&sbi->write_io[i][j].bio_list_lock); 4104 } 4105 } 4106 4107 init_rwsem(&sbi->cp_rwsem); 4108 init_rwsem(&sbi->quota_sem); 4109 init_waitqueue_head(&sbi->cp_wait); 4110 init_sb_info(sbi); 4111 4112 err = f2fs_init_iostat(sbi); 4113 if (err) 4114 goto free_bio_info; 4115 4116 err = init_percpu_info(sbi); 4117 if (err) 4118 goto free_iostat; 4119 4120 if (F2FS_IO_ALIGNED(sbi)) { 4121 sbi->write_io_dummy = 4122 mempool_create_page_pool(2 * (F2FS_IO_SIZE(sbi) - 1), 0); 4123 if (!sbi->write_io_dummy) { 4124 err = -ENOMEM; 4125 goto free_percpu; 4126 } 4127 } 4128 4129 /* init per sbi slab cache */ 4130 err = f2fs_init_xattr_caches(sbi); 4131 if (err) 4132 goto free_io_dummy; 4133 err = f2fs_init_page_array_cache(sbi); 4134 if (err) 4135 goto free_xattr_cache; 4136 4137 /* get an inode for meta space */ 4138 sbi->meta_inode = f2fs_iget(sb, F2FS_META_INO(sbi)); 4139 if (IS_ERR(sbi->meta_inode)) { 4140 f2fs_err(sbi, "Failed to read F2FS meta data inode"); 4141 err = PTR_ERR(sbi->meta_inode); 4142 goto free_page_array_cache; 4143 } 4144 4145 err = f2fs_get_valid_checkpoint(sbi); 4146 if (err) { 4147 f2fs_err(sbi, "Failed to get valid F2FS checkpoint"); 4148 goto free_meta_inode; 4149 } 4150 4151 if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_QUOTA_NEED_FSCK_FLAG)) 4152 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR); 4153 if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_DISABLED_QUICK_FLAG)) { 4154 set_sbi_flag(sbi, SBI_CP_DISABLED_QUICK); 4155 sbi->interval_time[DISABLE_TIME] = DEF_DISABLE_QUICK_INTERVAL; 4156 } 4157 4158 if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_FSCK_FLAG)) 4159 set_sbi_flag(sbi, SBI_NEED_FSCK); 4160 4161 /* Initialize device list */ 4162 err = f2fs_scan_devices(sbi); 4163 if (err) { 4164 f2fs_err(sbi, "Failed to find devices"); 4165 goto free_devices; 4166 } 4167 4168 err = f2fs_init_post_read_wq(sbi); 4169 if (err) { 4170 f2fs_err(sbi, "Failed to initialize post read workqueue"); 4171 goto free_devices; 4172 } 4173 4174 sbi->total_valid_node_count = 4175 le32_to_cpu(sbi->ckpt->valid_node_count); 4176 percpu_counter_set(&sbi->total_valid_inode_count, 4177 le32_to_cpu(sbi->ckpt->valid_inode_count)); 4178 sbi->user_block_count = le64_to_cpu(sbi->ckpt->user_block_count); 4179 sbi->total_valid_block_count = 4180 le64_to_cpu(sbi->ckpt->valid_block_count); 4181 sbi->last_valid_block_count = sbi->total_valid_block_count; 4182 sbi->reserved_blocks = 0; 4183 sbi->current_reserved_blocks = 0; 4184 limit_reserve_root(sbi); 4185 adjust_unusable_cap_perc(sbi); 4186 4187 for (i = 0; i < NR_INODE_TYPE; i++) { 4188 INIT_LIST_HEAD(&sbi->inode_list[i]); 4189 spin_lock_init(&sbi->inode_lock[i]); 4190 } 4191 mutex_init(&sbi->flush_lock); 4192 4193 f2fs_init_extent_cache_info(sbi); 4194 4195 f2fs_init_ino_entry_info(sbi); 4196 4197 