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_alloc, 142 Opt_fsync, 143 Opt_test_dummy_encryption, 144 Opt_inlinecrypt, 145 Opt_checkpoint_disable, 146 Opt_checkpoint_disable_cap, 147 Opt_checkpoint_disable_cap_perc, 148 Opt_checkpoint_enable, 149 Opt_checkpoint_merge, 150 Opt_nocheckpoint_merge, 151 Opt_compress_algorithm, 152 Opt_compress_log_size, 153 Opt_compress_extension, 154 Opt_nocompress_extension, 155 Opt_compress_chksum, 156 Opt_compress_mode, 157 Opt_compress_cache, 158 Opt_atgc, 159 Opt_gc_merge, 160 Opt_nogc_merge, 161 Opt_discard_unit, 162 Opt_err, 163 }; 164 165 static match_table_t f2fs_tokens = { 166 {Opt_gc_background, "background_gc=%s"}, 167 {Opt_disable_roll_forward, "disable_roll_forward"}, 168 {Opt_norecovery, "norecovery"}, 169 {Opt_discard, "discard"}, 170 {Opt_nodiscard, "nodiscard"}, 171 {Opt_noheap, "no_heap"}, 172 {Opt_heap, "heap"}, 173 {Opt_user_xattr, "user_xattr"}, 174 {Opt_nouser_xattr, "nouser_xattr"}, 175 {Opt_acl, "acl"}, 176 {Opt_noacl, "noacl"}, 177 {Opt_active_logs, "active_logs=%u"}, 178 {Opt_disable_ext_identify, "disable_ext_identify"}, 179 {Opt_inline_xattr, "inline_xattr"}, 180 {Opt_noinline_xattr, "noinline_xattr"}, 181 {Opt_inline_xattr_size, "inline_xattr_size=%u"}, 182 {Opt_inline_data, "inline_data"}, 183 {Opt_inline_dentry, "inline_dentry"}, 184 {Opt_noinline_dentry, "noinline_dentry"}, 185 {Opt_flush_merge, "flush_merge"}, 186 {Opt_noflush_merge, "noflush_merge"}, 187 {Opt_nobarrier, "nobarrier"}, 188 {Opt_fastboot, "fastboot"}, 189 {Opt_extent_cache, "extent_cache"}, 190 {Opt_noextent_cache, "noextent_cache"}, 191 {Opt_noinline_data, "noinline_data"}, 192 {Opt_data_flush, "data_flush"}, 193 {Opt_reserve_root, "reserve_root=%u"}, 194 {Opt_resgid, "resgid=%u"}, 195 {Opt_resuid, "resuid=%u"}, 196 {Opt_mode, "mode=%s"}, 197 {Opt_io_size_bits, "io_bits=%u"}, 198 {Opt_fault_injection, "fault_injection=%u"}, 199 {Opt_fault_type, "fault_type=%u"}, 200 {Opt_lazytime, "lazytime"}, 201 {Opt_nolazytime, "nolazytime"}, 202 {Opt_quota, "quota"}, 203 {Opt_noquota, "noquota"}, 204 {Opt_usrquota, "usrquota"}, 205 {Opt_grpquota, "grpquota"}, 206 {Opt_prjquota, "prjquota"}, 207 {Opt_usrjquota, "usrjquota=%s"}, 208 {Opt_grpjquota, "grpjquota=%s"}, 209 {Opt_prjjquota, "prjjquota=%s"}, 210 {Opt_offusrjquota, "usrjquota="}, 211 {Opt_offgrpjquota, "grpjquota="}, 212 {Opt_offprjjquota, "prjjquota="}, 213 {Opt_jqfmt_vfsold, "jqfmt=vfsold"}, 214 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"}, 215 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"}, 216 {Opt_alloc, "alloc_mode=%s"}, 217 {Opt_fsync, "fsync_mode=%s"}, 218 {Opt_test_dummy_encryption, "test_dummy_encryption=%s"}, 219 {Opt_test_dummy_encryption, "test_dummy_encryption"}, 220 {Opt_inlinecrypt, "inlinecrypt"}, 221 {Opt_checkpoint_disable, "checkpoint=disable"}, 222 {Opt_checkpoint_disable_cap, "checkpoint=disable:%u"}, 223 {Opt_checkpoint_disable_cap_perc, "checkpoint=disable:%u%%"}, 224 {Opt_checkpoint_enable, "checkpoint=enable"}, 225 {Opt_checkpoint_merge, "checkpoint_merge"}, 226 {Opt_nocheckpoint_merge, "nocheckpoint_merge"}, 227 {Opt_compress_algorithm, "compress_algorithm=%s"}, 228 {Opt_compress_log_size, "compress_log_size=%u"}, 229 {Opt_compress_extension, "compress_extension=%s"}, 230 {Opt_nocompress_extension, "nocompress_extension=%s"}, 231 {Opt_compress_chksum, "compress_chksum"}, 232 {Opt_compress_mode, "compress_mode=%s"}, 233 {Opt_compress_cache, "compress_cache"}, 234 {Opt_atgc, "atgc"}, 235 {Opt_gc_merge, "gc_merge"}, 236 {Opt_nogc_merge, "nogc_merge"}, 237 {Opt_discard_unit, "discard_unit=%s"}, 238 {Opt_err, NULL}, 239 }; 240 241 void f2fs_printk(struct f2fs_sb_info *sbi, const char *fmt, ...) 242 { 243 struct va_format vaf; 244 va_list args; 245 int level; 246 247 va_start(args, fmt); 248 249 level = printk_get_level(fmt); 250 vaf.fmt = printk_skip_level(fmt); 251 vaf.va = &args; 252 printk("%c%cF2FS-fs (%s): %pV\n", 253 KERN_SOH_ASCII, level, sbi->sb->s_id, &vaf); 254 255 va_end(args); 256 } 257 258 #if IS_ENABLED(CONFIG_UNICODE) 259 static const struct f2fs_sb_encodings { 260 __u16 magic; 261 char *name; 262 unsigned int version; 263 } f2fs_sb_encoding_map[] = { 264 {F2FS_ENC_UTF8_12_1, "utf8", UNICODE_AGE(12, 1, 0)}, 265 }; 266 267 static const struct f2fs_sb_encodings * 268 f2fs_sb_read_encoding(const struct f2fs_super_block *sb) 269 { 270 __u16 magic = le16_to_cpu(sb->s_encoding); 271 int i; 272 273 for (i = 0; i < ARRAY_SIZE(f2fs_sb_encoding_map); i++) 274 if (magic == f2fs_sb_encoding_map[i].magic) 275 return &f2fs_sb_encoding_map[i]; 276 277 return NULL; 278 } 279 280 struct kmem_cache *f2fs_cf_name_slab; 281 static int __init f2fs_create_casefold_cache(void) 282 { 283 f2fs_cf_name_slab = f2fs_kmem_cache_create("f2fs_casefolded_name", 284 F2FS_NAME_LEN); 285 if (!f2fs_cf_name_slab) 286 return -ENOMEM; 287 return 0; 288 } 289 290 static void f2fs_destroy_casefold_cache(void) 291 { 292 kmem_cache_destroy(f2fs_cf_name_slab); 293 } 294 #else 295 static int __init f2fs_create_casefold_cache(void) { return 0; } 296 static void f2fs_destroy_casefold_cache(void) { } 297 #endif 298 299 static inline void limit_reserve_root(struct f2fs_sb_info *sbi) 300 { 301 block_t limit = min((sbi->user_block_count << 1) / 1000, 302 sbi->user_block_count - sbi->reserved_blocks); 303 304 /* limit is 0.2% */ 305 if (test_opt(sbi, RESERVE_ROOT) && 306 F2FS_OPTION(sbi).root_reserved_blocks > limit) { 307 F2FS_OPTION(sbi).root_reserved_blocks = limit; 308 f2fs_info(sbi, "Reduce reserved blocks for root = %u", 309 F2FS_OPTION(sbi).root_reserved_blocks); 310 } 311 if (!test_opt(sbi, RESERVE_ROOT) && 312 (!uid_eq(F2FS_OPTION(sbi).s_resuid, 313 make_kuid(&init_user_ns, F2FS_DEF_RESUID)) || 314 !gid_eq(F2FS_OPTION(sbi).s_resgid, 315 make_kgid(&init_user_ns, F2FS_DEF_RESGID)))) 316 f2fs_info(sbi, "Ignore s_resuid=%u, s_resgid=%u w/o reserve_root", 317 from_kuid_munged(&init_user_ns, 318 F2FS_OPTION(sbi).s_resuid), 319 from_kgid_munged(&init_user_ns, 320 F2FS_OPTION(sbi).s_resgid)); 321 } 322 323 static inline int adjust_reserved_segment(struct f2fs_sb_info *sbi) 324 { 325 unsigned int sec_blks = sbi->blocks_per_seg * sbi->segs_per_sec; 326 unsigned int avg_vblocks; 327 unsigned int wanted_reserved_segments; 328 block_t avail_user_block_count; 329 330 if (!F2FS_IO_ALIGNED(sbi)) 331 return 0; 332 333 /* average valid block count in section in worst case */ 334 avg_vblocks = sec_blks / F2FS_IO_SIZE(sbi); 335 336 /* 337 * we need enough free space when migrating one section in worst case 338 */ 339 wanted_reserved_segments = (F2FS_IO_SIZE(sbi) / avg_vblocks) * 340 reserved_segments(sbi); 341 wanted_reserved_segments -= reserved_segments(sbi); 342 343 avail_user_block_count = sbi->user_block_count - 344 sbi->current_reserved_blocks - 345 F2FS_OPTION(sbi).root_reserved_blocks; 346 347 if (wanted_reserved_segments * sbi->blocks_per_seg > 348 avail_user_block_count) { 349 f2fs_err(sbi, "IO align feature can't grab additional reserved segment: %u, available segments: %u", 350 wanted_reserved_segments, 351 avail_user_block_count >> sbi->log_blocks_per_seg); 352 return -ENOSPC; 353 } 354 355 SM_I(sbi)->additional_reserved_segments = wanted_reserved_segments; 356 357 f2fs_info(sbi, "IO align feature needs additional reserved segment: %u", 358 wanted_reserved_segments); 359 360 return 0; 361 } 362 363 static inline void adjust_unusable_cap_perc(struct f2fs_sb_info *sbi) 364 { 365 if (!F2FS_OPTION(sbi).unusable_cap_perc) 366 return; 367 368 if (F2FS_OPTION(sbi).unusable_cap_perc == 100) 369 F2FS_OPTION(sbi).unusable_cap = sbi->user_block_count; 370 else 371 F2FS_OPTION(sbi).unusable_cap = (sbi->user_block_count / 100) * 372 F2FS_OPTION(sbi).unusable_cap_perc; 373 374 f2fs_info(sbi, "Adjust unusable cap for checkpoint=disable = %u / %u%%", 375 F2FS_OPTION(sbi).unusable_cap, 376 F2FS_OPTION(sbi).unusable_cap_perc); 377 } 378 379 static void init_once(void *foo) 380 { 381 struct f2fs_inode_info *fi = (struct f2fs_inode_info *) foo; 382 383 inode_init_once(&fi->vfs_inode); 384 } 385 386 #ifdef CONFIG_QUOTA 387 static const char * const quotatypes[] = INITQFNAMES; 388 #define QTYPE2NAME(t) (quotatypes[t]) 389 static int f2fs_set_qf_name(struct super_block *sb, int qtype, 390 substring_t *args) 391 { 392 struct f2fs_sb_info *sbi = F2FS_SB(sb); 393 char *qname; 394 int ret = -EINVAL; 395 396 if (sb_any_quota_loaded(sb) && !F2FS_OPTION(sbi).s_qf_names[qtype]) { 397 f2fs_err(sbi, "Cannot change journaled quota options when quota turned on"); 398 return -EINVAL; 399 } 400 if (f2fs_sb_has_quota_ino(sbi)) { 401 f2fs_info(sbi, "QUOTA feature is enabled, so ignore qf_name"); 402 return 0; 403 } 404 405 qname = match_strdup(args); 406 if (!qname) { 407 f2fs_err(sbi, "Not enough memory for storing quotafile name"); 408 return -ENOMEM; 409 } 410 if (F2FS_OPTION(sbi).s_qf_names[qtype]) { 411 if (strcmp(F2FS_OPTION(sbi).s_qf_names[qtype], qname) == 0) 412 ret = 0; 413 else 414 f2fs_err(sbi, "%s quota file already specified", 415 QTYPE2NAME(qtype)); 416 goto errout; 417 } 418 if (strchr(qname, '/')) { 419 f2fs_err(sbi, "quotafile must be on filesystem root"); 420 goto errout; 421 } 422 F2FS_OPTION(sbi).s_qf_names[qtype] = qname; 423 set_opt(sbi, QUOTA); 424 return 0; 425 errout: 426 kfree(qname); 427 return ret; 428 } 429 430 static int f2fs_clear_qf_name(struct super_block *sb, int qtype) 431 { 432 struct f2fs_sb_info *sbi = F2FS_SB(sb); 433 434 if (sb_any_quota_loaded(sb) && F2FS_OPTION(sbi).s_qf_names[qtype]) { 435 f2fs_err(sbi, "Cannot change journaled quota options when quota turned on"); 436 return -EINVAL; 437 } 438 kfree(F2FS_OPTION(sbi).s_qf_names[qtype]); 439 F2FS_OPTION(sbi).s_qf_names[qtype] = NULL; 440 return 0; 441 } 442 443 static int f2fs_check_quota_options(struct f2fs_sb_info *sbi) 444 { 445 /* 446 * We do the test below only for project quotas. 'usrquota' and 447 * 'grpquota' mount options are allowed even without quota feature 448 * to support legacy quotas in quota files. 449 */ 450 if (test_opt(sbi, PRJQUOTA) && !f2fs_sb_has_project_quota(sbi)) { 451 f2fs_err(sbi, "Project quota feature not enabled. Cannot enable project quota enforcement."); 452 return -1; 453 } 454 if (F2FS_OPTION(sbi).s_qf_names[USRQUOTA] || 455 F2FS_OPTION(sbi).s_qf_names[GRPQUOTA] || 456 F2FS_OPTION(sbi).s_qf_names[PRJQUOTA]) { 457 if (test_opt(sbi, USRQUOTA) && 458 F2FS_OPTION(sbi).s_qf_names[USRQUOTA]) 459 clear_opt(sbi, USRQUOTA); 460 461 if (test_opt(sbi, GRPQUOTA) && 462 F2FS_OPTION(sbi).s_qf_names[GRPQUOTA]) 463 clear_opt(sbi, GRPQUOTA); 464 465 if (test_opt(sbi, PRJQUOTA) && 466 F2FS_OPTION(sbi).s_qf_names[PRJQUOTA]) 467 clear_opt(sbi, PRJQUOTA); 468 469 if (test_opt(sbi, GRPQUOTA) || test_opt(sbi, USRQUOTA) || 470 test_opt(sbi, PRJQUOTA)) { 471 f2fs_err(sbi, "old and new quota format mixing"); 472 return -1; 473 } 474 475 if (!F2FS_OPTION(sbi).s_jquota_fmt) { 476 f2fs_err(sbi, "journaled quota format not specified"); 477 return -1; 478 } 479 } 480 481 if (f2fs_sb_has_quota_ino(sbi) && F2FS_OPTION(sbi).s_jquota_fmt) { 482 f2fs_info(sbi, "QUOTA feature is enabled, so ignore jquota_fmt"); 483 F2FS_OPTION(sbi).s_jquota_fmt = 0; 484 } 485 return 0; 486 } 487 #endif 488 489 static int f2fs_set_test_dummy_encryption(struct super_block *sb, 490 const char *opt, 491 const substring_t *arg, 492 bool is_remount) 493 { 494 struct f2fs_sb_info *sbi = F2FS_SB(sb); 495 #ifdef CONFIG_FS_ENCRYPTION 496 int err; 497 498 if (!f2fs_sb_has_encrypt(sbi)) { 499 f2fs_err(sbi, "Encrypt feature is off"); 500 return -EINVAL; 501 } 502 503 /* 504 * This mount option is just for testing, and it's not worthwhile to 505 * implement the extra complexity (e.g. RCU protection) that would be 506 * needed to allow it to be set or changed during remount. We do allow 507 * it to be specified during remount, but only if there is no change. 508 */ 509 if (is_remount && !F2FS_OPTION(sbi).dummy_enc_policy.policy) { 510 f2fs_warn(sbi, "Can't set test_dummy_encryption on remount"); 511 return -EINVAL; 512 } 513 err = fscrypt_set_test_dummy_encryption( 514 sb, arg->from, &F2FS_OPTION(sbi).dummy_enc_policy); 515 if (err) { 516 if (err == -EEXIST) 517 f2fs_warn(sbi, 518 "Can't change test_dummy_encryption on remount"); 519 else if (err == -EINVAL) 520 f2fs_warn(sbi, "Value of option \"%s\" is unrecognized", 521 opt); 522 else 523 f2fs_warn(sbi, "Error processing option \"%s\" [%d]", 524 opt, err); 525 return -EINVAL; 526 } 527 f2fs_warn(sbi, "Test dummy encryption mode enabled"); 528 return 0; 529 #else 530 f2fs_warn(sbi, "test_dummy_encryption option not supported"); 531 return -EINVAL; 532 #endif 533 } 534 535 #ifdef CONFIG_F2FS_FS_COMPRESSION 536 /* 537 * 1. The same extension name cannot not appear in both compress and non-compress extension 538 * at the same time. 539 * 2. If the compress extension specifies all files, the types specified by the non-compress 540 * extension will be treated as special cases and will not be compressed. 541 * 3. Don't allow the non-compress extension specifies all files. 