f2fs_init_fsync_node_info(sbi); 4198 4199 /* setup checkpoint request control and start checkpoint issue thread */ 4200 f2fs_init_ckpt_req_control(sbi); 4201 if (!f2fs_readonly(sb) && !test_opt(sbi, DISABLE_CHECKPOINT) && 4202 test_opt(sbi, MERGE_CHECKPOINT)) { 4203 err = f2fs_start_ckpt_thread(sbi); 4204 if (err) { 4205 f2fs_err(sbi, 4206 "Failed to start F2FS issue_checkpoint_thread (%d)", 4207 err); 4208 goto stop_ckpt_thread; 4209 } 4210 } 4211 4212 /* setup f2fs internal modules */ 4213 err = f2fs_build_segment_manager(sbi); 4214 if (err) { 4215 f2fs_err(sbi, "Failed to initialize F2FS segment manager (%d)", 4216 err); 4217 goto free_sm; 4218 } 4219 err = f2fs_build_node_manager(sbi); 4220 if (err) { 4221 f2fs_err(sbi, "Failed to initialize F2FS node manager (%d)", 4222 err); 4223 goto free_nm; 4224 } 4225 4226 err = adjust_reserved_segment(sbi); 4227 if (err) 4228 goto free_nm; 4229 4230 /* For write statistics */ 4231 sbi->sectors_written_start = f2fs_get_sectors_written(sbi); 4232 4233 /* Read accumulated write IO statistics if exists */ 4234 seg_i = CURSEG_I(sbi, CURSEG_HOT_NODE); 4235 if (__exist_node_summaries(sbi)) 4236 sbi->kbytes_written = 4237 le64_to_cpu(seg_i->journal->info.kbytes_written); 4238 4239 f2fs_build_gc_manager(sbi); 4240 4241 err = f2fs_build_stats(sbi); 4242 if (err) 4243 goto free_nm; 4244 4245 /* get an inode for node space */ 4246 sbi->node_inode = f2fs_iget(sb, F2FS_NODE_INO(sbi)); 4247 if (IS_ERR(sbi->node_inode)) { 4248 f2fs_err(sbi, "Failed to read node inode"); 4249 err = PTR_ERR(sbi->node_inode); 4250 goto free_stats; 4251 } 4252 4253 /* read root inode and dentry */ 4254 root = f2fs_iget(sb, F2FS_ROOT_INO(sbi)); 4255 if (IS_ERR(root)) { 4256 f2fs_err(sbi, "Failed to read root inode"); 4257 err = PTR_ERR(root); 4258 goto free_node_inode; 4259 } 4260 if (!S_ISDIR(root->i_mode) || !root->i_blocks || 4261 !root->i_size || !root->i_nlink) { 4262 iput(root); 4263 err = -EINVAL; 4264 goto free_node_inode; 4265 } 4266 4267 sb->s_root = d_make_root(root); /* allocate root dentry */ 4268 if (!sb->s_root) { 4269 err = -ENOMEM; 4270 goto free_node_inode; 4271 } 4272 4273 err = f2fs_init_compress_inode(sbi); 4274 if (err) 4275 goto free_root_inode; 4276 4277 err = f2fs_register_sysfs(sbi); 4278 if (err) 4279 goto free_compress_inode; 4280 4281 #ifdef CONFIG_QUOTA 4282 /* Enable quota usage during mount */ 4283 if (f2fs_sb_has_quota_ino(sbi) && !f2fs_readonly(sb)) { 4284 err = f2fs_enable_quotas(sb); 4285 if (err) 4286 f2fs_err(sbi, "Cannot turn on quotas: error %d", err); 4287 } 4288 #endif 4289 /* if there are any orphan inodes, free them */ 4290 err = f2fs_recover_orphan_inodes(sbi); 4291 if (err) 4292 goto free_meta; 4293 4294 if (unlikely(is_set_ckpt_flags(sbi, CP_DISABLED_FLAG))) 4295 goto reset_checkpoint; 4296 4297 /* recover fsynced data */ 4298 if (!test_opt(sbi, DISABLE_ROLL_FORWARD) && 4299 !test_opt(sbi, NORECOVERY)) { 4300 /* 4301 * mount should be failed, when device has readonly mode, and 4302 * previous checkpoint was not done by clean system shutdown. 