542 */ 543 static int f2fs_test_compress_extension(struct f2fs_sb_info *sbi) 544 { 545 unsigned char (*ext)[F2FS_EXTENSION_LEN]; 546 unsigned char (*noext)[F2FS_EXTENSION_LEN]; 547 int ext_cnt, noext_cnt, index = 0, no_index = 0; 548 549 ext = F2FS_OPTION(sbi).extensions; 550 ext_cnt = F2FS_OPTION(sbi).compress_ext_cnt; 551 noext = F2FS_OPTION(sbi).noextensions; 552 noext_cnt = F2FS_OPTION(sbi).nocompress_ext_cnt; 553 554 if (!noext_cnt) 555 return 0; 556 557 for (no_index = 0; no_index < noext_cnt; no_index++) { 558 if (!strcasecmp("*", noext[no_index])) { 559 f2fs_info(sbi, "Don't allow the nocompress extension specifies all files"); 560 return -EINVAL; 561 } 562 for (index = 0; index < ext_cnt; index++) { 563 if (!strcasecmp(ext[index], noext[no_index])) { 564 f2fs_info(sbi, "Don't allow the same extension %s appear in both compress and nocompress extension", 565 ext[index]); 566 return -EINVAL; 567 } 568 } 569 } 570 return 0; 571 } 572 573 #ifdef CONFIG_F2FS_FS_LZ4 574 static int f2fs_set_lz4hc_level(struct f2fs_sb_info *sbi, const char *str) 575 { 576 #ifdef CONFIG_F2FS_FS_LZ4HC 577 unsigned int level; 578 #endif 579 580 if (strlen(str) == 3) { 581 F2FS_OPTION(sbi).compress_level = 0; 582 return 0; 583 } 584 585 #ifdef CONFIG_F2FS_FS_LZ4HC 586 str += 3; 587 588 if (str[0] != ':') { 589 f2fs_info(sbi, "wrong format, e.g. <alg_name>:<compr_level>"); 590 return -EINVAL; 591 } 592 if (kstrtouint(str + 1, 10, &level)) 593 return -EINVAL; 594 595 if (level < LZ4HC_MIN_CLEVEL || level > LZ4HC_MAX_CLEVEL) { 596 f2fs_info(sbi, "invalid lz4hc compress level: %d", level); 597 return -EINVAL; 598 } 599 600 F2FS_OPTION(sbi).compress_level = level; 601 return 0; 602 #else 603 f2fs_info(sbi, "kernel doesn't support lz4hc compression"); 604 return -EINVAL; 605 #endif 606 } 607 #endif 608 609 #ifdef CONFIG_F2FS_FS_ZSTD 610 static int f2fs_set_zstd_level(struct f2fs_sb_info *sbi, const char *str) 611 { 612 unsigned int level; 613 int len = 4; 614 615 if (strlen(str) == len) { 616 F2FS_OPTION(sbi).compress_level = 0; 617 return 0; 618 } 619 620 str += len; 621 622 if (str[0] != ':') { 623 f2fs_info(sbi, "wrong format, e.g. <alg_name>:<compr_level>"); 624 return -EINVAL; 625 } 626 if (kstrtouint(str + 1, 10, &level)) 627 return -EINVAL; 628 629 if (!level || level > zstd_max_clevel()) { 630 f2fs_info(sbi, "invalid zstd compress level: %d", level); 631 return -EINVAL; 632 } 633 634 F2FS_OPTION(sbi).compress_level = level; 635 return 0; 636 } 637 #endif 638 #endif 639 640 static int parse_options(struct super_block *sb, char *options, bool is_remount) 641 { 642 struct f2fs_sb_info *sbi = F2FS_SB(sb); 643 substring_t args[MAX_OPT_ARGS]; 644 #ifdef CONFIG_F2FS_FS_COMPRESSION 645 unsigned char (*ext)[F2FS_EXTENSION_LEN]; 646 unsigned char (*noext)[F2FS_EXTENSION_LEN]; 647 int ext_cnt, noext_cnt; 648 #endif 649 char *p, *name; 650 int arg = 0; 651 kuid_t uid; 652 kgid_t gid; 653 int ret; 654 655 if (!options) 656 goto default_check; 657 658 while ((p = strsep(&options, ",")) != NULL) { 659 int token; 660 661 if (!*p) 662 continue; 663 /* 664 * Initialize args struct so we know whether arg was 665 * found; some options take optional arguments. 666 */ 667 args[0].to = args[0].from = NULL; 668 token = match_token(p, f2fs_tokens, args); 669 670 switch (token) { 671 case Opt_gc_background: 672 name = match_strdup(&args[0]); 673 674 if (!name) 675 return -ENOMEM; 676 if (!strcmp(name, "on")) { 677 F2FS_OPTION(sbi).bggc_mode = BGGC_MODE_ON; 678 } else if (!strcmp(name, "off")) { 679 F2FS_OPTION(sbi).bggc_mode = BGGC_MODE_OFF; 680 } else if (!strcmp(name, "sync")) { 681 F2FS_OPTION(sbi).bggc_mode = BGGC_MODE_SYNC; 682 } else { 683 kfree(name); 684 return -EINVAL; 685 } 686 kfree(name); 687 break; 688 case Opt_disable_roll_forward: 689 set_opt(sbi, DISABLE_ROLL_FORWARD); 690 break; 691 case Opt_norecovery: 692 /* this option mounts f2fs with ro */ 693 set_opt(sbi, NORECOVERY); 694 if (!f2fs_readonly(sb)) 695 return -EINVAL; 696 break; 697 case Opt_discard: 698 if (!f2fs_hw_support_discard(sbi)) { 699 f2fs_warn(sbi, "device does not support discard"); 700 break; 701 } 702 set_opt(sbi, DISCARD); 703 break; 704 case Opt_nodiscard: 705 if (f2fs_hw_should_discard(sbi)) { 706 f2fs_warn(sbi, "discard is required for zoned block devices"); 707 return -EINVAL; 708 } 709 clear_opt(sbi, DISCARD); 710 break; 711 case Opt_noheap: 712 set_opt(sbi, NOHEAP); 713 break; 714 case Opt_heap: 715 clear_opt(sbi, NOHEAP); 716 break; 717 #ifdef CONFIG_F2FS_FS_XATTR 718 case Opt_user_xattr: 719 set_opt(sbi, XATTR_USER); 720 break; 721 case Opt_nouser_xattr: 722 clear_opt(sbi, XATTR_USER); 723 break; 724 case Opt_inline_xattr: 725 set_opt(sbi, INLINE_XATTR); 726 break; 727 case Opt_noinline_xattr: 728 clear_opt(sbi, INLINE_XATTR); 729 break; 730 case Opt_inline_xattr_size: 731 if (args->from && match_int(args, &arg)) 732 return -EINVAL; 733 set_opt(sbi, INLINE_XATTR_SIZE); 734 F2FS_OPTION(sbi).inline_xattr_size = arg; 735 break; 736 #else 737 case Opt_user_xattr: 738 f2fs_info(sbi, "user_xattr options not supported"); 739 break; 740 case Opt_nouser_xattr: 741 f2fs_info(sbi, "nouser_xattr options not supported"); 742 break; 743 case Opt_inline_xattr: 744 f2fs_info(sbi, "inline_xattr options not supported"); 745 break; 746 case Opt_noinline_xattr: 747 f2fs_info(sbi, "noinline_xattr options not supported"); 748 break; 749 #endif 750 #ifdef CONFIG_F2FS_FS_POSIX_ACL 751 case Opt_acl: 752 set_opt(sbi, POSIX_ACL); 753 break; 754 case Opt_noacl: 755 clear_opt(sbi, POSIX_ACL); 756 break; 757 #else 758 case Opt_acl: 759 f2fs_info(sbi, "acl options not supported"); 760 break; 761 case Opt_noacl: 762 f2fs_info(sbi, "noacl options not supported"); 763 break; 764 #endif 765 case Opt_active_logs: 766 if (args->from && match_int(args, &arg)) 767 return -EINVAL; 768 if (arg != 2 && arg != 4 && 769 arg != NR_CURSEG_PERSIST_TYPE) 770 return -EINVAL; 771 F2FS_OPTION(sbi).active_logs = arg; 772 break; 773 case Opt_disable_ext_identify: 774 set_opt(sbi, DISABLE_EXT_IDENTIFY); 775 break; 776 case Opt_inline_data: 777 set_opt(sbi, INLINE_DATA); 778 break; 779 case Opt_inline_dentry: 780 set_opt(sbi, INLINE_DENTRY); 781 break; 782 case Opt_noinline_dentry: 783 clear_opt(sbi, INLINE_DENTRY); 784 break; 785 case Opt_flush_merge: 786 set_opt(sbi, FLUSH_MERGE); 787 break; 788 case Opt_noflush_merge: 789 clear_opt(sbi, FLUSH_MERGE); 790 break; 791 case Opt_nobarrier: 792 set_opt(sbi, NOBARRIER); 793 break; 794 case Opt_fastboot: 795 set_opt(sbi, FASTBOOT); 796 break; 797 case Opt_extent_cache: 798 set_opt(sbi, EXTENT_CACHE); 799 break; 800 case Opt_noextent_cache: 801 clear_opt(sbi, EXTENT_CACHE); 802 break; 803 case Opt_noinline_data: 804 clear_opt(sbi, INLINE_DATA); 805 break; 806 case Opt_data_flush: 807 set_opt(sbi, DATA_FLUSH); 808 break; 809 case Opt_reserve_root: 810 if (args->from && match_int(args, &arg)) 811 return -EINVAL; 812 if (test_opt(sbi, RESERVE_ROOT)) { 813 f2fs_info(sbi, "Preserve previous reserve_root=%u", 814 F2FS_OPTION(sbi).root_reserved_blocks); 815 } else { 816 F2FS_OPTION(sbi).root_reserved_blocks = arg; 817 set_opt(sbi, RESERVE_ROOT); 818 } 819 break; 820 case Opt_resuid: 821 if (args->from && match_int(args, &arg)) 822 return -EINVAL; 823 uid = make_kuid(current_user_ns(), arg); 824 if (!uid_valid(uid)) { 825 f2fs_err(sbi, "Invalid uid value %d", arg); 826 return -EINVAL; 827 } 828 F2FS_OPTION(sbi).s_resuid = uid; 829 break; 830 case Opt_resgid: 831 if (args->from && match_int(args, &arg)) 832 return -EINVAL; 833 gid = make_kgid(current_user_ns(), arg); 834 if (!gid_valid(gid)) { 835 f2fs_err(sbi, "Invalid gid value %d", arg); 836 return -EINVAL; 837 } 838 F2FS_OPTION(sbi).s_resgid = gid; 839 break; 840 case Opt_mode: 841 name = match_strdup(&args[0]); 842 843 if (!name) 844 return -ENOMEM; 845 if (!strcmp(name, "adaptive")) { 846 if (f2fs_sb_has_blkzoned(sbi)) { 847 f2fs_warn(sbi, "adaptive mode is not allowed with zoned block device feature"); 848 kfree(name); 849 return -EINVAL; 850 } 851 F2FS_OPTION(sbi).fs_mode = FS_MODE_ADAPTIVE; 852 } else if (!strcmp(name, "lfs")) { 853 F2FS_OPTION(sbi).fs_mode = FS_MODE_LFS; 854 } else if (!strcmp(name, "fragment:segment")) { 855 F2FS_OPTION(sbi).fs_mode = FS_MODE_FRAGMENT_SEG; 856 } else if (!strcmp(name, "fragment:block")) { 857 F2FS_OPTION(sbi).fs_mode = FS_MODE_FRAGMENT_BLK; 858 } else { 859 kfree(name); 860 return -EINVAL; 861 } 862 kfree(name); 863 break; 864 case Opt_io_size_bits: 865 if (args->from && match_int(args, &arg)) 866 return -EINVAL; 867 if (arg <= 0 || arg > __ilog2_u32(BIO_MAX_VECS)) { 868 f2fs_warn(sbi, "Not support %d, larger than %d", 869 1 << arg, BIO_MAX_VECS); 870 return -EINVAL; 871 } 872 F2FS_OPTION(sbi).write_io_size_bits = arg; 873 break; 874 #ifdef CONFIG_F2FS_FAULT_INJECTION 875 case Opt_fault_injection: 876 if (args->from && match_int(args, &arg)) 877 return -EINVAL; 878 f2fs_build_fault_attr(sbi, arg, F2FS_ALL_FAULT_TYPE); 879 set_opt(sbi, FAULT_INJECTION); 880 break; 881 882 case Opt_fault_type: 883 if (args->from && match_int(args, &arg)) 884 return -EINVAL; 885 f2fs_build_fault_attr(sbi, 0, arg); 886 set_opt(sbi, FAULT_INJECTION); 887 break; 888 #else 889 case Opt_fault_injection: 890 f2fs_info(sbi, "fault_injection options not supported"); 891 break; 892 893 case Opt_fault_type: 894 f2fs_info(sbi, "fault_type options not supported"); 895 break; 896 #endif 897 case Opt_lazytime: 898 sb->s_flags |= SB_LAZYTIME; 899 break; 900 case Opt_nolazytime: 901 sb->s_flags &= ~SB_LAZYTIME; 902 break; 903 #ifdef CONFIG_QUOTA 904 case Opt_quota: 905 case Opt_usrquota: 906 set_opt(sbi, USRQUOTA); 907 break; 908 case Opt_grpquota: 909 set_opt(sbi, GRPQUOTA); 910 break; 911 case Opt_prjquota: 912 set_opt(sbi, PRJQUOTA); 913 break; 914 case Opt_usrjquota: 915 ret = f2fs_set_qf_name(sb, USRQUOTA, &args[0]); 916 if (ret) 917 return ret; 918 break; 919 case Opt_grpjquota: 920 ret = f2fs_set_qf_name(sb, GRPQUOTA, &args[0]); 921 if (ret) 922 return ret; 923 break; 924 case Opt_prjjquota: 925 ret = f2fs_set_qf_name(sb, PRJQUOTA, &args[0]); 926 if (ret) 927 return ret; 928 break; 929 case Opt_offusrjquota: 930 ret = f2fs_clear_qf_name(sb, USRQUOTA); 931 if (ret) 932 return ret; 933 break; 934 case Opt_offgrpjquota: 935 ret = f2fs_clear_qf_name(sb, GRPQUOTA); 936 if (ret) 937 return ret; 938 break; 939 case Opt_offprjjquota: 940 ret = f2fs_clear_qf_name(sb, PRJQUOTA); 941 if (ret) 942 return ret; 943 break; 944 case Opt_jqfmt_vfsold: 945 F2FS_OPTION(sbi).s_jquota_fmt = QFMT_VFS_OLD; 946 break; 947 case Opt_jqfmt_vfsv0: 948 F2FS_OPTION(sbi).s_jquota_fmt = QFMT_VFS_V0; 949 break; 950 case Opt_jqfmt_vfsv1: 951 F2FS_OPTION(sbi).s_jquota_fmt = QFMT_VFS_V1; 952 break; 953 case Opt_noquota: 954 clear_opt(sbi, QUOTA); 955 clear_opt(sbi, USRQUOTA); 956 clear_opt(sbi, GRPQUOTA); 957 clear_opt(sbi, PRJQUOTA); 958 break; 959 #else 960 case Opt_quota: 961 case Opt_usrquota: 962 case Opt_grpquota: 963 case Opt_prjquota: 964 case Opt_usrjquota: 965 case Opt_grpjquota: 966 case Opt_prjjquota: 967 case Opt_offusrjquota: 968 case Opt_offgrpjquota: 969 case Opt_offprjjquota: 970 case Opt_jqfmt_vfsold: 971 case Opt_jqfmt_vfsv0: 972 case Opt_jqfmt_vfsv1: 973 case Opt_noquota: 974 f2fs_info(sbi, "quota operations not supported"); 975 break; 976 #endif 977 case Opt_alloc: 978 name = match_strdup(&args[0]); 979 if (!name) 980 return -ENOMEM; 981 982 if (!strcmp(name, "default")) { 983 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_DEFAULT; 984 } else if (!strcmp(name, "reuse")) { 985 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_REUSE; 986 } else { 987 kfree(name); 988 return -EINVAL; 989 } 990 kfree(name); 991 break; 992 case Opt_fsync: 993 name = match_strdup(&args[0]); 994 if (!name) 995 return -ENOMEM; 996 if (!strcmp(name, "posix")) { 997 F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_POSIX; 998 } else if (!strcmp(name, "strict")) { 999 F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_STRICT; 1000 } else if (!strcmp(name, "nobarrier")) { 1001 F2FS_OPTION(sbi).fsync_mode = 1002 FSYNC_MODE_NOBARRIER; 1003 } else { 1004 kfree(name); 1005 return -EINVAL; 1006 } 1007 kfree(name); 1008 break; 1009 case Opt_test_dummy_encryption: 1010 ret = f2fs_set_test_dummy_encryption(sb, p, &args[0], 1011 is_remount); 1012 if (ret) 1013 return ret; 1014 break; 1015 case Opt_inlinecrypt: 1016 #ifdef CONFIG_FS_ENCRYPTION_INLINE_CRYPT 1017 sb->s_flags |= SB_INLINECRYPT; 1018 #else 1019 f2fs_info(sbi, "inline encryption not supported"); 1020 #endif 1021 break; 1022 case Opt_checkpoint_disable_cap_perc: 1023 if (args->from && match_int(args, &arg)) 1024 return -EINVAL; 1025 if (arg < 0 || arg > 100) 1026 return -EINVAL; 1027 F2FS_OPTION(sbi).unusable_cap_perc = arg; 1028 set_opt(sbi, DISABLE_CHECKPOINT); 1029 break; 1030 case Opt_checkpoint_disable_cap: 1031 if (args->from && match_int(args, &arg)) 1032 return -EINVAL; 1033 F2FS_OPTION(sbi).unusable_cap = arg; 1034 set_opt(sbi, DISABLE_CHECKPOINT); 1035 break; 1036 case Opt_checkpoint_disable: 1037 set_opt(sbi, DISABLE_CHECKPOINT); 1038 break; 1039 case Opt_checkpoint_enable: 1040 clear_opt(sbi, DISABLE_CHECKPOINT); 1041 break; 1042 case Opt_checkpoint_merge: 1043 set_opt(sbi, MERGE_CHECKPOINT); 1044 break; 1045 case Opt_nocheckpoint_merge: 1046 clear_opt(sbi, MERGE_CHECKPOINT); 1047 break; 1048 #ifdef CONFIG_F2FS_FS_COMPRESSION 1049 case Opt_compress_algorithm: 1050 if (!f2fs_sb_has_compression(sbi)) { 1051 f2fs_info(sbi, "Image doesn't support compression"); 1052 break; 1053 } 1054 name = match_strdup(&args[0]); 1055 if (!name) 1056 return -ENOMEM; 1057 if (!strcmp(name, "lzo")) { 1058 #ifdef CONFIG_F2FS_FS_LZO 1059 F2FS_OPTION(sbi).compress_level = 0; 1060 F2FS_OPTION(sbi).compress_algorithm = 1061 COMPRESS_LZO; 1062 #else 1063 f2fs_info(sbi, "kernel doesn't support lzo compression"); 1064 #endif 1065 } else if (!strncmp(name, "lz4", 3)) { 1066 #ifdef CONFIG_F2FS_FS_LZ4 1067 ret = f2fs_set_lz4hc_level(sbi, name); 1068 if (ret) { 1069 kfree(name); 1070 return -EINVAL; 1071 } 1072 F2FS_OPTION(sbi).compress_algorithm = 1073 COMPRESS_LZ4; 1074 #else 1075 f2fs_info(sbi, "kernel doesn't support lz4 compression"); 1076 #endif 1077 } else if (!strncmp(name, "zstd", 4)) { 1078 #ifdef CONFIG_F2FS_FS_ZSTD 1079 ret = f2fs_set_zstd_level(sbi, name); 1080 if (ret) { 1081 kfree(name); 1082 return -EINVAL; 1083 } 1084 F2FS_OPTION(sbi).compress_algorithm = 1085 COMPRESS_ZSTD; 1086 #else 1087 f2fs_info(sbi, "kernel doesn't support zstd compression"); 1088 #endif 1089 } else if (!strcmp(name, "lzo-rle")) { 1090 #ifdef CONFIG_F2FS_FS_LZORLE 1091 F2FS_OPTION(sbi).compress_level = 0; 1092 F2FS_OPTION(sbi).compress_algorithm = 1093 COMPRESS_LZORLE; 1094 #else 1095 f2fs_info(sbi, "kernel doesn't support lzorle compression"); 1096 #endif 1097 } else { 1098 kfree(name); 1099 return -EINVAL; 1100 } 1101 kfree(name); 1102 break; 1103 case Opt_compress_log_size: 1104 if (!f2fs_sb_has_compression(sbi)) { 1105 f2fs_info(sbi, "Image doesn't support compression"); 1106 break; 1107 } 1108 if (args->from && match_int(args, &arg)) 1109 return -EINVAL; 1110 if (arg < MIN_COMPRESS_LOG_SIZE || 1111 arg > MAX_COMPRESS_LOG_SIZE) { 1112 f2fs_err(sbi, 1113 "Compress cluster log size is out of range"); 1114 return -EINVAL; 1115 } 1116 F2FS_OPTION(sbi).compress_log_size = arg; 1117 break; 1118 case Opt_compress_extension: 1119 if (!f2fs_sb_has_compression(sbi)) { 1120 f2fs_info(sbi, "Image doesn't support compression"); 1121 break; 1122 } 1123 name = match_strdup(&args[0]); 1124 if (!name) 1125 return -ENOMEM; 1126 1127 ext = F2FS_OPTION(sbi).extensions; 1128 ext_cnt = F2FS_OPTION(sbi).compress_ext_cnt; 1129 1130 if (strlen(name) >= F2FS_EXTENSION_LEN || 1131 ext_cnt >= COMPRESS_EXT_NUM) { 1132 f2fs_err(sbi, 1133 "invalid extension length/number"); 1134 kfree(name); 1135 return -EINVAL; 1136 } 1137 1138 strcpy(ext[ext_cnt], name); 1139 F2FS_OPTION(sbi).compress_ext_cnt++; 1140 kfree(name); 1141 break; 1142 case Opt_nocompress_extension: 1143 if (!f2fs_sb_has_compression(sbi)) { 1144 f2fs_info(sbi, "Image doesn't support compression"); 1145 break; 1146 } 1147 name = match_strdup(&args[0]); 1148 if (!name) 1149 return -ENOMEM; 1150 1151 noext = F2FS_OPTION(sbi).noextensions; 1152 noext_cnt = F2FS_OPTION(sbi).nocompress_ext_cnt; 1153 1154 if (strlen(name) >= F2FS_EXTENSION_LEN || 1155 noext_cnt >= COMPRESS_EXT_NUM) { 1156 f2fs_err(sbi, 1157 "invalid extension length/number"); 1158 kfree(name); 1159 return -EINVAL; 1160 } 1161 1162 strcpy(noext[noext_cnt], name); 1163 F2FS_OPTION(sbi).nocompress_ext_cnt++; 1164 kfree(name); 1165 break; 1166 case Opt_compress_chksum: 1167 F2FS_OPTION(sbi).compress_chksum = true; 1168 break; 1169 case Opt_compress_mode: 1170 name = match_strdup(&args[0]); 1171 if (!name) 1172 return -ENOMEM; 1173 if (!strcmp(name, "fs")) { 1174 F2FS_OPTION(sbi).compress_mode = COMPR_MODE_FS; 1175 } else if (!strcmp(name, "user")) { 1176 F2FS_OPTION(sbi).compress_mode = COMPR_MODE_USER; 1177 } else { 1178 kfree(name); 1179 return -EINVAL; 1180 } 1181 kfree(name); 1182 break; 1183 case Opt_compress_cache: 1184 set_opt(sbi, COMPRESS_CACHE); 1185 break; 1186 #else 1187 case Opt_compress_algorithm: 1188 case Opt_compress_log_size: 1189 case Opt_compress_extension: 1190 case Opt_nocompress_extension: 1191 case Opt_compress_chksum: 1192 case Opt_compress_mode: 1193 case Opt_compress_cache: 1194 f2fs_info(sbi, "compression options not supported"); 1195 break; 1196 #endif 1197 case Opt_atgc: 1198 set_opt(sbi, ATGC); 1199 break; 1200 case Opt_gc_merge: 1201 set_opt(sbi, GC_MERGE); 1202 break; 1203 case Opt_nogc_merge: 1204 clear_opt(sbi, GC_MERGE); 1205 break; 1206 case Opt_discard_unit: 1207 name = match_strdup(&args[0]); 1208 if (!name) 1209 return -ENOMEM; 1210 if (!strcmp(name, "block")) { 1211 F2FS_OPTION(sbi).discard_unit = 1212 DISCARD_UNIT_BLOCK; 1213 } else if (!strcmp(name, "segment")) { 1214 F2FS_OPTION(sbi).discard_unit = 1215 DISCARD_UNIT_SEGMENT; 1216 } else if (!strcmp(name, "section")) { 1217 F2FS_OPTION(sbi).discard_unit = 1218 DISCARD_UNIT_SECTION; 1219 } else { 1220 kfree(name); 1221 return -EINVAL; 1222 } 1223 kfree(name); 1224 break; 1225 default: 1226 f2fs_err(sbi, "Unrecognized mount option \"%s\" or missing value", 1227 p); 1228 return -EINVAL; 1229 } 1230 } 1231 default_check: 1232 #ifdef CONFIG_QUOTA 1233 if (f2fs_check_quota_options(sbi)) 1234 return -EINVAL; 1235 #else 1236 if (f2fs_sb_has_quota_ino(sbi) && !f2fs_readonly(sbi->sb)) { 1237 f2fs_info(sbi, "Filesystem with quota feature cannot be mounted RDWR without CONFIG_QUOTA"); 1238 return -EINVAL; 1239 } 1240 if (f2fs_sb_has_project_quota(sbi) && !f2fs_readonly(sbi->sb)) { 1241 f2fs_err(sbi, "Filesystem with project quota feature cannot be mounted RDWR without CONFIG_QUOTA"); 1242 return -EINVAL; 1243 } 1244 #endif 1245 #if !IS_ENABLED(CONFIG_UNICODE) 1246 if (f2fs_sb_has_casefold(sbi)) { 1247 f2fs_err(sbi, 1248 "Filesystem with casefold feature cannot be mounted without CONFIG_UNICODE"); 1249 return -EINVAL; 1250 } 1251 #endif 1252 /* 1253 * The BLKZONED feature indicates that the drive was formatted with 1254 * zone alignment optimization. This is optional for host-aware 1255 * devices, but mandatory for host-managed zoned block devices. 1256 */ 1257 #ifndef CONFIG_BLK_DEV_ZONED 1258 if (f2fs_sb_has_blkzoned(sbi)) { 1259 f2fs_err(sbi, "Zoned block device support is not enabled"); 1260 return -EINVAL; 1261 } 1262 #endif 1263 if (f2fs_sb_has_blkzoned(sbi)) { 1264 if (F2FS_OPTION(sbi).discard_unit != 1265 DISCARD_UNIT_SECTION) { 1266 f2fs_info(sbi, "Zoned block device doesn't need small discard, set discard_unit=section by default"); 1267 F2FS_OPTION(sbi).discard_unit = 1268 DISCARD_UNIT_SECTION; 1269 } 1270 } 1271 1272 #ifdef CONFIG_F2FS_FS_COMPRESSION 1273 if (f2fs_test_compress_extension(sbi)) { 1274 f2fs_err(sbi, "invalid compress or nocompress extension"); 1275 return -EINVAL; 1276 } 1277 #endif 1278 1279 if (F2FS_IO_SIZE_BITS(sbi) && !f2fs_lfs_mode(sbi)) { 1280 f2fs_err(sbi, "Should set mode=lfs with %uKB-sized IO", 1281 F2FS_IO_SIZE_KB(sbi)); 1282 return -EINVAL; 1283 } 1284 1285 if (test_opt(sbi, INLINE_XATTR_SIZE)) { 1286 int min_size, max_size; 1287 1288 if (!f2fs_sb_has_extra_attr(sbi) || 1289 !f2fs_sb_has_flexible_inline_xattr(sbi)) { 1290 f2fs_err(sbi, "extra_attr or flexible_inline_xattr feature is off"); 1291 return -EINVAL; 1292 } 1293 if (!test_opt(sbi, INLINE_XATTR)) { 1294 f2fs_err(sbi, "inline_xattr_size option should be set with inline_xattr option"); 1295 return -EINVAL; 1296 } 1297 1298 min_size = sizeof(struct f2fs_xattr_header) / sizeof(__le32); 1299 max_size = MAX_INLINE_XATTR_SIZE; 1300 1301 if (F2FS_OPTION(sbi).inline_xattr_size < min_size || 1302 F2FS_OPTION(sbi).inline_xattr_size > max_size) { 1303 f2fs_err(sbi, "inline xattr size is out of range: %d ~ %d", 1304 min_size, max_size); 1305 return -EINVAL; 1306 } 1307 } 1308 1309 if (test_opt(sbi, DISABLE_CHECKPOINT) && f2fs_lfs_mode(sbi)) { 1310 f2fs_err(sbi, "LFS not compatible with checkpoint=disable"); 1311 return -EINVAL; 1312 } 1313 1314 if (f2fs_sb_has_readonly(sbi) && !f2fs_readonly(sbi->sb)) { 1315 f2fs_err(sbi, "Allow to mount readonly mode only"); 1316 return -EROFS; 1317 } 1318 return 0; 1319 } 1320 1321 static struct inode *f2fs_alloc_inode(struct super_block *sb) 1322 { 1323 struct f2fs_inode_info *fi; 1324 1325 if (time_to_inject(F2FS_SB(sb), FAULT_SLAB_ALLOC)) { 1326 f2fs_show_injection_info(F2FS_SB(sb), FAULT_SLAB_ALLOC); 1327 return NULL; 1328 } 1329 1330 fi = alloc_inode_sb(sb, f2fs_inode_cachep, GFP_F2FS_ZERO); 1331 if (!fi) 1332 return NULL; 1333 1334 init_once((void *) fi); 1335 1336 /* Initialize f2fs-specific inode info */ 1337 atomic_set(&fi->dirty_pages, 0); 1338 atomic_set(&fi->i_compr_blocks, 0); 1339 init_f2fs_rwsem(&fi->i_sem); 1340 spin_lock_init(&fi->i_size_lock); 1341 INIT_LIST_HEAD(&fi->dirty_list); 1342 INIT_LIST_HEAD(&fi->gdirty_list); 1343 init_f2fs_rwsem(&fi->i_gc_rwsem[READ]); 1344 init_f2fs_rwsem(&fi->i_gc_rwsem[WRITE]); 1345 init_f2fs_rwsem(&fi->i_xattr_sem); 1346 1347 /* Will be used by directory only */ 1348 fi->i_dir_level = F2FS_SB(sb)->dir_level; 1349 1350 return &fi->vfs_inode; 1351 } 1352 1353 static int f2fs_drop_inode(struct inode *inode) 1354 { 1355 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1356 int ret; 1357 1358 /* 1359 * during filesystem shutdown, if checkpoint is disabled, 1360 * drop useless meta/node dirty pages. 1361 */ 1362 if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED))) { 1363 if (inode->i_ino == F2FS_NODE_INO(sbi) || 1364 inode->i_ino == F2FS_META_INO(sbi)) { 1365 trace_f2fs_drop_inode(inode, 1); 1366 return 1; 1367 } 1368 } 1369 1370 /* 1371 * This is to avoid a deadlock condition like below. 1372 * writeback_single_inode(inode) 1373 * - f2fs_write_data_page 1374 * - f2fs_gc -> iput -> evict 1375 * - inode_wait_for_writeback(inode) 1376 */ 1377 if ((!inode_unhashed(inode) && inode->i_state & I_SYNC)) { 1378 if (!inode->i_nlink && !is_bad_inode(inode)) { 1379 /* to avoid evict_inode call simultaneously */ 1380 atomic_inc(&inode->i_count); 1381 spin_unlock(&inode->i_lock); 1382 1383 if (f2fs_is_atomic_file(inode)) 1384 f2fs_abort_atomic_write(inode, true); 1385 1386 /* should remain fi->extent_tree for writepage */ 1387 f2fs_destroy_extent_node(inode); 1388 1389 sb_start_intwrite(inode->i_sb); 1390 f2fs_i_size_write(inode, 0); 1391 1392 f2fs_submit_merged_write_cond(F2FS_I_SB(inode), 1393 inode, NULL, 0, DATA); 1394 truncate_inode_pages_final(inode->i_mapping); 1395 1396 if (F2FS_HAS_BLOCKS(inode)) 1397 f2fs_truncate(inode); 1398 1399 sb_end_intwrite(inode->i_sb); 1400 1401 spin_lock(&inode->i_lock); 1402 atomic_dec(&inode->i_count); 1403 } 1404 trace_f2fs_drop_inode(inode, 0); 1405 return 0; 1406 } 1407 ret = generic_drop_inode(inode); 1408 if (!ret) 1409 ret = fscrypt_drop_inode(inode); 1410 trace_f2fs_drop_inode(inode, ret); 1411 return ret; 1412 } 1413 1414 int f2fs_inode_dirtied(struct inode *inode, bool sync) 1415 { 1416 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1417 int ret = 0; 1418 1419 spin_lock(&sbi->inode_lock[DIRTY_META]); 1420 if (is_inode_flag_set(inode, FI_DIRTY_INODE)) { 1421 ret = 1; 1422 } else { 1423 set_inode_flag(inode, FI_DIRTY_INODE); 1424 stat_inc_dirty_inode(sbi, DIRTY_META); 1425 } 1426 if (sync && list_empty(&F2FS_I(inode)->gdirty_list)) { 1427 list_add_tail(&F2FS_I(inode)->gdirty_list, 1428 &sbi->inode_list[DIRTY_META]); 1429 inc_page_count(sbi, F2FS_DIRTY_IMETA); 1430 } 1431 spin_unlock(&sbi->inode_lock[DIRTY_META]); 1432 return ret; 1433 } 1434 1435 void f2fs_inode_synced(struct inode *inode) 1436 { 1437 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1438 1439 spin_lock(&sbi->inode_lock[DIRTY_META]); 1440 if (!is_inode_flag_set(inode, FI_DIRTY_INODE)) { 1441 spin_unlock(&sbi->inode_lock[DIRTY_META]); 1442 return; 1443 } 1444 if (!list_empty(&F2FS_I(inode)->gdirty_list)) { 1445 list_del_init(&F2FS_I(inode)->gdirty_list); 1446 dec_page_count(sbi, F2FS_DIRTY_IMETA); 1447 } 1448 clear_inode_flag(inode, FI_DIRTY_INODE); 1449 clear_inode_flag(inode, FI_AUTO_RECOVER); 1450 stat_dec_dirty_inode(F2FS_I_SB(inode), DIRTY_META); 1451 spin_unlock(&sbi->inode_lock[DIRTY_META]); 1452 } 1453 1454 /* 1455 * f2fs_dirty_inode() is called from __mark_inode_dirty() 1456 * 1457 * We should call set_dirty_inode to write the dirty inode through write_inode. 1458 */ 1459 static void f2fs_dirty_inode(struct inode *inode, int flags) 1460 { 1461 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1462 1463 if (inode->i_ino == F2FS_NODE_INO(sbi) || 1464 inode->i_ino == F2FS_META_INO(sbi)) 1465 return; 1466 1467 if (is_inode_flag_set(inode, FI_AUTO_RECOVER)) 1468 clear_inode_flag(inode, FI_AUTO_RECOVER); 1469 1470 f2fs_inode_dirtied(inode, false); 1471 } 1472 1473 static void f2fs_free_inode(struct inode *inode) 1474 { 1475 fscrypt_free_inode(inode); 1476 kmem_cache_free(f2fs_inode_cachep, F2FS_I(inode)); 1477 } 1478 1479 static void destroy_percpu_info(struct f2fs_sb_info *sbi) 1480 { 1481 percpu_counter_destroy(&sbi->total_valid_inode_count); 1482 percpu_counter_destroy(&sbi->rf_node_block_count); 1483 percpu_counter_destroy(&sbi->alloc_valid_block_count); 1484 } 1485 1486 static void destroy_device_list(struct f2fs_sb_info *sbi) 1487 { 1488 int i; 1489 1490 for (i = 0; i < sbi->s_ndevs; i++) { 1491 blkdev_put(FDEV(i).bdev, FMODE_EXCL); 1492 #ifdef CONFIG_BLK_DEV_ZONED 1493 kvfree(FDEV(i).blkz_seq); 1494 kfree(FDEV(i).zone_capacity_blocks); 1495 #endif 1496 } 1497 kvfree(sbi->devs); 1498 } 1499 1500 static void f2fs_put_super(struct super_block *sb) 1501 { 1502 struct f2fs_sb_info *sbi = F2FS_SB(sb); 1503 int i; 1504 bool dropped; 1505 1506 /* unregister procfs/sysfs entries in advance to avoid race case */ 1507 f2fs_unregister_sysfs(sbi); 1508 1509 f2fs_quota_off_umount(sb); 1510 1511 /* prevent remaining shrinker jobs */ 1512 mutex_lock(&sbi->umount_mutex); 1513 1514 /* 1515 * flush all issued checkpoints and stop checkpoint issue thread. 1516 * after then, all checkpoints should be done by each process context. 1517 */ 1518 f2fs_stop_ckpt_thread(sbi); 1519 1520 /* 1521 * We don't need to do checkpoint when superblock is clean. 1522 * But, the previous checkpoint was not done by umount, it needs to do 1523 * clean checkpoint again. 1524 */ 1525 if ((is_sbi_flag_set(sbi, SBI_IS_DIRTY) || 1526 !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG))) { 1527 struct cp_control cpc = { 1528 .reason = CP_UMOUNT, 1529 }; 1530 f2fs_write_checkpoint(sbi, &cpc); 1531 } 1532 1533 /* be sure to wait for any on-going discard commands */ 1534 dropped = f2fs_issue_discard_timeout(sbi); 1535 1536 if ((f2fs_hw_support_discard(sbi) || f2fs_hw_should_discard(sbi)) && 1537 !sbi->discard_blks && !dropped) { 1538 struct cp_control cpc = { 1539 .reason = CP_UMOUNT | CP_TRIMMED, 1540 }; 1541 f2fs_write_checkpoint(sbi, &cpc); 1542 } 1543 1544 /* 1545 * normally superblock is clean, so we need to release this. 1546 * In addition, EIO will skip do checkpoint, we need this as well. 1547 */ 1548 f2fs_release_ino_entry(sbi, true); 1549 1550 f2fs_leave_shrinker(sbi); 1551 mutex_unlock(&sbi->umount_mutex); 1552 1553 /* our cp_error case, we can wait for any writeback page */ 1554 f2fs_flush_merged_writes(sbi); 1555 1556 f2fs_wait_on_all_pages(sbi, F2FS_WB_CP_DATA); 1557 1558 f2fs_bug_on(sbi, sbi->fsync_node_num); 1559 1560 f2fs_destroy_compress_inode(sbi); 1561 1562 iput(sbi->node_inode); 1563 sbi->node_inode = NULL; 1564 1565 iput(sbi->meta_inode); 1566 sbi->meta_inode = NULL; 1567 1568 /* 1569 * iput() can update stat information, if f2fs_write_checkpoint() 1570 * above failed with error. 1571 */ 1572 f2fs_destroy_stats(sbi); 1573 1574 /* destroy f2fs internal modules */ 1575 f2fs_destroy_node_manager(sbi); 1576 f2fs_destroy_segment_manager(sbi); 1577 1578 f2fs_destroy_post_read_wq(sbi); 1579 1580 kvfree(sbi->ckpt); 1581 1582 sb->s_fs_info = NULL; 1583 if (sbi->s_chksum_driver) 1584 crypto_free_shash(sbi->s_chksum_driver); 1585 kfree(sbi->raw_super); 1586 1587 destroy_device_list(sbi); 1588 f2fs_destroy_page_array_cache(sbi); 1589 f2fs_destroy_xattr_caches(sbi); 1590 mempool_destroy(sbi->write_io_dummy); 1591 #ifdef CONFIG_QUOTA 1592 for (i = 0; i < MAXQUOTAS; i++) 1593 kfree(F2FS_OPTION(sbi).s_qf_names[i]); 1594 #endif 1595 fscrypt_free_dummy_policy(&F2FS_OPTION(sbi).dummy_enc_policy); 1596 destroy_percpu_info(sbi); 1597 f2fs_destroy_iostat(sbi); 1598 for (i = 0; i < NR_PAGE_TYPE; i++) 1599 kvfree(sbi->write_io[i]); 1600 #if IS_ENABLED(CONFIG_UNICODE) 1601 utf8_unload(sb->s_encoding); 1602 #endif 1603 kfree(sbi); 1604 } 1605 1606 int f2fs_sync_fs(struct super_block *sb, int sync) 1607 { 1608 struct f2fs_sb_info *sbi = F2FS_SB(sb); 1609 int err = 0; 1610 1611 if (unlikely(f2fs_cp_error(sbi))) 1612 return 0; 1613 if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED))) 1614 return 0; 1615 1616 trace_f2fs_sync_fs(sb, sync); 1617 1618 if (unlikely(is_sbi_flag_set(sbi, SBI_POR_DOING))) 1619 return -EAGAIN; 1620 1621 if (sync) 1622 err = f2fs_issue_checkpoint(sbi); 1623 1624 return err; 1625 } 1626 1627 static int f2fs_freeze(struct super_block *sb) 1628 { 1629 if (f2fs_readonly(sb)) 1630 return 0; 1631 1632 /* IO error happened before */ 1633 if (unlikely(f2fs_cp_error(F2FS_SB(sb)))) 1634 return -EIO; 1635 1636 /* must be clean, since sync_filesystem() was already called */ 1637 if (is_sbi_flag_set(F2FS_SB(sb), SBI_IS_DIRTY)) 1638 return -EINVAL; 1639 1640 /* ensure no checkpoint required */ 1641 if (!llist_empty(&F2FS_SB(sb)->cprc_info.issue_list)) 1642 return -EINVAL; 1643 1644 /* to avoid deadlock on f2fs_evict_inode->SB_FREEZE_FS */ 1645 set_sbi_flag(F2FS_SB(sb), SBI_IS_FREEZING); 1646 return 0; 1647 } 1648 1649 static int f2fs_unfreeze(struct super_block *sb) 1650 { 1651 clear_sbi_flag(F2FS_SB(sb), SBI_IS_FREEZING); 1652 return 0; 1653 } 1654 1655 #ifdef CONFIG_QUOTA 1656 static int f2fs_statfs_project(struct super_block *sb, 1657 kprojid_t projid, struct kstatfs *buf) 1658 { 1659 struct kqid qid; 1660 struct dquot *dquot; 1661 u64 limit; 1662 u64 curblock; 1663 1664 qid = make_kqid_projid(projid); 1665 dquot = dqget(sb, qid); 1666 if (IS_ERR(dquot)) 1667 return PTR_ERR(dquot); 1668 spin_lock(&dquot->dq_dqb_lock); 1669 1670 limit = min_not_zero(dquot->dq_dqb.dqb_bsoftlimit, 1671 dquot->dq_dqb.dqb_bhardlimit); 1672 if (limit) 1673 limit >>= sb->s_blocksize_bits; 1674 1675 if (limit && buf->f_blocks > limit) { 1676 curblock = (dquot->dq_dqb.dqb_curspace + 1677 dquot->dq_dqb.dqb_rsvspace) >> sb->s_blocksize_bits; 1678 buf->f_blocks = limit; 1679 buf->f_bfree = buf->f_bavail = 1680 (buf->f_blocks > curblock) ? 