4303 */ 4304 if (f2fs_hw_is_readonly(sbi)) { 4305 if (!is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG)) { 4306 err = f2fs_recover_fsync_data(sbi, true); 4307 if (err > 0) { 4308 err = -EROFS; 4309 f2fs_err(sbi, "Need to recover fsync data, but " 4310 "write access unavailable, please try " 4311 "mount w/ disable_roll_forward or norecovery"); 4312 } 4313 if (err < 0) 4314 goto free_meta; 4315 } 4316 f2fs_info(sbi, "write access unavailable, skipping recovery"); 4317 goto reset_checkpoint; 4318 } 4319 4320 if (need_fsck) 4321 set_sbi_flag(sbi, SBI_NEED_FSCK); 4322 4323 if (skip_recovery) 4324 goto reset_checkpoint; 4325 4326 err = f2fs_recover_fsync_data(sbi, false); 4327 if (err < 0) { 4328 if (err != -ENOMEM) 4329 skip_recovery = true; 4330 need_fsck = true; 4331 f2fs_err(sbi, "Cannot recover all fsync data errno=%d", 4332 err); 4333 goto free_meta; 4334 } 4335 } else { 4336 err = f2fs_recover_fsync_data(sbi, true); 4337 4338 if (!f2fs_readonly(sb) && err > 0) { 4339 err = -EINVAL; 4340 f2fs_err(sbi, "Need to recover fsync data"); 4341 goto free_meta; 4342 } 4343 } 4344 4345 /* 4346 * If the f2fs is not readonly and fsync data recovery succeeds, 4347 * check zoned block devices' write pointer consistency. 4348 */ 4349 if (!err && !f2fs_readonly(sb) && f2fs_sb_has_blkzoned(sbi)) { 4350 err = f2fs_check_write_pointer(sbi); 4351 if (err) 4352 goto free_meta; 4353 } 4354 4355 reset_checkpoint: 4356 f2fs_init_inmem_curseg(sbi); 4357 4358 /* f2fs_recover_fsync_data() cleared this already */ 4359 clear_sbi_flag(sbi, SBI_POR_DOING); 4360 4361 if (test_opt(sbi, DISABLE_CHECKPOINT)) { 4362 err = f2fs_disable_checkpoint(sbi); 4363 if (err) 4364 goto sync_free_meta; 4365 } else if (is_set_ckpt_flags(sbi, CP_DISABLED_FLAG)) { 4366 f2fs_enable_checkpoint(sbi); 4367 } 4368 4369 /* 4370 * If filesystem is not mounted as read-only then 4371 * do start the gc_thread. 4372 */ 4373 if ((F2FS_OPTION(sbi).bggc_mode != BGGC_MODE_OFF || 4374 test_opt(sbi, GC_MERGE)) && !f2fs_readonly(sb)) { 4375 /* After POR, we can run background GC thread.*/ 4376 err = f2fs_start_gc_thread(sbi); 4377 if (err) 4378 goto sync_free_meta; 4379 } 4380 kvfree(options); 4381 4382 /* recover broken superblock */ 4383 if (recovery) { 4384 err = f2fs_commit_super(sbi, true); 4385 f2fs_info(sbi, "Try to recover %dth superblock, ret: %d", 4386 sbi->valid_super_block ? 