1681 (buf->f_blocks - curblock) : 0; 1682 } 1683 1684 limit = min_not_zero(dquot->dq_dqb.dqb_isoftlimit, 1685 dquot->dq_dqb.dqb_ihardlimit); 1686 1687 if (limit && buf->f_files > limit) { 1688 buf->f_files = limit; 1689 buf->f_ffree = 1690 (buf->f_files > dquot->dq_dqb.dqb_curinodes) ? 1691 (buf->f_files - dquot->dq_dqb.dqb_curinodes) : 0; 1692 } 1693 1694 spin_unlock(&dquot->dq_dqb_lock); 1695 dqput(dquot); 1696 return 0; 1697 } 1698 #endif 1699 1700 static int f2fs_statfs(struct dentry *dentry, struct kstatfs *buf) 1701 { 1702 struct super_block *sb = dentry->d_sb; 1703 struct f2fs_sb_info *sbi = F2FS_SB(sb); 1704 u64 id = huge_encode_dev(sb->s_bdev->bd_dev); 1705 block_t total_count, user_block_count, start_count; 1706 u64 avail_node_count; 1707 unsigned int total_valid_node_count; 1708 1709 total_count = le64_to_cpu(sbi->raw_super->block_count); 1710 start_count = le32_to_cpu(sbi->raw_super->segment0_blkaddr); 1711 buf->f_type = F2FS_SUPER_MAGIC; 1712 buf->f_bsize = sbi->blocksize; 1713 1714 buf->f_blocks = total_count - start_count; 1715 1716 spin_lock(&sbi->stat_lock); 1717 1718 user_block_count = sbi->user_block_count; 1719 total_valid_node_count = valid_node_count(sbi); 1720 avail_node_count = sbi->total_node_count - F2FS_RESERVED_NODE_NUM; 1721 buf->f_bfree = user_block_count - valid_user_blocks(sbi) - 1722 sbi->current_reserved_blocks; 1723 1724 if (unlikely(buf->f_bfree <= sbi->unusable_block_count)) 1725 buf->f_bfree = 0; 1726 else 1727 buf->f_bfree -= sbi->unusable_block_count; 1728 spin_unlock(&sbi->stat_lock); 1729 1730 if (buf->f_bfree > F2FS_OPTION(sbi).root_reserved_blocks) 1731 buf->f_bavail = buf->f_bfree - 1732 F2FS_OPTION(sbi).root_reserved_blocks; 1733 else 1734 buf->f_bavail = 0; 1735 1736 if (avail_node_count > user_block_count) { 1737 buf->f_files = user_block_count; 1738 buf->f_ffree = buf->f_bavail; 1739 } else { 1740 buf->f_files = avail_node_count; 1741 buf->f_ffree = min(avail_node_count - total_valid_node_count, 1742 buf->f_bavail); 1743 } 1744 1745 buf->f_namelen = F2FS_NAME_LEN; 1746 buf->f_fsid = u64_to_fsid(id); 1747 1748 #ifdef CONFIG_QUOTA 1749 if (is_inode_flag_set(dentry->d_inode, FI_PROJ_INHERIT) && 1750 sb_has_quota_limits_enabled(sb, PRJQUOTA)) { 1751 f2fs_statfs_project(sb, F2FS_I(dentry->d_inode)->i_projid, buf); 1752 } 1753 #endif 1754 return 0; 1755 } 1756 1757 static inline void f2fs_show_quota_options(struct seq_file *seq, 1758 struct super_block *sb) 1759 { 1760 #ifdef CONFIG_QUOTA 1761 struct f2fs_sb_info *sbi = F2FS_SB(sb); 1762 1763 if (F2FS_OPTION(sbi).s_jquota_fmt) { 1764 char *fmtname = ""; 1765 1766 switch (F2FS_OPTION(sbi).s_jquota_fmt) { 1767 case QFMT_VFS_OLD: 1768 fmtname = "vfsold"; 1769 break; 1770 case QFMT_VFS_V0: 1771 fmtname = "vfsv0"; 1772 break; 1773 case QFMT_VFS_V1: 1774 fmtname = "vfsv1"; 1775 break; 1776 } 1777 seq_printf(seq, ",jqfmt=%s", fmtname); 1778 } 1779 1780 if (F2FS_OPTION(sbi).s_qf_names[USRQUOTA]) 1781 seq_show_option(seq, "usrjquota", 1782 F2FS_OPTION(sbi).s_qf_names[USRQUOTA]); 1783 1784 if (F2FS_OPTION(sbi).s_qf_names[GRPQUOTA]) 1785 seq_show_option(seq, "grpjquota", 1786 F2FS_OPTION(sbi).s_qf_names[GRPQUOTA]); 1787 1788 if (F2FS_OPTION(sbi).s_qf_names[PRJQUOTA]) 1789 seq_show_option(seq, "prjjquota", 1790 F2FS_OPTION(sbi).s_qf_names[PRJQUOTA]); 1791 #endif 1792 } 1793 1794 #ifdef CONFIG_F2FS_FS_COMPRESSION 1795 static inline void f2fs_show_compress_options(struct seq_file *seq, 1796 struct super_block *sb) 1797 { 1798 struct f2fs_sb_info *sbi = F2FS_SB(sb); 1799 char *algtype = ""; 1800 int i; 1801 1802 if (!f2fs_sb_has_compression(sbi)) 1803 return; 1804 1805 switch (F2FS_OPTION(sbi).compress_algorithm) { 1806 case COMPRESS_LZO: 1807 algtype = "lzo"; 1808 break; 1809 case COMPRESS_LZ4: 1810 algtype = "lz4"; 1811 break; 1812 case COMPRESS_ZSTD: 1813 algtype = "zstd"; 1814 break; 1815 case COMPRESS_LZORLE: 1816 algtype = "lzo-rle"; 1817 break; 1818 } 1819 seq_printf(seq, ",compress_algorithm=%s", algtype); 1820 1821 if (F2FS_OPTION(sbi).compress_level) 1822 seq_printf(seq, ":%d", F2FS_OPTION(sbi).compress_level); 1823 1824 seq_printf(seq, ",compress_log_size=%u", 1825 F2FS_OPTION(sbi).compress_log_size); 1826 1827 for (i = 0; i < F2FS_OPTION(sbi).compress_ext_cnt; i++) { 1828 seq_printf(seq, ",compress_extension=%s", 1829 F2FS_OPTION(sbi).extensions[i]); 1830 } 1831 1832 for (i = 0; i < F2FS_OPTION(sbi).nocompress_ext_cnt; i++) { 1833 seq_printf(seq, ",nocompress_extension=%s", 1834 F2FS_OPTION(sbi).noextensions[i]); 1835 } 1836 1837 if (F2FS_OPTION(sbi).compress_chksum) 1838 seq_puts(seq, ",compress_chksum"); 1839 1840 if (F2FS_OPTION(sbi).compress_mode == COMPR_MODE_FS) 1841 seq_printf(seq, ",compress_mode=%s", "fs"); 1842 else if (F2FS_OPTION(sbi).compress_mode == COMPR_MODE_USER) 1843 seq_printf(seq, ",compress_mode=%s", "user"); 1844 1845 if (test_opt(sbi, COMPRESS_CACHE)) 1846 seq_puts(seq, ",compress_cache"); 1847 } 1848 #endif 1849 1850 static int f2fs_show_options(struct seq_file *seq, struct dentry *root) 1851 { 1852 struct f2fs_sb_info *sbi = F2FS_SB(root->d_sb); 1853 1854 if (F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_SYNC) 1855 seq_printf(seq, ",background_gc=%s", "sync"); 1856 else if (F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_ON) 1857 seq_printf(seq, ",background_gc=%s", "on"); 1858 else if (F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_OFF) 1859 seq_printf(seq, ",background_gc=%s", "off"); 1860 1861 if (test_opt(sbi, GC_MERGE)) 1862 seq_puts(seq, ",gc_merge"); 1863 1864 if (test_opt(sbi, DISABLE_ROLL_FORWARD)) 1865 seq_puts(seq, ",disable_roll_forward"); 1866 if (test_opt(sbi, NORECOVERY)) 1867 seq_puts(seq, ",norecovery"); 1868 if (test_opt(sbi, DISCARD)) 1869 seq_puts(seq, ",discard"); 1870 else 1871 seq_puts(seq, ",nodiscard"); 1872 if (test_opt(sbi, NOHEAP)) 1873 seq_puts(seq, ",no_heap"); 1874 else 1875 seq_puts(seq, ",heap"); 1876 #ifdef CONFIG_F2FS_FS_XATTR 1877 if (test_opt(sbi, XATTR_USER)) 1878 seq_puts(seq, ",user_xattr"); 1879 else 1880 seq_puts(seq, ",nouser_xattr"); 1881 if (test_opt(sbi, INLINE_XATTR)) 1882 seq_puts(seq, ",inline_xattr"); 1883 else 1884 seq_puts(seq, ",noinline_xattr"); 1885 if (test_opt(sbi, INLINE_XATTR_SIZE)) 1886 seq_printf(seq, ",inline_xattr_size=%u", 1887 F2FS_OPTION(sbi).inline_xattr_size); 1888 #endif 1889 #ifdef CONFIG_F2FS_FS_POSIX_ACL 1890 if (test_opt(sbi, POSIX_ACL)) 1891 seq_puts(seq, ",acl"); 1892 else 1893 seq_puts(seq, ",noacl"); 1894 #endif 1895 if (test_opt(sbi, DISABLE_EXT_IDENTIFY)) 1896 seq_puts(seq, ",disable_ext_identify"); 1897 if (test_opt(sbi, INLINE_DATA)) 1898 seq_puts(seq, ",inline_data"); 1899 else 1900 seq_puts(seq, ",noinline_data"); 1901 if (test_opt(sbi, INLINE_DENTRY)) 1902 seq_puts(seq, ",inline_dentry"); 1903 else 1904 seq_puts(seq, ",noinline_dentry"); 1905 if (!f2fs_readonly(sbi->sb) && test_opt(sbi, FLUSH_MERGE)) 1906 seq_puts(seq, ",flush_merge"); 1907 if (test_opt(sbi, NOBARRIER)) 1908 seq_puts(seq, ",nobarrier"); 1909 if (test_opt(sbi, FASTBOOT)) 1910 seq_puts(seq, ",fastboot"); 1911 if (test_opt(sbi, EXTENT_CACHE)) 1912 seq_puts(seq, ",extent_cache"); 1913 else 1914 seq_puts(seq, ",noextent_cache"); 1915 if (test_opt(sbi, DATA_FLUSH)) 1916 seq_puts(seq, ",data_flush"); 1917 1918 seq_puts(seq, ",mode="); 1919 if (F2FS_OPTION(sbi).fs_mode == FS_MODE_ADAPTIVE) 1920 seq_puts(seq, "adaptive"); 1921 else if (F2FS_OPTION(sbi).fs_mode == FS_MODE_LFS) 1922 seq_puts(seq, "lfs"); 1923 else if (F2FS_OPTION(sbi).fs_mode == FS_MODE_FRAGMENT_SEG) 1924 seq_puts(seq, "fragment:segment"); 1925 else if (F2FS_OPTION(sbi).fs_mode == FS_MODE_FRAGMENT_BLK) 1926 seq_puts(seq, "fragment:block"); 1927 seq_printf(seq, ",active_logs=%u", F2FS_OPTION(sbi).active_logs); 1928 if (test_opt(sbi, RESERVE_ROOT)) 1929 seq_printf(seq, ",reserve_root=%u,resuid=%u,resgid=%u", 1930 F2FS_OPTION(sbi).root_reserved_blocks, 1931 from_kuid_munged(&init_user_ns, 1932 F2FS_OPTION(sbi).s_resuid), 1933 from_kgid_munged(&init_user_ns, 1934 F2FS_OPTION(sbi).s_resgid)); 1935 if (F2FS_IO_SIZE_BITS(sbi)) 1936 seq_printf(seq, ",io_bits=%u", 1937 F2FS_OPTION(sbi).write_io_size_bits); 1938 #ifdef CONFIG_F2FS_FAULT_INJECTION 1939 if (test_opt(sbi, FAULT_INJECTION)) { 1940 seq_printf(seq, ",fault_injection=%u", 1941 F2FS_OPTION(sbi).fault_info.inject_rate); 1942 seq_printf(seq, ",fault_type=%u", 1943 F2FS_OPTION(sbi).fault_info.inject_type); 1944 } 1945 #endif 1946 #ifdef CONFIG_QUOTA 1947 if (test_opt(sbi, QUOTA)) 1948 seq_puts(seq, ",quota"); 1949 if (test_opt(sbi, USRQUOTA)) 1950 seq_puts(seq, ",usrquota"); 1951 if (test_opt(sbi, GRPQUOTA)) 1952 seq_puts(seq, ",grpquota"); 1953 if (test_opt(sbi, PRJQUOTA)) 1954 seq_puts(seq, ",prjquota"); 1955 #endif 1956 f2fs_show_quota_options(seq, sbi->sb); 1957 1958 fscrypt_show_test_dummy_encryption(seq, ',', sbi->sb); 1959 1960 if (sbi->sb->s_flags & SB_INLINECRYPT) 1961 seq_puts(seq, ",inlinecrypt"); 1962 1963 if (F2FS_OPTION(sbi).alloc_mode == ALLOC_MODE_DEFAULT) 1964 seq_printf(seq, ",alloc_mode=%s", "default"); 1965 else if (F2FS_OPTION(sbi).alloc_mode == ALLOC_MODE_REUSE) 1966 seq_printf(seq, ",alloc_mode=%s", "reuse"); 1967 1968 if (test_opt(sbi, DISABLE_CHECKPOINT)) 1969 seq_printf(seq, ",checkpoint=disable:%u", 1970 F2FS_OPTION(sbi).unusable_cap); 1971 if (test_opt(sbi, MERGE_CHECKPOINT)) 1972 seq_puts(seq, ",checkpoint_merge"); 1973 else 1974 seq_puts(seq, ",nocheckpoint_merge"); 1975 if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_POSIX) 1976 seq_printf(seq, ",fsync_mode=%s", "posix"); 1977 else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_STRICT) 1978 seq_printf(seq, ",fsync_mode=%s", "strict"); 1979 else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_NOBARRIER) 1980 seq_printf(seq, ",fsync_mode=%s", "nobarrier"); 1981 1982 #ifdef CONFIG_F2FS_FS_COMPRESSION 1983 f2fs_show_compress_options(seq, sbi->sb); 1984 #endif 1985 1986 if (test_opt(sbi, ATGC)) 1987 seq_puts(seq, ",atgc"); 1988 1989 if (F2FS_OPTION(sbi).discard_unit == DISCARD_UNIT_BLOCK) 1990 seq_printf(seq, ",discard_unit=%s", "block"); 1991 else if (F2FS_OPTION(sbi).discard_unit == DISCARD_UNIT_SEGMENT) 1992 seq_printf(seq, ",discard_unit=%s", "segment"); 1993 else if (F2FS_OPTION(sbi).discard_unit == DISCARD_UNIT_SECTION) 1994 seq_printf(seq, ",discard_unit=%s", "section"); 1995 1996 return 0; 1997 } 1998 1999 static void default_options(struct f2fs_sb_info *sbi) 2000 { 2001 /* init some FS parameters */ 2002 if (f2fs_sb_has_readonly(sbi)) 2003 F2FS_OPTION(sbi).active_logs = NR_CURSEG_RO_TYPE; 2004 else 2005 F2FS_OPTION(sbi).active_logs = NR_CURSEG_PERSIST_TYPE; 2006 2007 F2FS_OPTION(sbi).inline_xattr_size = DEFAULT_INLINE_XATTR_ADDRS; 2008 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_DEFAULT; 2009 F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_POSIX; 2010 F2FS_OPTION(sbi).s_resuid = make_kuid(&init_user_ns, F2FS_DEF_RESUID); 2011 F2FS_OPTION(sbi).s_resgid = make_kgid(&init_user_ns, F2FS_DEF_RESGID); 2012 F2FS_OPTION(sbi).compress_algorithm = COMPRESS_LZ4; 2013 F2FS_OPTION(sbi).compress_log_size = MIN_COMPRESS_LOG_SIZE; 2014 F2FS_OPTION(sbi).compress_ext_cnt = 0; 2015 F2FS_OPTION(sbi).compress_mode = COMPR_MODE_FS; 2016 F2FS_OPTION(sbi).bggc_mode = BGGC_MODE_ON; 2017 2018 sbi->sb->s_flags &= ~SB_INLINECRYPT; 2019 2020 set_opt(sbi, INLINE_XATTR); 2021 set_opt(sbi, INLINE_DATA); 2022 set_opt(sbi, INLINE_DENTRY); 2023 set_opt(sbi, EXTENT_CACHE); 2024 set_opt(sbi, NOHEAP); 2025 clear_opt(sbi, DISABLE_CHECKPOINT); 2026 set_opt(sbi, MERGE_CHECKPOINT); 2027 F2FS_OPTION(sbi).unusable_cap = 0; 2028 sbi->sb->s_flags |= SB_LAZYTIME; 2029 set_opt(sbi, FLUSH_MERGE); 2030 if (f2fs_hw_support_discard(sbi) || f2fs_hw_should_discard(sbi)) 2031 set_opt(sbi, DISCARD); 2032 if (f2fs_sb_has_blkzoned(sbi)) { 2033 F2FS_OPTION(sbi).fs_mode = FS_MODE_LFS; 2034 F2FS_OPTION(sbi).discard_unit = DISCARD_UNIT_SECTION; 2035 } else { 2036 F2FS_OPTION(sbi).fs_mode = FS_MODE_ADAPTIVE; 2037 F2FS_OPTION(sbi).discard_unit = DISCARD_UNIT_BLOCK; 2038 } 2039 2040 #ifdef CONFIG_F2FS_FS_XATTR 2041 set_opt(sbi, XATTR_USER); 2042 #endif 2043 #ifdef CONFIG_F2FS_FS_POSIX_ACL 2044 set_opt(sbi, POSIX_ACL); 2045 #endif 2046 2047 f2fs_build_fault_attr(sbi, 0, 0); 2048 } 2049 2050 #ifdef CONFIG_QUOTA 2051 static int f2fs_enable_quotas(struct super_block *sb); 2052 #endif 2053 2054 static int f2fs_disable_checkpoint(struct f2fs_sb_info *sbi) 2055 { 2056 unsigned int s_flags = sbi->sb->s_flags; 2057 struct cp_control cpc; 2058 unsigned int gc_mode = sbi->gc_mode; 2059 int err = 0; 2060 int ret; 2061 block_t unusable; 2062 2063 if (s_flags & SB_RDONLY) { 2064 f2fs_err(sbi, "checkpoint=disable on readonly fs"); 2065 return -EINVAL; 2066 } 2067 sbi->sb->s_flags |= SB_ACTIVE; 2068 2069 /* check if we need more GC first */ 2070 unusable = f2fs_get_unusable_blocks(sbi); 2071 if (!f2fs_disable_cp_again(sbi, unusable)) 2072 goto skip_gc; 2073 2074 f2fs_update_time(sbi, DISABLE_TIME); 2075 2076 sbi->gc_mode = GC_URGENT_HIGH; 2077 2078 while (!f2fs_time_over(sbi, DISABLE_TIME)) { 2079 struct f2fs_gc_control gc_control = { 2080 .victim_segno = NULL_SEGNO, 2081 .init_gc_type = FG_GC, 2082 .should_migrate_blocks = false, 2083 .err_gc_skipped = true, 2084 .nr_free_secs = 1 }; 2085 2086 f2fs_down_write(&sbi->gc_lock); 2087 err = f2fs_gc(sbi, &gc_control); 2088 if (err == -ENODATA) { 2089 err = 0; 2090 break; 2091 } 2092 if (err && err != -EAGAIN) 2093 break; 2094 } 2095 2096 ret = sync_filesystem(sbi->sb); 2097 if (ret || err) { 2098 err = ret ? ret : err; 2099 goto restore_flag; 2100 } 2101 2102 unusable = f2fs_get_unusable_blocks(sbi); 2103 if (f2fs_disable_cp_again(sbi, unusable)) { 2104 err = -EAGAIN; 2105 goto restore_flag; 2106 } 2107 2108 skip_gc: 2109 f2fs_down_write(&sbi->gc_lock); 2110 cpc.reason = CP_PAUSE; 2111 set_sbi_flag(sbi, SBI_CP_DISABLED); 2112 err = f2fs_write_checkpoint(sbi, &cpc); 2113 if (err) 2114 goto out_unlock; 2115 2116 spin_lock(&sbi->stat_lock); 2117 sbi->unusable_block_count = unusable; 2118 spin_unlock(&sbi->stat_lock); 2119 2120 out_unlock: 2121 f2fs_up_write(&sbi->gc_lock); 2122 restore_flag: 2123 sbi->gc_mode = gc_mode; 2124 sbi->sb->s_flags = s_flags; /* Restore SB_RDONLY status */ 2125 return err; 2126 } 2127 2128 static void f2fs_enable_checkpoint(struct f2fs_sb_info *sbi) 2129 { 2130 int retry = DEFAULT_RETRY_IO_COUNT; 2131 2132 /* we should flush all the data to keep data consistency */ 2133 do { 2134 sync_inodes_sb(sbi->sb); 2135 f2fs_io_schedule_timeout(DEFAULT_IO_TIMEOUT); 2136 } while (get_pages(sbi, F2FS_DIRTY_DATA) && retry--); 2137 2138 if (unlikely(retry < 0)) 2139 f2fs_warn(sbi, "checkpoint=enable has some unwritten data."); 2140 2141 f2fs_down_write(&sbi->gc_lock); 2142 f2fs_dirty_to_prefree(sbi); 2143 2144 clear_sbi_flag(sbi, SBI_CP_DISABLED); 2145 set_sbi_flag(sbi, SBI_IS_DIRTY); 2146 f2fs_up_write(&sbi->gc_lock); 2147 2148 f2fs_sync_fs(sbi->sb, 1); 2149 } 2150 2151 static int f2fs_remount(struct super_block *sb, int *flags, char *data) 2152 { 2153 struct f2fs_sb_info *sbi = F2FS_SB(sb); 2154 struct f2fs_mount_info org_mount_opt; 2155 unsigned long old_sb_flags; 2156 int err; 2157 bool need_restart_gc = false, need_stop_gc = false; 2158 bool need_restart_ckpt = false, need_stop_ckpt = false; 2159 bool need_restart_flush = false, need_stop_flush = false; 2160 bool need_restart_discard = false, need_stop_discard = false; 2161 bool no_extent_cache = !test_opt(sbi, EXTENT_CACHE); 2162 bool enable_checkpoint = !test_opt(sbi, DISABLE_CHECKPOINT); 2163 bool no_io_align = !F2FS_IO_ALIGNED(sbi); 2164 bool no_atgc = !test_opt(sbi, ATGC); 2165 bool no_discard = !test_opt(sbi, DISCARD); 2166 bool no_compress_cache = !test_opt(sbi, COMPRESS_CACHE); 2167 bool block_unit_discard = f2fs_block_unit_discard(sbi); 2168 struct discard_cmd_control *dcc; 2169 #ifdef CONFIG_QUOTA 2170 int i, j; 2171 #endif 2172 2173 /* 2174 * Save the old mount options in case we 2175 * need to restore them. 2176 */ 2177 org_mount_opt = sbi->mount_opt; 2178 old_sb_flags = sb->s_flags; 2179 2180 #ifdef CONFIG_QUOTA 2181 org_mount_opt.s_jquota_fmt = F2FS_OPTION(sbi).s_jquota_fmt; 2182 for (i = 0; i < MAXQUOTAS; i++) { 2183 if (F2FS_OPTION(sbi).s_qf_names[i]) { 2184 org_mount_opt.s_qf_names[i] = 2185 kstrdup(F2FS_OPTION(sbi).s_qf_names[i], 2186 GFP_KERNEL); 2187 if (!org_mount_opt.s_qf_names[i]) { 2188 for (j = 0; j < i; j++) 2189 kfree(org_mount_opt.s_qf_names[j]); 2190 return -ENOMEM; 2191 } 2192 } else { 2193 org_mount_opt.s_qf_names[i] = NULL; 2194 } 2195 } 2196 #endif 2197 2198 /* recover superblocks we couldn't write due to previous RO mount */ 2199 if (!(*flags & SB_RDONLY) && is_sbi_flag_set(sbi, SBI_NEED_SB_WRITE)) { 2200 err = f2fs_commit_super(sbi, false); 2201 f2fs_info(sbi, "Try to recover all the superblocks, ret: %d", 2202 err); 2203 if (!err) 2204 clear_sbi_flag(sbi, SBI_NEED_SB_WRITE); 2205 } 2206 2207 default_options(sbi); 2208 2209 /* parse mount options */ 2210 err = parse_options(sb, data, true); 2211 if (err) 2212 goto restore_opts; 2213 2214 /* 2215 * Previous and new state of filesystem is RO, 2216 * so skip checking GC and FLUSH_MERGE conditions. 2217 */ 2218 if (f2fs_readonly(sb) && (*flags & SB_RDONLY)) 2219 goto skip; 2220 2221 if (f2fs_sb_has_readonly(sbi) && !(*flags & SB_RDONLY)) { 2222 err = -EROFS; 2223 goto restore_opts; 2224 } 2225 2226 #ifdef CONFIG_QUOTA 2227 if (!f2fs_readonly(sb) && (*flags & SB_RDONLY)) { 2228 err = dquot_suspend(sb, -1); 2229 if (err < 0) 2230 goto restore_opts; 2231 } else if (f2fs_readonly(sb) && !(*flags & SB_RDONLY)) { 2232 /* dquot_resume needs RW */ 2233 sb->s_flags &= ~SB_RDONLY; 2234 if (sb_any_quota_suspended(sb)) { 2235 dquot_resume(sb, -1); 2236 } else if (f2fs_sb_has_quota_ino(sbi)) { 2237 err = f2fs_enable_quotas(sb); 2238 if (err) 2239 goto restore_opts; 2240 } 2241 } 2242 #endif 2243 /* disallow enable atgc dynamically */ 2244 if (no_atgc == !!test_opt(sbi, ATGC)) { 2245 err = -EINVAL; 2246 f2fs_warn(sbi, "switch atgc option is not allowed"); 2247 goto restore_opts; 2248 } 2249 2250 /* disallow enable/disable extent_cache dynamically */ 2251 if (no_extent_cache == !!test_opt(sbi, EXTENT_CACHE)) { 2252 err = -EINVAL; 2253 f2fs_warn(sbi, "switch extent_cache option is not allowed"); 2254 goto restore_opts; 2255 } 2256 2257 if (no_io_align == !!F2FS_IO_ALIGNED(sbi)) { 2258 err = -EINVAL; 2259 f2fs_warn(sbi, "switch io_bits option is not allowed"); 2260 goto restore_opts; 2261 } 2262 2263 if (no_compress_cache == !!test_opt(sbi, COMPRESS_CACHE)) { 2264 err = -EINVAL; 2265 f2fs_warn(sbi, "switch compress_cache option is not allowed"); 2266 goto restore_opts; 2267 } 2268 2269 if (block_unit_discard != f2fs_block_unit_discard(sbi)) { 2270 err = -EINVAL; 2271 f2fs_warn(sbi, "switch discard_unit option is not allowed"); 2272 goto restore_opts; 2273 } 2274 2275 if ((*flags & SB_RDONLY) && test_opt(sbi, DISABLE_CHECKPOINT)) { 2276 err = -EINVAL; 2277 f2fs_warn(sbi, "disabling checkpoint not compatible with read-only"); 2278 goto restore_opts; 2279 } 2280 2281 /* 2282 * We stop the GC thread if FS is mounted as RO 2283 * or if background_gc = off is passed in mount 2284 * option. Also sync the filesystem. 2285 */ 2286 if ((*flags & SB_RDONLY) || 2287 (F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_OFF && 2288 !test_opt(sbi, GC_MERGE))) { 2289 if (sbi->gc_thread) { 2290 f2fs_stop_gc_thread(sbi); 2291 need_restart_gc = true; 2292 } 2293 } else if (!sbi->gc_thread) { 2294 err = f2fs_start_gc_thread(sbi); 2295 if (err) 2296 goto restore_opts; 2297 need_stop_gc = true; 2298 } 2299 2300 if (*flags & SB_RDONLY) { 2301 sync_inodes_sb(sb); 2302 2303 set_sbi_flag(sbi, SBI_IS_DIRTY); 2304 set_sbi_flag(sbi, SBI_IS_CLOSE); 2305 f2fs_sync_fs(sb, 1); 2306 clear_sbi_flag(sbi, SBI_IS_CLOSE); 2307 } 2308 2309 if ((*flags & SB_RDONLY) || test_opt(sbi, DISABLE_CHECKPOINT) || 2310 !test_opt(sbi, MERGE_CHECKPOINT)) { 2311 f2fs_stop_ckpt_thread(sbi); 2312 need_restart_ckpt = true; 2313 } else { 2314 err = f2fs_start_ckpt_thread(sbi); 2315 if (err) { 2316 f2fs_err(sbi, 2317 "Failed to start F2FS issue_checkpoint_thread (%d)", 2318 err); 2319 goto restore_gc; 2320 } 2321 need_stop_ckpt = true; 2322 } 2323 2324 /* 2325 * We stop issue flush thread if FS is mounted as RO 2326 * or if flush_merge is not passed in mount option. 2327 */ 2328 if ((*flags & SB_RDONLY) || !test_opt(sbi, FLUSH_MERGE)) { 2329 clear_opt(sbi, FLUSH_MERGE); 2330 f2fs_destroy_flush_cmd_control(sbi, false); 2331 need_restart_flush = true; 2332 } else { 2333 err = f2fs_create_flush_cmd_control(sbi); 2334 if (err) 2335 goto restore_ckpt; 2336 need_stop_flush = true; 2337 } 2338 2339 if (no_discard == !!test_opt(sbi, DISCARD)) { 2340 if (test_opt(sbi, DISCARD)) { 2341 err = f2fs_start_discard_thread(sbi); 2342 if (err) 2343 goto restore_flush; 2344 need_stop_discard = true; 2345 } else { 2346 dcc = SM_I(sbi)->dcc_info; 2347 f2fs_stop_discard_thread(sbi); 2348 if (atomic_read(&dcc->discard_cmd_cnt)) 2349 f2fs_issue_discard_timeout(sbi); 2350 need_restart_discard = true; 2351 } 2352 } 2353 2354 if (enable_checkpoint == !!test_opt(sbi, DISABLE_CHECKPOINT)) { 2355 if (test_opt(sbi, DISABLE_CHECKPOINT)) { 2356 err = f2fs_disable_checkpoint(sbi); 2357 if (err) 2358 goto restore_discard; 2359 } else { 2360 f2fs_enable_checkpoint(sbi); 2361 } 2362 } 2363 2364 skip: 2365 #ifdef CONFIG_QUOTA 2366 /* Release old quota file names */ 2367 for (i = 0; i < MAXQUOTAS; i++) 2368 kfree(org_mount_opt.s_qf_names[i]); 2369 #endif 2370 /* Update the POSIXACL Flag */ 2371 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) | 2372 (test_opt(sbi, POSIX_ACL) ? SB_POSIXACL : 0); 2373 2374 limit_reserve_root(sbi); 2375 adjust_unusable_cap_perc(sbi); 2376 *flags = (*flags & ~SB_LAZYTIME) | (sb->s_flags & SB_LAZYTIME); 2377 return 0; 2378 restore_discard: 2379 if (need_restart_discard) { 2380 if (f2fs_start_discard_thread(sbi)) 2381 f2fs_warn(sbi, "discard has been stopped"); 2382 } else if (need_stop_discard) { 2383 f2fs_stop_discard_thread(sbi); 2384 } 2385 restore_flush: 2386 if (need_restart_flush) { 2387 if (f2fs_create_flush_cmd_control(sbi)) 2388 f2fs_warn(sbi, "background flush thread has stopped"); 2389 } else if (need_stop_flush) { 2390 clear_opt(sbi, FLUSH_MERGE); 2391 f2fs_destroy_flush_cmd_control(sbi, false); 2392 } 2393 restore_ckpt: 2394 if (need_restart_ckpt) { 2395 if (f2fs_start_ckpt_thread(sbi)) 2396 f2fs_warn(sbi, "background ckpt thread has stopped"); 2397 } else if (need_stop_ckpt) { 2398 f2fs_stop_ckpt_thread(sbi); 2399 } 2400 restore_gc: 2401 if (need_restart_gc) { 2402 if (f2fs_start_gc_thread(sbi)) 2403 f2fs_warn(sbi, "background gc thread has stopped"); 2404 } else if (need_stop_gc) { 2405 f2fs_stop_gc_thread(sbi); 2406 } 2407 restore_opts: 2408 #ifdef CONFIG_QUOTA 2409 F2FS_OPTION(sbi).s_jquota_fmt = org_mount_opt.s_jquota_fmt; 2410 for (i = 0; i < MAXQUOTAS; i++) { 2411 kfree(F2FS_OPTION(sbi).s_qf_names[i]); 2412 F2FS_OPTION(sbi).s_qf_names[i] = org_mount_opt.s_qf_names[i]; 2413 } 2414 #endif 2415 sbi->mount_opt = org_mount_opt; 2416 sb->s_flags = old_sb_flags; 2417 return err; 2418 } 2419 2420 #ifdef CONFIG_QUOTA 2421 /* Read data from quotafile */ 2422 static ssize_t f2fs_quota_read(struct super_block *sb, int type, char *data, 2423 size_t len, loff_t off) 2424 { 2425 struct inode *inode = sb_dqopt(sb)->files[type]; 2426 struct address_space *mapping = inode->i_mapping; 2427 block_t blkidx = F2FS_BYTES_TO_BLK(off); 2428 int offset = off & (sb->s_blocksize - 1); 2429 int tocopy; 2430 size_t toread; 2431 loff_t i_size = i_size_read(inode); 2432 struct page *page; 2433 char *kaddr; 2434 2435 if (off > i_size) 2436 return 0; 2437 2438 if (off + len > i_size) 2439 len = i_size - off; 2440 toread = len; 2441 while (toread > 0) { 2442 tocopy = min_t(unsigned long, sb->s_blocksize - offset, toread); 2443 repeat: 2444 page = read_cache_page_gfp(mapping, blkidx, GFP_NOFS); 2445 if (IS_ERR(page)) { 2446 if (PTR_ERR(page) == -ENOMEM) { 2447 memalloc_retry_wait(GFP_NOFS); 2448 goto repeat; 2449 } 2450 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR); 2451 return PTR_ERR(page); 2452 } 2453 2454 lock_page(page); 2455 2456 if (unlikely(page->mapping != mapping)) { 2457 f2fs_put_page(page, 1); 2458 goto repeat; 2459 } 2460 if (unlikely(!PageUptodate(page))) { 2461 f2fs_put_page(page, 1); 2462 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR); 2463 return -EIO; 2464 } 2465 2466 kaddr = kmap_atomic(page); 2467 memcpy(data, kaddr + offset, tocopy); 2468 kunmap_atomic(kaddr); 2469 f2fs_put_page(page, 1); 2470 2471 offset = 0; 2472 toread -= tocopy; 2473 data += tocopy; 2474 blkidx++; 2475 } 2476 return len; 2477 } 2478 2479 /* Write to quotafile */ 2480 static ssize_t f2fs_quota_write(struct super_block *sb, int type, 2481 const char *data, size_t len, loff_t off) 2482 { 2483 struct inode *inode = sb_dqopt(sb)->files[type]; 2484 struct address_space *mapping = inode->i_mapping; 2485 const struct address_space_operations *a_ops = mapping->a_ops; 2486 int offset = off & (sb->s_blocksize - 1); 2487 size_t towrite = len; 2488 struct page *page; 2489 void *fsdata = NULL; 2490 char *kaddr; 2491 int err = 0; 2492 int tocopy; 2493 2494 while (towrite > 0) { 2495 tocopy = min_t(unsigned long, sb->s_blocksize - offset, 2496 towrite); 2497 retry: 2498 err = a_ops->write_begin(NULL, mapping, off, tocopy, 2499 &page, &fsdata); 2500 if (unlikely(err)) { 2501 if (err == -ENOMEM) { 2502 f2fs_io_schedule_timeout(DEFAULT_IO_TIMEOUT); 2503 goto retry; 2504 } 2505 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR); 2506 break; 2507 } 2508 2509 kaddr = kmap_atomic(page); 2510 memcpy(kaddr + offset, data, tocopy); 2511 kunmap_atomic(kaddr); 2512 flush_dcache_page(page); 2513 2514 a_ops->write_end(NULL, mapping, off, tocopy, tocopy, 2515 page, fsdata); 2516 offset = 0; 2517 towrite -= tocopy; 2518 off += tocopy; 2519 data += tocopy; 2520 cond_resched(); 2521 } 2522 2523 if (len == towrite) 2524 return err; 2525 inode->i_mtime = inode->i_ctime = current_time(inode); 2526 f2fs_mark_inode_dirty_sync(inode, false); 2527 return len - towrite; 2528 } 2529 2530 int f2fs_dquot_initialize(struct inode *inode) 2531 { 2532 if (time_to_inject(F2FS_I_SB(inode), FAULT_DQUOT_INIT)) { 2533 f2fs_show_injection_info(F2FS_I_SB(inode), FAULT_DQUOT_INIT); 2534 return -ESRCH; 2535 } 2536 2537 return dquot_initialize(inode); 2538 } 2539 2540 static struct dquot **f2fs_get_dquots(struct inode *inode) 2541 { 2542 return F2FS_I(inode)->i_dquot; 2543 } 2544 2545 static qsize_t *f2fs_get_reserved_space(struct inode *inode) 2546 { 2547 return &F2FS_I(inode)->i_reserved_quota; 2548 } 2549 2550 static int f2fs_quota_on_mount(struct f2fs_sb_info *sbi, int type) 2551 { 2552 if (is_set_ckpt_flags(sbi, CP_QUOTA_NEED_FSCK_FLAG)) { 2553 f2fs_err(sbi, "quota sysfile may be corrupted, skip loading it"); 2554 return 0; 2555 } 2556 2557 return dquot_quota_on_mount(sbi->sb, F2FS_OPTION(sbi).s_qf_names[type], 2558 F2FS_OPTION(sbi).s_jquota_fmt, type); 2559 } 2560 2561 int f2fs_enable_quota_files(struct f2fs_sb_info *sbi, bool rdonly) 2562 { 2563 int enabled = 0; 2564 int i, err; 2565 2566 if (f2fs_sb_has_quota_ino(sbi) && rdonly) { 2567 err = f2fs_enable_quotas(sbi->sb); 2568 if (err) { 2569 f2fs_err(sbi, "Cannot turn on quota_ino: %d", err); 2570 return 0; 2571 } 2572 return 1; 2573 } 2574 2575 for (i = 0; i < MAXQUOTAS; i++) { 2576 if (F2FS_OPTION(sbi).s_qf_names[i]) { 2577 err = f2fs_quota_on_mount(sbi, i); 2578 if (!err) { 2579 enabled = 1; 2580 continue; 2581 } 2582 f2fs_err(sbi, "Cannot turn on quotas: %d on %d", 2583 err, i); 2584 } 2585 } 2586 return enabled; 2587 } 2588 2589 static int