1 : 2, err); 4387 } 4388 4389 f2fs_join_shrinker(sbi); 4390 4391 f2fs_tuning_parameters(sbi); 4392 4393 f2fs_notice(sbi, "Mounted with checkpoint version = %llx", 4394 cur_cp_version(F2FS_CKPT(sbi))); 4395 f2fs_update_time(sbi, CP_TIME); 4396 f2fs_update_time(sbi, REQ_TIME); 4397 clear_sbi_flag(sbi, SBI_CP_DISABLED_QUICK); 4398 return 0; 4399 4400 sync_free_meta: 4401 /* safe to flush all the data */ 4402 sync_filesystem(sbi->sb); 4403 retry_cnt = 0; 4404 4405 free_meta: 4406 #ifdef CONFIG_QUOTA 4407 f2fs_truncate_quota_inode_pages(sb); 4408 if (f2fs_sb_has_quota_ino(sbi) && !f2fs_readonly(sb)) 4409 f2fs_quota_off_umount(sbi->sb); 4410 #endif 4411 /* 4412 * Some dirty meta pages can be produced by f2fs_recover_orphan_inodes() 4413 * failed by EIO. Then, iput(node_inode) can trigger balance_fs_bg() 4414 * followed by f2fs_write_checkpoint() through f2fs_write_node_pages(), which 4415 * falls into an infinite loop in f2fs_sync_meta_pages(). 4416 */ 4417 truncate_inode_pages_final(META_MAPPING(sbi)); 4418 /* evict some inodes being cached by GC */ 4419 evict_inodes(sb); 4420 f2fs_unregister_sysfs(sbi); 4421 free_compress_inode: 4422 f2fs_destroy_compress_inode(sbi); 4423 free_root_inode: 4424 dput(sb->s_root); 4425 sb->s_root = NULL; 4426 free_node_inode: 4427 f2fs_release_ino_entry(sbi, true); 4428 truncate_inode_pages_final(NODE_MAPPING(sbi)); 4429 iput(sbi->node_inode); 4430 sbi->node_inode = NULL; 4431 free_stats: 4432 f2fs_destroy_stats(sbi); 4433 free_nm: 4434 /* stop discard thread before destroying node manager */ 4435 f2fs_stop_discard_thread(sbi); 4436 f2fs_destroy_node_manager(sbi); 4437 free_sm: 4438 f2fs_destroy_segment_manager(sbi); 4439 f2fs_destroy_post_read_wq(sbi); 4440 stop_ckpt_thread: 4441 f2fs_stop_ckpt_thread(sbi); 4442 free_devices: 4443 destroy_device_list(sbi); 4444 kvfree(sbi->ckpt); 4445 free_meta_inode: 4446 make_bad_inode(sbi->meta_inode); 4447 iput(sbi->meta_inode); 4448 sbi->meta_inode = NULL; 4449 free_page_array_cache: 4450 f2fs_destroy_page_array_cache(sbi); 4451 free_xattr_cache: 4452 f2fs_destroy_xattr_caches(sbi); 4453 free_io_dummy: 4454 mempool_destroy(sbi->write_io_dummy); 4455 free_percpu: 4456 destroy_percpu_info(sbi); 4457 free_iostat: 4458 f2fs_destroy_iostat(sbi); 4459 free_bio_info: 4460 for (i = 0; i < NR_PAGE_TYPE; i++) 4461 kvfree(sbi->write_io[i]); 4462 4463 #if IS_ENABLED(CONFIG_UNICODE) 4464 utf8_unload(sb->s_encoding); 4465 sb->s_encoding = NULL; 4466 #endif 4467 free_options: 4468 #ifdef CONFIG_QUOTA 4469 for (i = 0; i < MAXQUOTAS; i++) 4470 kfree(F2FS_OPTION(sbi).s_qf_names[i]); 4471 #endif 4472 fscrypt_free_dummy_policy(&F2FS_OPTION(sbi).dummy_enc_policy); 4473 kvfree(options); 4474 free_sb_buf: 4475 kfree(raw_super); 4476 free_sbi: 4477 if (sbi->s_chksum_driver) 4478 crypto_free_shash(sbi->s_chksum_driver); 4479 kfree(sbi); 4480 4481 /* give only one another chance */ 4482 if (retry_cnt > 0 && skip_recovery) { 4483 retry_cnt--; 4484 shrink_dcache_sb(sb); 4485 goto try_onemore; 4486 } 4487 return err; 4488 } 4489 4490 static struct dentry *f2fs_mount(struct file_system_type *fs_type, int flags, 4491 const char *dev_name, void *data) 4492 { 4493 return mount_bdev(fs_type, flags, dev_name, data, f2fs_fill_super); 4494 } 4495 4496 static void kill_f2fs_super(struct super_block *sb) 4497 { 4498 if (sb->s_root) { 4499 struct f2fs_sb_info *sbi = F2FS_SB(sb); 4500 4501 set_sbi_flag(sbi, SBI_IS_CLOSE); 4502 f2fs_stop_gc_thread(sbi); 4503 f2fs_stop_discard_thread(sbi); 4504 4505 #ifdef CONFIG_F2FS_FS_COMPRESSION 4506 /* 4507 * latter evict_inode() can bypass checking and invalidating 4508 * compress inode cache. 4509 */ 4510 if (test_opt(sbi, COMPRESS_CACHE)) 4511 truncate_inode_pages_final(COMPRESS_MAPPING(sbi)); 4512 #endif 4513 4514 if (is_sbi_flag_set(sbi, SBI_IS_DIRTY) || 4515 !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG)) { 4516 struct cp_control cpc = { 4517 .reason = CP_UMOUNT, 4518 }; 4519 f2fs_write_checkpoint(sbi, &cpc); 4520 } 4521 4522 if (is_sbi_flag_set(sbi, SBI_IS_RECOVERED) && f2fs_readonly(sb)) 4523 sb->s_flags &= ~SB_RDONLY; 4524 } 4525 kill_block_super(sb); 4526 } 4527 4528 static struct file_system_type f2fs_fs_type = { 4529 .owner = THIS_MODULE, 4530 .name = "f2fs", 4531 .mount = f2fs_mount, 4532 .kill_sb = kill_f2fs_super, 4533 .fs_flags = FS_REQUIRES_DEV, 4534 }; 4535 MODULE_ALIAS_FS("f2fs"); 4536 4537 static int __init init_inodecache(void) 4538 { 4539 f2fs_inode_cachep = kmem_cache_create("f2fs_inode_cache", 4540 sizeof(struct f2fs_inode_info), 0, 4541 SLAB_RECLAIM_ACCOUNT|SLAB_ACCOUNT, NULL); 4542 if (!f2fs_inode_cachep) 4543 return -ENOMEM; 4544 return 0; 4545 } 4546 4547 static void destroy_inodecache(void) 4548 { 4549 /* 4550 * Make sure all delayed rcu free inodes are flushed before we 4551 * destroy cache. 4552 */ 4553 rcu_barrier(); 4554 kmem_cache_destroy(f2fs_inode_cachep); 4555 } 4556 4557 static int __init init_f2fs_fs(void) 4558 { 4559 int err; 4560 4561 if (PAGE_SIZE != F2FS_BLKSIZE) { 4562 printk("F2FS not supported on PAGE_SIZE(%lu) != %d\n", 4563 PAGE_SIZE, F2FS_BLKSIZE); 4564 return -EINVAL; 4565 } 4566 4567 err = init_inodecache(); 4568 if (err) 4569 goto fail; 4570 err = f2fs_create_node_manager_caches(); 4571 if (err) 4572 goto free_inodecache; 4573 err = f2fs_create_segment_manager_caches(); 4574 if (err) 4575 goto free_node_manager_caches; 4576 err = f2fs_create_checkpoint_caches(); 4577 if (err) 4578 goto free_segment_manager_caches; 4579 err = f2fs_create_recovery_cache(); 4580 if (err) 4581 goto free_checkpoint_caches; 4582 err = f2fs_create_extent_cache(); 4583 if (err) 4584 goto free_recovery_cache; 4585 err = f2fs_create_garbage_collection_cache(); 4586 if (err) 4587 goto free_extent_cache; 4588 err = f2fs_init_sysfs(); 4589 if (err) 4590 goto free_garbage_collection_cache; 4591 err = register_shrinker(&f2fs_shrinker_info); 