f2fs_quota_enable(struct super_block *sb, int type, int format_id, 2590 unsigned int flags) 2591 { 2592 struct inode *qf_inode; 2593 unsigned long qf_inum; 2594 int err; 2595 2596 BUG_ON(!f2fs_sb_has_quota_ino(F2FS_SB(sb))); 2597 2598 qf_inum = f2fs_qf_ino(sb, type); 2599 if (!qf_inum) 2600 return -EPERM; 2601 2602 qf_inode = f2fs_iget(sb, qf_inum); 2603 if (IS_ERR(qf_inode)) { 2604 f2fs_err(F2FS_SB(sb), "Bad quota inode %u:%lu", type, qf_inum); 2605 return PTR_ERR(qf_inode); 2606 } 2607 2608 /* Don't account quota for quota files to avoid recursion */ 2609 qf_inode->i_flags |= S_NOQUOTA; 2610 err = dquot_load_quota_inode(qf_inode, type, format_id, flags); 2611 iput(qf_inode); 2612 return err; 2613 } 2614 2615 static int f2fs_enable_quotas(struct super_block *sb) 2616 { 2617 struct f2fs_sb_info *sbi = F2FS_SB(sb); 2618 int type, err = 0; 2619 unsigned long qf_inum; 2620 bool quota_mopt[MAXQUOTAS] = { 2621 test_opt(sbi, USRQUOTA), 2622 test_opt(sbi, GRPQUOTA), 2623 test_opt(sbi, PRJQUOTA), 2624 }; 2625 2626 if (is_set_ckpt_flags(F2FS_SB(sb), CP_QUOTA_NEED_FSCK_FLAG)) { 2627 f2fs_err(sbi, "quota file may be corrupted, skip loading it"); 2628 return 0; 2629 } 2630 2631 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE; 2632 2633 for (type = 0; type < MAXQUOTAS; type++) { 2634 qf_inum = f2fs_qf_ino(sb, type); 2635 if (qf_inum) { 2636 err = f2fs_quota_enable(sb, type, QFMT_VFS_V1, 2637 DQUOT_USAGE_ENABLED | 2638 (quota_mopt[type] ? DQUOT_LIMITS_ENABLED : 0)); 2639 if (err) { 2640 f2fs_err(sbi, "Failed to enable quota tracking (type=%d, err=%d). Please run fsck to fix.", 2641 type, err); 2642 for (type--; type >= 0; type--) 2643 dquot_quota_off(sb, type); 2644 set_sbi_flag(F2FS_SB(sb), 2645 SBI_QUOTA_NEED_REPAIR); 2646 return err; 2647 } 2648 } 2649 } 2650 return 0; 2651 } 2652 2653 static int f2fs_quota_sync_file(struct f2fs_sb_info *sbi, int type) 2654 { 2655 struct quota_info *dqopt = sb_dqopt(sbi->sb); 2656 struct address_space *mapping = dqopt->files[type]->i_mapping; 2657 int ret = 0; 2658 2659 ret = dquot_writeback_dquots(sbi->sb, type); 2660 if (ret) 2661 goto out; 2662 2663 ret = filemap_fdatawrite(mapping); 2664 if (ret) 2665 goto out; 2666 2667 /* if we are using journalled quota */ 2668 if (is_journalled_quota(sbi)) 2669 goto out; 2670 2671 ret = filemap_fdatawait(mapping); 2672 2673 truncate_inode_pages(&dqopt->files[type]->i_data, 0); 2674 out: 2675 if (ret) 2676 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR); 2677 return ret; 2678 } 2679 2680 int f2fs_quota_sync(struct super_block *sb, int type) 2681 { 2682 struct f2fs_sb_info *sbi = F2FS_SB(sb); 2683 struct quota_info *dqopt = sb_dqopt(sb); 2684 int cnt; 2685 int ret = 0; 2686 2687 /* 2688 * Now when everything is written we can discard the pagecache so 2689 * that userspace sees the changes. 2690 */ 2691 for (cnt = 0; cnt < MAXQUOTAS; cnt++) { 2692 2693 if (type != -1 && cnt != type) 2694 continue; 2695 2696 if (!sb_has_quota_active(sb, cnt)) 2697 continue; 2698 2699 if (!f2fs_sb_has_quota_ino(sbi)) 2700 inode_lock(dqopt->files[cnt]); 2701 2702 /* 2703 * do_quotactl 2704 * f2fs_quota_sync 2705 * f2fs_down_read(quota_sem) 2706 * dquot_writeback_dquots() 2707 * f2fs_dquot_commit 2708 * block_operation 2709 * f2fs_down_read(quota_sem) 2710 */ 2711 f2fs_lock_op(sbi); 2712 f2fs_down_read(&sbi->quota_sem); 2713 2714 ret = f2fs_quota_sync_file(sbi, cnt); 2715 2716 f2fs_up_read(&sbi->quota_sem); 2717 f2fs_unlock_op(sbi); 2718 2719 if (!f2fs_sb_has_quota_ino(sbi)) 2720 inode_unlock(dqopt->files[cnt]); 2721 2722 if (ret) 2723 break; 2724 } 2725 return ret; 2726 } 2727 2728 static int f2fs_quota_on(struct super_block *sb, int type, int format_id, 2729 const struct path *path) 2730 { 2731 struct inode *inode; 2732 int err; 2733 2734 /* if quota sysfile exists, deny enabling quota with specific file */ 2735 if (f2fs_sb_has_quota_ino(F2FS_SB(sb))) { 2736 f2fs_err(F2FS_SB(sb), "quota sysfile already exists"); 2737 return -EBUSY; 2738 } 2739 2740 err = f2fs_quota_sync(sb, type); 2741 if (err) 2742 return err; 2743 2744 err = dquot_quota_on(sb, type, format_id, path); 2745 if (err) 2746 return err; 2747 2748 inode = d_inode(path->dentry); 2749 2750 inode_lock(inode); 2751 F2FS_I(inode)->i_flags |= F2FS_NOATIME_FL | F2FS_IMMUTABLE_FL; 2752 f2fs_set_inode_flags(inode); 2753 inode_unlock(inode); 2754 f2fs_mark_inode_dirty_sync(inode, false); 2755 2756 return 0; 2757 } 2758 2759 static int __f2fs_quota_off(struct super_block *sb, int type) 2760 { 2761 struct inode *inode = sb_dqopt(sb)->files[type]; 2762 int err; 2763 2764 if (!inode || !igrab(inode)) 2765 return dquot_quota_off(sb, type); 2766 2767 err = f2fs_quota_sync(sb, type); 2768 if (err) 2769 goto out_put; 2770 2771 err = dquot_quota_off(sb, type); 2772 if (err || f2fs_sb_has_quota_ino(F2FS_SB(sb))) 2773 goto out_put; 2774 2775 inode_lock(inode); 2776 F2FS_I(inode)->i_flags &= ~(F2FS_NOATIME_FL | F2FS_IMMUTABLE_FL); 2777 f2fs_set_inode_flags(inode); 2778 inode_unlock(inode); 2779 f2fs_mark_inode_dirty_sync(inode, false); 2780 out_put: 2781 iput(inode); 2782 return err; 2783 } 2784 2785 static int f2fs_quota_off(struct super_block *sb, int type) 2786 { 2787 struct f2fs_sb_info *sbi = F2FS_SB(sb); 2788 int err; 2789 2790 err = __f2fs_quota_off(sb, type); 2791 2792 /* 2793 * quotactl can shutdown journalled quota, result in inconsistence 2794 * between quota record and fs data by following updates, tag the 2795 * flag to let fsck be aware of it. 2796 */ 2797 if (is_journalled_quota(sbi)) 2798 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR); 2799 return err; 2800 } 2801 2802 void f2fs_quota_off_umount(struct super_block *sb) 2803 { 2804 int type; 2805 int err; 2806 2807 for (type = 0; type < MAXQUOTAS; type++) { 2808 err = __f2fs_quota_off(sb, type); 2809 if (err) { 2810 int ret = dquot_quota_off(sb, type); 2811 2812 f2fs_err(F2FS_SB(sb), "Fail to turn off disk quota (type: %d, err: %d, ret:%d), Please run fsck to fix it.", 2813 type, err, ret); 2814 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR); 2815 } 2816 } 2817 /* 2818 * In case of checkpoint=disable, we must flush quota blocks. 2819 * This can cause NULL exception for node_inode in end_io, since 2820 * put_super already dropped it. 2821 */ 2822 sync_filesystem(sb); 2823 } 2824 2825 static void f2fs_truncate_quota_inode_pages(struct super_block *sb) 2826 { 2827 struct quota_info *dqopt = sb_dqopt(sb); 2828 int type; 2829 2830 for (type = 0; type < MAXQUOTAS; type++) { 2831 if (!dqopt->files[type]) 2832 continue; 2833 f2fs_inode_synced(dqopt->files[type]); 2834 } 2835 } 2836 2837 static int f2fs_dquot_commit(struct dquot *dquot) 2838 { 2839 struct f2fs_sb_info *sbi = F2FS_SB(dquot->dq_sb); 2840 int ret; 2841 2842 f2fs_down_read_nested(&sbi->quota_sem, SINGLE_DEPTH_NESTING); 2843 ret = dquot_commit(dquot); 2844 if (ret < 0) 2845 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR); 2846 f2fs_up_read(&sbi->quota_sem); 2847 return ret; 2848 } 2849 2850 static int f2fs_dquot_acquire(struct dquot *dquot) 2851 { 2852 struct f2fs_sb_info *sbi = F2FS_SB(dquot->dq_sb); 2853 int ret; 2854 2855 f2fs_down_read(&sbi->quota_sem); 2856 ret = dquot_acquire(dquot); 2857 if (ret < 0) 2858 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR); 2859 f2fs_up_read(&sbi->quota_sem); 2860 return ret; 2861 } 2862 2863 static int f2fs_dquot_release(struct dquot *dquot) 2864 { 2865 struct f2fs_sb_info *sbi = F2FS_SB(dquot->dq_sb); 2866 int ret = dquot_release(dquot); 2867 2868 if (ret < 0) 2869 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR); 2870 return ret; 2871 } 2872 2873 static int f2fs_dquot_mark_dquot_dirty(struct dquot *dquot) 2874 { 2875 struct super_block *sb = dquot->dq_sb; 2876 struct f2fs_sb_info *sbi = F2FS_SB(sb); 2877 int ret = dquot_mark_dquot_dirty(dquot); 2878 2879 /* if we are using journalled quota */ 2880 if (is_journalled_quota(sbi)) 2881 set_sbi_flag(sbi, SBI_QUOTA_NEED_FLUSH); 2882 2883 return ret; 2884 } 2885 2886 static int f2fs_dquot_commit_info(struct super_block *sb, int type) 2887 { 2888 struct f2fs_sb_info *sbi = F2FS_SB(sb); 2889 int ret = dquot_commit_info(sb, type); 2890 2891 if (ret < 0) 2892 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR); 2893 return ret; 2894 } 2895 2896 static int f2fs_get_projid(struct inode *inode, kprojid_t *projid) 2897 { 2898 *projid = F2FS_I(inode)->i_projid; 2899 return 0; 2900 } 2901 2902 static const struct dquot_operations f2fs_quota_operations = { 2903 .get_reserved_space = f2fs_get_reserved_space, 2904 .write_dquot = f2fs_dquot_commit, 2905 .acquire_dquot = f2fs_dquot_acquire, 2906 .release_dquot = f2fs_dquot_release, 2907 .mark_dirty = f2fs_dquot_mark_dquot_dirty, 2908 .write_info = f2fs_dquot_commit_info, 2909 .alloc_dquot = dquot_alloc, 2910 .destroy_dquot = dquot_destroy, 2911 .get_projid = f2fs_get_projid, 2912 .get_next_id = dquot_get_next_id, 2913 }; 2914 2915 static const struct quotactl_ops f2fs_quotactl_ops = { 2916 .quota_on = f2fs_quota_on, 2917 .quota_off = f2fs_quota_off, 2918 .quota_sync = f2fs_quota_sync, 2919 .get_state = dquot_get_state, 2920 .set_info = dquot_set_dqinfo, 2921 .get_dqblk = dquot_get_dqblk, 2922 .set_dqblk = dquot_set_dqblk, 2923 .get_nextdqblk = dquot_get_next_dqblk, 2924 }; 2925 #else 2926 int f2fs_dquot_initialize(struct inode *inode) 2927 { 2928 return 0; 2929 } 2930 2931 int f2fs_quota_sync(struct super_block *sb, int type) 2932 { 2933 return 0; 2934 } 2935 2936 void f2fs_quota_off_umount(struct super_block *sb) 2937 { 2938 } 2939 #endif 2940 2941 static const struct super_operations f2fs_sops = { 2942 .alloc_inode = f2fs_alloc_inode, 2943 .free_inode = f2fs_free_inode, 2944 .drop_inode = f2fs_drop_inode, 2945 .write_inode = f2fs_write_inode, 2946 .dirty_inode = f2fs_dirty_inode, 2947 .show_options = f2fs_show_options, 2948 #ifdef CONFIG_QUOTA 2949 .quota_read = f2fs_quota_read, 2950 .quota_write = f2fs_quota_write, 2951 .get_dquots = f2fs_get_dquots, 2952 #endif 2953 .evict_inode = f2fs_evict_inode, 2954 .put_super = f2fs_put_super, 2955 .sync_fs = f2fs_sync_fs, 2956 .freeze_fs = f2fs_freeze, 2957 .unfreeze_fs = f2fs_unfreeze, 2958 .statfs = f2fs_statfs, 2959 .remount_fs = f2fs_remount, 2960 }; 2961 2962 #ifdef CONFIG_FS_ENCRYPTION 2963 static int f2fs_get_context(struct inode *inode, void *ctx, size_t len) 2964 { 2965 return f2fs_getxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION, 2966 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT, 2967 ctx, len, NULL); 2968 } 2969 2970 static int f2fs_set_context(struct inode *inode, const void *ctx, size_t len, 2971 void *fs_data) 2972 { 2973 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 2974 2975 /* 2976 * Encrypting the root directory is not allowed because fsck 2977 * expects lost+found directory to exist and remain unencrypted 2978 * if LOST_FOUND feature is enabled. 2979 * 2980 */ 2981 if (f2fs_sb_has_lost_found(sbi) && 2982 inode->i_ino == F2FS_ROOT_INO(sbi)) 2983 return -EPERM; 2984 2985 return f2fs_setxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION, 2986 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT, 2987 ctx, len, fs_data, XATTR_CREATE); 2988 } 2989 2990 static const union fscrypt_policy *f2fs_get_dummy_policy(struct super_block *sb) 2991 { 2992 return F2FS_OPTION(F2FS_SB(sb)).dummy_enc_policy.policy; 2993 } 2994 2995 static bool f2fs_has_stable_inodes(struct super_block *sb) 2996 { 2997 return true; 2998 } 2999 3000 static void f2fs_get_ino_and_lblk_bits(struct super_block *sb, 3001 int *ino_bits_ret, int *lblk_bits_ret) 3002 { 3003 *ino_bits_ret = 8 * sizeof(nid_t); 3004 *lblk_bits_ret = 8 * sizeof(block_t); 3005 } 3006 3007 static int f2fs_get_num_devices(struct super_block *sb) 3008 { 3009 struct f2fs_sb_info *sbi = F2FS_SB(sb); 3010 3011 if (f2fs_is_multi_device(sbi)) 3012 return sbi->s_ndevs; 3013 return 1; 3014 } 3015 3016 static void f2fs_get_devices(struct super_block *sb, 3017 struct request_queue **devs) 3018 { 3019 struct f2fs_sb_info *sbi = F2FS_SB(sb); 3020 int i; 3021 3022 for (i = 0; i < sbi->s_ndevs; i++) 3023 devs[i] = bdev_get_queue(FDEV(i).bdev); 3024 } 3025 3026 static const struct fscrypt_operations f2fs_cryptops = { 3027 .key_prefix = "f2fs:", 3028 .get_context = f2fs_get_context, 3029 .set_context = f2fs_set_context, 3030 .get_dummy_policy = f2fs_get_dummy_policy, 3031 .empty_dir = f2fs_empty_dir, 3032 .has_stable_inodes = f2fs_has_stable_inodes, 3033 .get_ino_and_lblk_bits = f2fs_get_ino_and_lblk_bits, 3034 .get_num_devices = f2fs_get_num_devices, 3035 .get_devices = f2fs_get_devices, 3036 }; 3037 #endif 3038 3039 static struct inode *f2fs_nfs_get_inode(struct super_block *sb, 3040 u64 ino, u32 generation) 3041 { 3042 struct f2fs_sb_info *sbi = F2FS_SB(sb); 3043 struct inode *inode; 3044 3045 if (f2fs_check_nid_range(sbi, ino)) 3046 return ERR_PTR(-ESTALE); 3047 3048 /* 3049 * f2fs_iget isn't quite right if the inode is currently unallocated! 3050 * However f2fs_iget currently does appropriate checks to handle stale 3051 * inodes so everything is OK. 3052 */ 3053 inode = f2fs_iget(sb, ino); 3054 if (IS_ERR(inode)) 3055 return ERR_CAST(inode); 3056 if (unlikely(generation && inode->i_generation != generation)) { 3057 /* we didn't find the right inode.. */ 3058 iput(inode); 3059 return ERR_PTR(-ESTALE); 3060 } 3061 return inode; 3062 } 3063 3064 static struct dentry *f2fs_fh_to_dentry(struct super_block *sb, struct fid *fid, 3065 int fh_len, int fh_type) 3066 { 3067 return generic_fh_to_dentry(sb, fid, fh_len, fh_type, 3068 f2fs_nfs_get_inode); 3069 } 3070 3071 static struct dentry *f2fs_fh_to_parent(struct super_block *sb, struct fid *fid, 3072 int fh_len, int fh_type) 3073 { 3074 return generic_fh_to_parent(sb, fid, fh_len, fh_type, 3075 f2fs_nfs_get_inode); 3076 } 3077 3078 static const struct export_operations f2fs_export_ops = { 3079 .fh_to_dentry = f2fs_fh_to_dentry, 3080 .fh_to_parent = f2fs_fh_to_parent, 3081 .get_parent = f2fs_get_parent, 3082 }; 3083 3084 loff_t max_file_blocks(struct inode *inode) 3085 { 3086 loff_t result = 0; 3087 loff_t leaf_count; 3088 3089 /* 3090 * note: previously, result is equal to (DEF_ADDRS_PER_INODE - 3091 * DEFAULT_INLINE_XATTR_ADDRS), but now f2fs try to reserve more 3092 * space in inode.i_addr, it will be more safe to reassign 3093 * result as zero. 3094 */ 3095 3096 if (inode && f2fs_compressed_file(inode)) 3097 leaf_count = ADDRS_PER_BLOCK(inode); 3098 else 3099 leaf_count = DEF_ADDRS_PER_BLOCK; 3100 3101 /* two direct node blocks */ 3102 result += (leaf_count * 2); 3103 3104 /* two indirect node blocks */ 3105 leaf_count *= NIDS_PER_BLOCK; 3106 result += (leaf_count * 2); 3107 3108 /* one double indirect node block */ 3109 leaf_count *= NIDS_PER_BLOCK; 3110 result += leaf_count; 3111 3112 return result; 3113 } 3114 3115 static int __f2fs_commit_super(struct buffer_head *bh, 3116 struct f2fs_super_block *super) 3117 { 3118 lock_buffer(bh); 3119 if (super) 3120 memcpy(bh->b_data + F2FS_SUPER_OFFSET, super, sizeof(*super)); 3121 set_buffer_dirty(bh); 3122 unlock_buffer(bh); 3123 3124 /* it's rare case, we can do fua all the time */ 3125 return __sync_dirty_buffer(bh, REQ_SYNC | REQ_PREFLUSH | REQ_FUA); 3126 } 3127 3128 static inline bool sanity_check_area_boundary(struct f2fs_sb_info *sbi, 3129 struct buffer_head *bh) 3130 { 3131 struct f2fs_super_block *raw_super = (struct f2fs_super_block *) 3132 (bh->b_data + F2FS_SUPER_OFFSET); 3133 struct super_block *sb = sbi->sb; 3134 u32 segment0_blkaddr = le32_to_cpu(raw_super->segment0_blkaddr); 3135 u32 cp_blkaddr = le32_to_cpu(raw_super->cp_blkaddr); 3136 u32 sit_blkaddr = le32_to_cpu(raw_super->sit_blkaddr); 3137 u32 nat_blkaddr = le32_to_cpu(raw_super->nat_blkaddr); 3138 u32 ssa_blkaddr = le32_to_cpu(raw_super->ssa_blkaddr); 3139 u32 main_blkaddr = le32_to_cpu(raw_super->main_blkaddr); 3140 u32 segment_count_ckpt = le32_to_cpu(raw_super->segment_count_ckpt); 3141 u32 segment_count_sit = le32_to_cpu(raw_super->segment_count_sit); 3142 u32 segment_count_nat = le32_to_cpu(raw_super->segment_count_nat); 3143 u32 segment_count_ssa = le32_to_cpu(raw_super->segment_count_ssa); 3144 u32 segment_count_main = le32_to_cpu(raw_super->segment_count_main); 3145 u32 segment_count = le32_to_cpu(raw_super->segment_count); 3146 u32 log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg); 3147 u64 main_end_blkaddr = main_blkaddr + 3148 (segment_count_main << log_blocks_per_seg); 3149 u64 seg_end_blkaddr = segment0_blkaddr + 3150 (segment_count << log_blocks_per_seg); 3151 3152 if (segment0_blkaddr != cp_blkaddr) { 3153 f2fs_info(sbi, "Mismatch start address, segment0(%u) cp_blkaddr(%u)", 3154 segment0_blkaddr, cp_blkaddr); 3155 return true; 3156 } 3157 3158 if (cp_blkaddr + (segment_count_ckpt << log_blocks_per_seg) != 3159 sit_blkaddr) { 3160 f2fs_info(sbi, "Wrong CP boundary, start(%u) end(%u) blocks(%u)", 3161 cp_blkaddr, sit_blkaddr, 3162 segment_count_ckpt << log_blocks_per_seg); 3163 return true; 3164 } 3165 3166 if (sit_blkaddr + (segment_count_sit << log_blocks_per_seg) != 3167 nat_blkaddr) { 3168 f2fs_info(sbi, "Wrong SIT boundary, start(%u) end(%u) blocks(%u)", 3169 sit_blkaddr, nat_blkaddr, 3170 segment_count_sit << log_blocks_per_seg); 3171 return true; 3172 } 3173 3174 if (nat_blkaddr + (segment_count_nat << log_blocks_per_seg) != 3175 ssa_blkaddr) { 3176 f2fs_info(sbi, "Wrong NAT boundary, start(%u) end(%u) blocks(%u)", 3177 nat_blkaddr, ssa_blkaddr, 3178 segment_count_nat << log_blocks_per_seg); 3179 return true; 3180 } 3181 3182 if (ssa_blkaddr + (segment_count_ssa << log_blocks_per_seg) != 3183 main_blkaddr) { 3184 f2fs_info(sbi, "Wrong SSA boundary, start(%u) end(%u) blocks(%u)", 3185 ssa_blkaddr, main_blkaddr, 3186 segment_count_ssa << log_blocks_per_seg); 3187 return true; 3188 } 3189 3190 if (main_end_blkaddr > seg_end_blkaddr) { 3191 f2fs_info(sbi, "Wrong MAIN_AREA boundary, start(%u) end(%llu) block(%u)", 3192 main_blkaddr, seg_end_blkaddr, 3193 segment_count_main << log_blocks_per_seg); 3194 return true; 3195 } else if (main_end_blkaddr < seg_end_blkaddr) { 3196 int err = 0; 3197 char *res; 3198 3199 /* fix in-memory information all the time */ 3200 raw_super->segment_count = cpu_to_le32((main_end_blkaddr - 3201 segment0_blkaddr) >> log_blocks_per_seg); 3202 3203 if (f2fs_readonly(sb) || bdev_read_only(sb->s_bdev)) { 3204 set_sbi_flag(sbi, SBI_NEED_SB_WRITE); 3205 res = "internally"; 3206 } else { 3207 err = __f2fs_commit_super(bh, NULL); 3208 res = err ? "failed" : "done"; 3209 } 3210 f2fs_info(sbi, "Fix alignment : %s, start(%u) end(%llu) block(%u)", 3211 res, main_blkaddr, seg_end_blkaddr, 3212 segment_count_main << log_blocks_per_seg); 3213 if (err) 3214 return true; 3215 } 3216 return false; 3217 } 3218 3219 static int sanity_check_raw_super(struct f2fs_sb_info *sbi, 3220 struct buffer_head *bh) 3221 { 3222 block_t segment_count, segs_per_sec, secs_per_zone, segment_count_main; 3223 block_t total_sections, blocks_per_seg; 3224 struct f2fs_super_block *raw_super = (struct f2fs_super_block *) 3225 (bh->b_data + F2FS_SUPER_OFFSET); 3226 size_t crc_offset = 0; 3227 __u32 crc = 0; 3228 3229 if (le32_to_cpu(raw_super->magic) != F2FS_SUPER_MAGIC) { 3230 f2fs_info(sbi, "Magic Mismatch, valid(0x%x) - read(0x%x)", 3231 F2FS_SUPER_MAGIC, le32_to_cpu(raw_super->magic)); 3232 return -EINVAL; 3233 } 3234 3235 /* Check checksum_offset and crc in superblock */ 3236 if (__F2FS_HAS_FEATURE(raw_super, F2FS_FEATURE_SB_CHKSUM)) { 3237 crc_offset = le32_to_cpu(raw_super->checksum_offset); 3238 if (crc_offset != 3239 offsetof(struct f2fs_super_block, crc)) { 3240 f2fs_info(sbi, "Invalid SB checksum offset: %zu", 3241 crc_offset); 3242 return -EFSCORRUPTED; 3243 } 3244 crc = le32_to_cpu(raw_super->crc); 3245 if (!f2fs_crc_valid(sbi, crc, raw_super, crc_offset)) { 3246 f2fs_info(sbi, "Invalid SB checksum value: %u", crc); 3247 return -EFSCORRUPTED; 3248 } 3249 } 3250 3251 /* Currently, support only 4KB block size */ 3252 if (le32_to_cpu(raw_super->log_blocksize) != F2FS_BLKSIZE_BITS) { 3253 f2fs_info(sbi, "Invalid log_blocksize (%u), supports only %u", 3254 le32_to_cpu(raw_super->log_blocksize), 3255 F2FS_BLKSIZE_BITS); 3256 return -EFSCORRUPTED; 3257 } 3258 3259 /* check log blocks per segment */ 3260 if (le32_to_cpu(raw_super->log_blocks_per_seg) != 9) { 3261 f2fs_info(sbi, "Invalid log blocks per segment (%u)", 3262 le32_to_cpu(raw_super->log_blocks_per_seg)); 3263 return -EFSCORRUPTED; 3264 } 3265 3266 /* Currently, support 512/1024/2048/4096 bytes sector size */ 3267 if (le32_to_cpu(raw_super->log_sectorsize) > 3268 F2FS_MAX_LOG_SECTOR_SIZE || 3269 le32_to_cpu(raw_super->log_sectorsize) < 3270 F2FS_MIN_LOG_SECTOR_SIZE) { 3271 f2fs_info(sbi, "Invalid log sectorsize (%u)", 3272 le32_to_cpu(raw_super->log_sectorsize)); 3273 return -EFSCORRUPTED; 3274 } 3275 if (le32_to_cpu(raw_super->log_sectors_per_block) + 3276 le32_to_cpu(raw_super->log_sectorsize) != 3277 F2FS_MAX_LOG_SECTOR_SIZE) { 3278 f2fs_info(sbi, "Invalid log sectors per block(%u) log sectorsize(%u)", 3279 le32_to_cpu(raw_super->log_sectors_per_block), 3280 le32_to_cpu(raw_super->log_sectorsize)); 3281 return -EFSCORRUPTED; 3282 } 3283 3284 segment_count = le32_to_cpu(raw_super->segment_count); 3285 segment_count_main = le32_to_cpu(raw_super->segment_count_main); 3286 segs_per_sec = le32_to_cpu(raw_super->segs_per_sec); 3287 secs_per_zone = le32_to_cpu(raw_super->secs_per_zone); 3288 total_sections = le32_to_cpu(raw_super->section_count); 3289 3290 /* blocks_per_seg should be 512, given the above check */ 3291 blocks_per_seg = 1 << le32_to_cpu(raw_super->log_blocks_per_seg); 3292 3293 if (segment_count > F2FS_MAX_SEGMENT || 3294 segment_count < F2FS_MIN_SEGMENTS) { 3295 f2fs_info(sbi, "Invalid segment count (%u)", segment_count); 3296 return -EFSCORRUPTED; 3297 } 3298 3299 if (total_sections > segment_count_main || total_sections < 1 || 3300 segs_per_sec > segment_count || !segs_per_sec) { 3301 f2fs_info(sbi, "Invalid segment/section count (%u, %u x %u)", 3302 segment_count, total_sections, segs_per_sec); 3303 return -EFSCORRUPTED; 3304 } 3305 3306 if (segment_count_main != total_sections * segs_per_sec) { 3307 f2fs_info(sbi, "Invalid segment/section count (%u != %u * %u)", 3308 segment_count_main, total_sections, segs_per_sec); 3309 return -EFSCORRUPTED; 3310 } 3311 3312 if ((segment_count / segs_per_sec) < total_sections) { 3313 f2fs_info(sbi, "Small segment_count (%u < %u * %u)", 3314 segment_count, segs_per_sec, total_sections); 3315 return -EFSCORRUPTED; 3316 } 3317 3318 if (segment_count > (le64_to_cpu(raw_super->block_count) >> 9)) { 3319 f2fs_info(sbi, "Wrong segment_count / block_count (%u > %llu)", 3320 segment_count, le64_to_cpu(raw_super->block_count)); 3321 return -EFSCORRUPTED; 3322 } 3323 3324 if (RDEV(0).path[0]) { 3325 block_t dev_seg_count = le32_to_cpu(RDEV(0).total_segments); 3326 int i = 1; 3327 3328 while (i < MAX_DEVICES && RDEV(i).path[0]) { 3329 dev_seg_count += le32_to_cpu(RDEV(i).total_segments); 3330 i++; 3331 } 3332 if (segment_count != dev_seg_count) { 3333 f2fs_info(sbi, "Segment count (%u) mismatch with total segments from devices (%u)", 3334 segment_count, dev_seg_count); 3335 return -EFSCORRUPTED; 3336 } 3337 } else { 3338 if (__F2FS_HAS_FEATURE(raw_super, F2FS_FEATURE_BLKZONED) && 3339 !bdev_is_zoned(sbi->sb->s_bdev)) { 3340 f2fs_info(sbi, "Zoned block device path is missing"); 3341 return -EFSCORRUPTED; 3342 } 3343 } 3344 3345 if (secs_per_zone > total_sections || !secs_per_zone) { 3346 f2fs_info(sbi, "Wrong secs_per_zone / total_sections (%u, %u)", 3347 secs_per_zone, total_sections); 3348 return -EFSCORRUPTED; 3349 } 3350 if (le32_to_cpu(raw_super->extension_count) > F2FS_MAX_EXTENSION || 3351 raw_super->hot_ext_count > F2FS_MAX_EXTENSION || 3352 (le32_to_cpu(raw_super->extension_count) + 3353 raw_super->hot_ext_count) > F2FS_MAX_EXTENSION) { 3354 f2fs_info(sbi, "Corrupted extension count (%u + %u > %u)", 3355 le32_to_cpu(raw_super->extension_count), 3356 raw_super->hot_ext_count, 3357 F2FS_MAX_EXTENSION); 3358 return -EFSCORRUPTED; 3359 } 3360 3361 if (le32_to_cpu(raw_super->cp_payload) >= 3362 (blocks_per_seg - F2FS_CP_PACKS - 3363 NR_CURSEG_PERSIST_TYPE)) { 3364 f2fs_info(sbi, "Insane cp_payload (%u >= %u)", 3365 le32_to_cpu(raw_super->cp_payload), 3366 blocks_per_seg - F2FS_CP_PACKS - 3367 NR_CURSEG_PERSIST_TYPE); 3368 return -EFSCORRUPTED; 3369 } 3370 3371 /* check reserved ino info */ 3372 if (le32_to_cpu(raw_super->node_ino) != 1 || 3373 le32_to_cpu(raw_super->meta_ino) != 2 || 3374 le32_to_cpu(raw_super->root_ino) != 3) { 3375 f2fs_info(sbi, "Invalid Fs Meta Ino: node(%u) meta(%u) root(%u)", 3376 le32_to_cpu(raw_super->node_ino), 3377 le32_to_cpu(raw_super->meta_ino), 3378 le32_to_cpu(raw_super->root_ino)); 3379 return -EFSCORRUPTED; 3380 } 3381 3382 /* check CP/SIT/NAT/SSA/MAIN_AREA area boundary */ 3383 if (sanity_check_area_boundary(sbi, bh)) 3384 return -EFSCORRUPTED; 3385 3386 return 0; 3387 } 3388 3389 int f2fs_sanity_check_ckpt(struct f2fs_sb_info *sbi) 3390 { 3391 unsigned int total, fsmeta; 3392 struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi); 3393 struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi); 3394 unsigned int ovp_segments, reserved_segments; 3395 unsigned int main_segs, blocks_per_seg; 3396 unsigned int sit_segs, nat_segs; 3397 unsigned int sit_bitmap_size, nat_bitmap_size; 3398 unsigned int log_blocks_per_seg; 3399 unsigned int segment_count_main; 3400 unsigned int cp_pack_start_sum, cp_payload; 3401 block_t user_block_count, valid_user_blocks; 3402 block_t avail_node_count, valid_node_count; 3403 unsigned int nat_blocks, nat_bits_bytes, nat_bits_blocks; 3404 int i, j; 3405 3406 total = le32_to_cpu(raw_super->segment_count); 3407 fsmeta = le32_to_cpu(raw_super->segment_count_ckpt); 3408 sit_segs = le32_to_cpu(raw_super->segment_count_sit); 3409 fsmeta += sit_segs; 3410 nat_segs = le32_to_cpu(raw_super->segment_count_nat); 3411 fsmeta += nat_segs; 3412 fsmeta += le32_to_cpu(ckpt->rsvd_segment_count); 3413 fsmeta += le32_to_cpu(raw_super->segment_count_ssa); 3414 3415 if (unlikely(fsmeta >= total)) 3416 return 1; 3417 3418 ovp_segments = le32_to_cpu(ckpt->overprov_segment_count); 3419 reserved_segments = le32_to_cpu(ckpt->rsvd_segment_count); 3420 3421 if (!f2fs_sb_has_readonly(sbi) && 3422 unlikely(fsmeta < F2FS_MIN_META_SEGMENTS || 3423 ovp_segments == 0 || reserved_segments == 0)) { 3424 f2fs_err(sbi, "Wrong layout: check mkfs.f2fs version"); 3425 return 1; 3426 } 3427 user_block_count = le64_to_cpu(ckpt->user_block_count); 3428 segment_count_main = le32_to_cpu(raw_super->segment_count_main) + 3429 (f2fs_sb_has_readonly(sbi) ? 1 : 0); 3430 log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg); 3431 if (!user_block_count || user_block_count >= 3432 segment_count_main << log_blocks_per_seg) { 3433 f2fs_err(sbi, "Wrong user_block_count: %u", 3434 user_block_count); 3435 return 1; 3436 } 3437 3438 valid_user_blocks = le64_to_cpu(ckpt->valid_block_count); 3439 if (valid_user_blocks > user_block_count) { 3440 f2fs_err(sbi, "Wrong valid_user_blocks: %u, user_block_count: %u", 3441 valid_user_blocks, user_block_count); 3442 return 1; 3443 } 3444 3445 valid_node_count = le32_to_cpu(ckpt->valid_node_count); 3446 avail_node_count = sbi->total_node_count - F2FS_RESERVED_NODE_NUM; 3447 if (valid_node_count > avail_node_count) { 3448 f2fs_err(sbi, "Wrong valid_node_count: %u, avail_node_count: %u", 3449 valid_node_count, avail_node_count); 3450 return 1; 3451 } 3452 3453 main_segs = le32_to_cpu(raw_super->segment_count_main); 3454 blocks_per_seg = sbi->blocks_per_seg; 3455 3456 for (i = 0; i < NR_CURSEG_NODE_TYPE; i++) { 3457 if (le32_to_cpu(ckpt->cur_node_segno[i]) >= main_segs || 3458 le16_to_cpu(ckpt->cur_node_blkoff[i]) >= blocks_per_seg) 3459 return 1; 3460 3461 if (f2fs_sb_has_readonly(sbi)) 3462 goto check_data; 3463 3464 for (j = i + 1; j < NR_CURSEG_NODE_TYPE; j++) { 3465 if (le32_to_cpu(ckpt->cur_node_segno[i]) == 3466 le32_to_cpu(ckpt->cur_node_segno[j])) { 3467 f2fs_err(sbi, "Node segment (%u, %u) has the same segno: %u", 3468 i, j, 3469 le32_to_cpu(ckpt->cur_node_segno[i])); 3470 return 1; 3471 } 3472 } 3473 } 3474 check_data: 3475 for (i = 0; i < NR_CURSEG_DATA_TYPE; i++) { 3476 if (le32_to_cpu(ckpt->cur_data_segno[i]) >= main_segs || 3477 le16_to_cpu(ckpt->cur_data_blkoff[i]) >= blocks_per_seg) 3478 return 1; 3479 3480 if (f2fs_sb_has_readonly(sbi)) 3481 goto skip_cross; 3482 3483 for (j = i + 1; j < NR_CURSEG_DATA_TYPE; j++) { 3484 if (le32_to_cpu(ckpt->cur_data_segno[i]) == 3485 le32_to_cpu(ckpt->cur_data_segno[j])) { 3486 f2fs_err(sbi, "Data segment (%u, %u) has the same segno: %u", 3487 i, j, 3488 le32_to_cpu(ckpt->cur_data_segno[i])); 3489 return 1; 3490 } 3491 } 3492 } 3493 for (i = 0; i < NR_CURSEG_NODE_TYPE; i++) { 3494 for (j = 0; j < NR_CURSEG_DATA_TYPE; j++) { 3495 if (le32_to_cpu(ckpt->cur_node_segno[i]) == 3496 le32_to_cpu(ckpt->cur_data_segno[j])) { 