4592 if (err) 4593 goto free_sysfs; 4594 err = register_filesystem(&f2fs_fs_type); 4595 if (err) 4596 goto free_shrinker; 4597 f2fs_create_root_stats(); 4598 err = f2fs_init_post_read_processing(); 4599 if (err) 4600 goto free_root_stats; 4601 err = f2fs_init_iostat_processing(); 4602 if (err) 4603 goto free_post_read; 4604 err = f2fs_init_bio_entry_cache(); 4605 if (err) 4606 goto free_iostat; 4607 err = f2fs_init_bioset(); 4608 if (err) 4609 goto free_bio_enrty_cache; 4610 err = f2fs_init_compress_mempool(); 4611 if (err) 4612 goto free_bioset; 4613 err = f2fs_init_compress_cache(); 4614 if (err) 4615 goto free_compress_mempool; 4616 err = f2fs_create_casefold_cache(); 4617 if (err) 4618 goto free_compress_cache; 4619 return 0; 4620 free_compress_cache: 4621 f2fs_destroy_compress_cache(); 4622 free_compress_mempool: 4623 f2fs_destroy_compress_mempool(); 4624 free_bioset: 4625 f2fs_destroy_bioset(); 4626 free_bio_enrty_cache: 4627 f2fs_destroy_bio_entry_cache(); 4628 free_iostat: 4629 f2fs_destroy_iostat_processing(); 4630 free_post_read: 4631 f2fs_destroy_post_read_processing(); 4632 free_root_stats: 4633 f2fs_destroy_root_stats(); 4634 unregister_filesystem(&f2fs_fs_type); 4635 free_shrinker: 4636 unregister_shrinker(&f2fs_shrinker_info); 4637 free_sysfs: 4638 f2fs_exit_sysfs(); 4639 free_garbage_collection_cache: 4640 f2fs_destroy_garbage_collection_cache(); 4641 free_extent_cache: 4642 f2fs_destroy_extent_cache(); 4643 free_recovery_cache: 4644 f2fs_destroy_recovery_cache(); 4645 free_checkpoint_caches: 4646 f2fs_destroy_checkpoint_caches(); 4647 free_segment_manager_caches: 4648 f2fs_destroy_segment_manager_caches(); 4649 free_node_manager_caches: 4650 f2fs_destroy_node_manager_caches(); 4651 free_inodecache: 4652 destroy_inodecache(); 4653 fail: 4654 return err; 4655 } 4656 4657 static void __exit exit_f2fs_fs(void) 4658 { 4659 f2fs_destroy_casefold_cache(); 4660 f2fs_destroy_compress_cache(); 4661 f2fs_destroy_compress_mempool(); 4662 f2fs_destroy_bioset(); 4663 f2fs_destroy_bio_entry_cache(); 4664 f2fs_destroy_iostat_processing(); 4665 f2fs_destroy_post_read_processing(); 4666 f2fs_destroy_root_stats(); 4667 unregister_filesystem(&f2fs_fs_type); 4668 unregister_shrinker(&f2fs_shrinker_info); 4669 f2fs_exit_sysfs(); 4670 f2fs_destroy_garbage_collection_cache(); 4671 f2fs_destroy_extent_cache(); 4672 f2fs_destroy_recovery_cache(); 4673 f2fs_destroy_checkpoint_caches(); 4674 f2fs_destroy_segment_manager_caches(); 4675 f2fs_destroy_node_manager_caches(); 4676 destroy_inodecache(); 4677 } 4678 4679 module_init(init_f2fs_fs) 4680 module_exit(exit_f2fs_fs) 4681 4682 MODULE_AUTHOR("Samsung Electronics's Praesto Team"); 4683 MODULE_DESCRIPTION("Flash Friendly File System"); 4684 MODULE_LICENSE("GPL"); 4685 MODULE_SOFTDEP("pre: crc32"); 4686 4687