3497 f2fs_err(sbi, "Node segment (%u) and Data segment (%u) has the same segno: %u", 3498 i, j, 3499 le32_to_cpu(ckpt->cur_node_segno[i])); 3500 return 1; 3501 } 3502 } 3503 } 3504 skip_cross: 3505 sit_bitmap_size = le32_to_cpu(ckpt->sit_ver_bitmap_bytesize); 3506 nat_bitmap_size = le32_to_cpu(ckpt->nat_ver_bitmap_bytesize); 3507 3508 if (sit_bitmap_size != ((sit_segs / 2) << log_blocks_per_seg) / 8 || 3509 nat_bitmap_size != ((nat_segs / 2) << log_blocks_per_seg) / 8) { 3510 f2fs_err(sbi, "Wrong bitmap size: sit: %u, nat:%u", 3511 sit_bitmap_size, nat_bitmap_size); 3512 return 1; 3513 } 3514 3515 cp_pack_start_sum = __start_sum_addr(sbi); 3516 cp_payload = __cp_payload(sbi); 3517 if (cp_pack_start_sum < cp_payload + 1 || 3518 cp_pack_start_sum > blocks_per_seg - 1 - 3519 NR_CURSEG_PERSIST_TYPE) { 3520 f2fs_err(sbi, "Wrong cp_pack_start_sum: %u", 3521 cp_pack_start_sum); 3522 return 1; 3523 } 3524 3525 if (__is_set_ckpt_flags(ckpt, CP_LARGE_NAT_BITMAP_FLAG) && 3526 le32_to_cpu(ckpt->checksum_offset) != CP_MIN_CHKSUM_OFFSET) { 3527 f2fs_warn(sbi, "using deprecated layout of large_nat_bitmap, " 3528 "please run fsck v1.13.0 or higher to repair, chksum_offset: %u, " 3529 "fixed with patch: \"f2fs-tools: relocate chksum_offset for large_nat_bitmap feature\"", 3530 le32_to_cpu(ckpt->checksum_offset)); 3531 return 1; 3532 } 3533 3534 nat_blocks = nat_segs << log_blocks_per_seg; 3535 nat_bits_bytes = nat_blocks / BITS_PER_BYTE; 3536 nat_bits_blocks = F2FS_BLK_ALIGN((nat_bits_bytes << 1) + 8); 3537 if (__is_set_ckpt_flags(ckpt, CP_NAT_BITS_FLAG) && 3538 (cp_payload + F2FS_CP_PACKS + 3539 NR_CURSEG_PERSIST_TYPE + nat_bits_blocks >= blocks_per_seg)) { 3540 f2fs_warn(sbi, "Insane cp_payload: %u, nat_bits_blocks: %u)", 3541 cp_payload, nat_bits_blocks); 3542 return 1; 3543 } 3544 3545 if (unlikely(f2fs_cp_error(sbi))) { 3546 f2fs_err(sbi, "A bug case: need to run fsck"); 3547 return 1; 3548 } 3549 return 0; 3550 } 3551 3552 static void init_sb_info(struct f2fs_sb_info *sbi) 3553 { 3554 struct f2fs_super_block *raw_super = sbi->raw_super; 3555 int i; 3556 3557 sbi->log_sectors_per_block = 3558 le32_to_cpu(raw_super->log_sectors_per_block); 3559 sbi->log_blocksize = le32_to_cpu(raw_super->log_blocksize); 3560 sbi->blocksize = 1 << sbi->log_blocksize; 3561 sbi->log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg); 3562 sbi->blocks_per_seg = 1 << sbi->log_blocks_per_seg; 3563 sbi->segs_per_sec = le32_to_cpu(raw_super->segs_per_sec); 3564 sbi->secs_per_zone = le32_to_cpu(raw_super->secs_per_zone); 3565 sbi->total_sections = le32_to_cpu(raw_super->section_count); 3566 sbi->total_node_count = 3567 (le32_to_cpu(raw_super->segment_count_nat) / 2) 3568 * sbi->blocks_per_seg * NAT_ENTRY_PER_BLOCK; 3569 F2FS_ROOT_INO(sbi) = le32_to_cpu(raw_super->root_ino); 3570 F2FS_NODE_INO(sbi) = le32_to_cpu(raw_super->node_ino); 3571 F2FS_META_INO(sbi) = le32_to_cpu(raw_super->meta_ino); 3572 sbi->cur_victim_sec = NULL_SECNO; 3573 sbi->gc_mode = GC_NORMAL; 3574 sbi->next_victim_seg[BG_GC] = NULL_SEGNO; 3575 sbi->next_victim_seg[FG_GC] = NULL_SEGNO; 3576 sbi->max_victim_search = DEF_MAX_VICTIM_SEARCH; 3577 sbi->migration_granularity = sbi->segs_per_sec; 3578 sbi->seq_file_ra_mul = MIN_RA_MUL; 3579 sbi->max_fragment_chunk = DEF_FRAGMENT_SIZE; 3580 sbi->max_fragment_hole = DEF_FRAGMENT_SIZE; 3581 spin_lock_init(&sbi->gc_urgent_high_lock); 3582 3583 sbi->dir_level = DEF_DIR_LEVEL; 3584 sbi->interval_time[CP_TIME] = DEF_CP_INTERVAL; 3585 sbi->interval_time[REQ_TIME] = DEF_IDLE_INTERVAL; 3586 sbi->interval_time[DISCARD_TIME] = DEF_IDLE_INTERVAL; 3587 sbi->interval_time[GC_TIME] = DEF_IDLE_INTERVAL; 3588 sbi->interval_time[DISABLE_TIME] = DEF_DISABLE_INTERVAL; 3589 sbi->interval_time[UMOUNT_DISCARD_TIMEOUT] = 3590 DEF_UMOUNT_DISCARD_TIMEOUT; 3591 clear_sbi_flag(sbi, SBI_NEED_FSCK); 3592 3593 for (i = 0; i < NR_COUNT_TYPE; i++) 3594 atomic_set(&sbi->nr_pages[i], 0); 3595 3596 for (i = 0; i < META; i++) 3597 atomic_set(&sbi->wb_sync_req[i], 0); 3598 3599 INIT_LIST_HEAD(&sbi->s_list); 3600 mutex_init(&sbi->umount_mutex); 3601 init_f2fs_rwsem(&sbi->io_order_lock); 3602 spin_lock_init(&sbi->cp_lock); 3603 3604 sbi->dirty_device = 0; 3605 spin_lock_init(&sbi->dev_lock); 3606 3607 init_f2fs_rwsem(&sbi->sb_lock); 3608 init_f2fs_rwsem(&sbi->pin_sem); 3609 } 3610 3611 static int init_percpu_info(struct f2fs_sb_info *sbi) 3612 { 3613 int err; 3614 3615 err = percpu_counter_init(&sbi->alloc_valid_block_count, 0, GFP_KERNEL); 3616 if (err) 3617 return err; 3618 3619 err = percpu_counter_init(&sbi->rf_node_block_count, 0, GFP_KERNEL); 3620 if (err) 3621 goto err_valid_block; 3622 3623 err = percpu_counter_init(&sbi->total_valid_inode_count, 0, 3624 GFP_KERNEL); 3625 if (err) 3626 goto err_node_block; 3627 return 0; 3628 3629 err_node_block: 3630 percpu_counter_destroy(&sbi->rf_node_block_count); 3631 err_valid_block: 3632 percpu_counter_destroy(&sbi->alloc_valid_block_count); 3633 return err; 3634 } 3635 3636 #ifdef CONFIG_BLK_DEV_ZONED 3637 3638 struct f2fs_report_zones_args { 3639 struct f2fs_dev_info *dev; 3640 bool zone_cap_mismatch; 3641 }; 3642 3643 static int f2fs_report_zone_cb(struct blk_zone *zone, unsigned int idx, 3644 void *data) 3645 { 3646 struct f2fs_report_zones_args *rz_args = data; 3647 3648 if (zone->type == BLK_ZONE_TYPE_CONVENTIONAL) 3649 return 0; 3650 3651 set_bit(idx, rz_args->dev->blkz_seq); 3652 rz_args->dev->zone_capacity_blocks[idx] = zone->capacity >> 3653 F2FS_LOG_SECTORS_PER_BLOCK; 3654 if (zone->len != zone->capacity && !rz_args->zone_cap_mismatch) 3655 rz_args->zone_cap_mismatch = true; 3656 3657 return 0; 3658 } 3659 3660 static int init_blkz_info(struct f2fs_sb_info *sbi, int devi) 3661 { 3662 struct block_device *bdev = FDEV(devi).bdev; 3663 sector_t nr_sectors = bdev_nr_sectors(bdev); 3664 struct f2fs_report_zones_args rep_zone_arg; 3665 u64 zone_sectors; 3666 int ret; 3667 3668 if (!f2fs_sb_has_blkzoned(sbi)) 3669 return 0; 3670 3671 zone_sectors = bdev_zone_sectors(bdev); 3672 if (!is_power_of_2(zone_sectors)) { 3673 f2fs_err(sbi, "F2FS does not support non power of 2 zone sizes\n"); 3674 return -EINVAL; 3675 } 3676 3677 if (sbi->blocks_per_blkz && sbi->blocks_per_blkz != 3678 SECTOR_TO_BLOCK(zone_sectors)) 3679 return -EINVAL; 3680 sbi->blocks_per_blkz = SECTOR_TO_BLOCK(zone_sectors); 3681 if (sbi->log_blocks_per_blkz && sbi->log_blocks_per_blkz != 3682 __ilog2_u32(sbi->blocks_per_blkz)) 3683 return -EINVAL; 3684 sbi->log_blocks_per_blkz = __ilog2_u32(sbi->blocks_per_blkz); 3685 FDEV(devi).nr_blkz = SECTOR_TO_BLOCK(nr_sectors) >> 3686 sbi->log_blocks_per_blkz; 3687 if (nr_sectors & (zone_sectors - 1)) 3688 FDEV(devi).nr_blkz++; 3689 3690 FDEV(devi).blkz_seq = f2fs_kvzalloc(sbi, 3691 BITS_TO_LONGS(FDEV(devi).nr_blkz) 3692 * sizeof(unsigned long), 3693 GFP_KERNEL); 3694 if (!FDEV(devi).blkz_seq) 3695 return -ENOMEM; 3696 3697 /* Get block zones type and zone-capacity */ 3698 FDEV(devi).zone_capacity_blocks = f2fs_kzalloc(sbi, 3699 FDEV(devi).nr_blkz * sizeof(block_t), 3700 GFP_KERNEL); 3701 if (!FDEV(devi).zone_capacity_blocks) 3702 return -ENOMEM; 3703 3704 rep_zone_arg.dev = &FDEV(devi); 3705 rep_zone_arg.zone_cap_mismatch = false; 3706 3707 ret = blkdev_report_zones(bdev, 0, BLK_ALL_ZONES, f2fs_report_zone_cb, 3708 &rep_zone_arg); 3709 if (ret < 0) 3710 return ret; 3711 3712 if (!rep_zone_arg.zone_cap_mismatch) { 3713 kfree(FDEV(devi).zone_capacity_blocks); 3714 FDEV(devi).zone_capacity_blocks = NULL; 3715 } 3716 3717 return 0; 3718 } 3719 #endif 3720 3721 /* 3722 * Read f2fs raw super block. 3723 * Because we have two copies of super block, so read both of them 3724 * to get the first valid one. If any one of them is broken, we pass 3725 * them recovery flag back to the caller. 3726 */ 3727 static int read_raw_super_block(struct f2fs_sb_info *sbi, 3728 struct f2fs_super_block **raw_super, 3729 int *valid_super_block, int *recovery) 3730 { 3731 struct super_block *sb = sbi->sb; 3732 int block; 3733 struct buffer_head *bh; 3734 struct f2fs_super_block *super; 3735 int err = 0; 3736 3737 super = kzalloc(sizeof(struct f2fs_super_block), GFP_KERNEL); 3738 if (!super) 3739 return -ENOMEM; 3740 3741 for (block = 0; block < 2; block++) { 3742 bh = sb_bread(sb, block); 3743 if (!bh) { 3744 f2fs_err(sbi, "Unable to read %dth superblock", 3745 block + 1); 3746 err = -EIO; 3747 *recovery = 1; 3748 continue; 3749 } 3750 3751 /* sanity checking of raw super */ 3752 err = sanity_check_raw_super(sbi, bh); 3753 if (err) { 3754 f2fs_err(sbi, "Can't find valid F2FS filesystem in %dth superblock", 3755 block + 1); 3756 brelse(bh); 3757 *recovery = 1; 3758 continue; 3759 } 3760 3761 if (!*raw_super) { 3762 memcpy(super, bh->b_data + F2FS_SUPER_OFFSET, 3763 sizeof(*super)); 3764 *valid_super_block = block; 3765 *raw_super = super; 3766 } 3767 brelse(bh); 3768 } 3769 3770 /* No valid superblock */ 3771 if (!*raw_super) 3772 kfree(super); 3773 else 3774 err = 0; 3775 3776 return err; 3777 } 3778 3779 int f2fs_commit_super(struct f2fs_sb_info *sbi, bool recover) 3780 { 3781 struct buffer_head *bh; 3782 __u32 crc = 0; 3783 int err; 3784 3785 if ((recover && f2fs_readonly(sbi->sb)) || 3786 bdev_read_only(sbi->sb->s_bdev)) { 3787 set_sbi_flag(sbi, SBI_NEED_SB_WRITE); 3788 return -EROFS; 3789 } 3790 3791 /* we should update superblock crc here */ 3792 if (!recover && f2fs_sb_has_sb_chksum(sbi)) { 3793 crc = f2fs_crc32(sbi, F2FS_RAW_SUPER(sbi), 3794 offsetof(struct f2fs_super_block, crc)); 3795 F2FS_RAW_SUPER(sbi)->crc = cpu_to_le32(crc); 3796 } 3797 3798 /* write back-up superblock first */ 3799 bh = sb_bread(sbi->sb, sbi->valid_super_block ? 0 : 1); 3800 if (!bh) 3801 return -EIO; 3802 err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi)); 3803 brelse(bh); 3804 3805 /* if we are in recovery path, skip writing valid superblock */ 3806 if (recover || err) 3807 return err; 3808 3809 /* write current valid superblock */ 3810 bh = sb_bread(sbi->sb, sbi->valid_super_block); 3811 if (!bh) 3812 return -EIO; 3813 err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi)); 3814 brelse(bh); 3815 return err; 3816 } 3817 3818 static int f2fs_scan_devices(struct f2fs_sb_info *sbi) 3819 { 3820 struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi); 3821 unsigned int max_devices = MAX_DEVICES; 3822 unsigned int logical_blksize; 3823 int i; 3824 3825 /* Initialize single device information */ 3826 if (!RDEV(0).path[0]) { 3827 if (!bdev_is_zoned(sbi->sb->s_bdev)) 3828 return 0; 3829 max_devices = 1; 3830 } 3831 3832 /* 3833 * Initialize multiple devices information, or single 3834 * zoned block device information. 3835 */ 3836 sbi->devs = f2fs_kzalloc(sbi, 3837 array_size(max_devices, 3838 sizeof(struct f2fs_dev_info)), 3839 GFP_KERNEL); 3840 if (!sbi->devs) 3841 return -ENOMEM; 3842 3843 logical_blksize = bdev_logical_block_size(sbi->sb->s_bdev); 3844 sbi->aligned_blksize = true; 3845 3846 for (i = 0; i < max_devices; i++) { 3847 3848 if (i > 0 && !RDEV(i).path[0]) 3849 break; 3850 3851 if (max_devices == 1) { 3852 /* Single zoned block device mount */ 3853 FDEV(0).bdev = 3854 blkdev_get_by_dev(sbi->sb->s_bdev->bd_dev, 3855 sbi->sb->s_mode, sbi->sb->s_type); 3856 } else { 3857 /* Multi-device mount */ 3858 memcpy(FDEV(i).path, RDEV(i).path, MAX_PATH_LEN); 3859 FDEV(i).total_segments = 3860 le32_to_cpu(RDEV(i).total_segments); 3861 if (i == 0) { 3862 FDEV(i).start_blk = 0; 3863 FDEV(i).end_blk = FDEV(i).start_blk + 3864 (FDEV(i).total_segments << 3865 sbi->log_blocks_per_seg) - 1 + 3866 le32_to_cpu(raw_super->segment0_blkaddr); 3867 } else { 3868 FDEV(i).start_blk = FDEV(i - 1).end_blk + 1; 3869 FDEV(i).end_blk = FDEV(i).start_blk + 3870 (FDEV(i).total_segments << 3871 sbi->log_blocks_per_seg) - 1; 3872 } 3873 FDEV(i).bdev = blkdev_get_by_path(FDEV(i).path, 3874 sbi->sb->s_mode, sbi->sb->s_type); 3875 } 3876 if (IS_ERR(FDEV(i).bdev)) 3877 return PTR_ERR(FDEV(i).bdev); 3878 3879 /* to release errored devices */ 3880 sbi->s_ndevs = i + 1; 3881 3882 if (logical_blksize != bdev_logical_block_size(FDEV(i).bdev)) 3883 sbi->aligned_blksize = false; 3884 3885 #ifdef CONFIG_BLK_DEV_ZONED 3886 if (bdev_zoned_model(FDEV(i).bdev) == BLK_ZONED_HM && 3887 !f2fs_sb_has_blkzoned(sbi)) { 3888 f2fs_err(sbi, "Zoned block device feature not enabled"); 3889 return -EINVAL; 3890 } 3891 if (bdev_zoned_model(FDEV(i).bdev) != BLK_ZONED_NONE) { 3892 if (init_blkz_info(sbi, i)) { 3893 f2fs_err(sbi, "Failed to initialize F2FS blkzone information"); 3894 return -EINVAL; 3895 } 3896 if (max_devices == 1) 3897 break; 3898 f2fs_info(sbi, "Mount Device [%2d]: %20s, %8u, %8x - %8x (zone: %s)", 3899 i, FDEV(i).path, 3900 FDEV(i).total_segments, 3901 FDEV(i).start_blk, FDEV(i).end_blk, 3902 bdev_zoned_model(FDEV(i).bdev) == BLK_ZONED_HA ? 3903 "Host-aware" : "Host-managed"); 3904 continue; 3905 } 3906 #endif 3907 f2fs_info(sbi, "Mount Device [%2d]: %20s, %8u, %8x - %8x", 3908 i, FDEV(i).path, 3909 FDEV(i).total_segments, 3910 FDEV(i).start_blk, FDEV(i).end_blk); 3911 } 3912 f2fs_info(sbi, 3913 "IO Block Size: %8d KB", F2FS_IO_SIZE_KB(sbi)); 3914 return 0; 3915 } 3916 3917 static int f2fs_setup_casefold(struct f2fs_sb_info *sbi) 3918 { 3919 #if IS_ENABLED(CONFIG_UNICODE) 3920 if (f2fs_sb_has_casefold(sbi) && !sbi->sb->s_encoding) { 3921 const struct f2fs_sb_encodings *encoding_info; 3922 struct unicode_map *encoding; 3923 __u16 encoding_flags; 3924 3925 encoding_info = f2fs_sb_read_encoding(sbi->raw_super); 3926 if (!encoding_info) { 3927 f2fs_err(sbi, 3928 "Encoding requested by superblock is unknown"); 3929 return -EINVAL; 3930 } 3931 3932 encoding_flags = le16_to_cpu(sbi->raw_super->s_encoding_flags); 3933 encoding = utf8_load(encoding_info->version); 3934 if (IS_ERR(encoding)) { 3935 f2fs_err(sbi, 3936 "can't mount with superblock charset: %s-%u.%u.%u " 3937 "not supported by the kernel. flags: 0x%x.", 3938 encoding_info->name, 3939 unicode_major(encoding_info->version), 3940 unicode_minor(encoding_info->version), 3941 unicode_rev(encoding_info->version), 3942 encoding_flags); 3943 return PTR_ERR(encoding); 3944 } 3945 f2fs_info(sbi, "Using encoding defined by superblock: " 3946 "%s-%u.%u.%u with flags 0x%hx", encoding_info->name, 3947 unicode_major(encoding_info->version), 3948 unicode_minor(encoding_info->version), 3949 unicode_rev(encoding_info->version), 3950 encoding_flags); 3951 3952 sbi->sb->s_encoding = encoding; 3953 sbi->sb->s_encoding_flags = encoding_flags; 3954 } 3955 #else 3956 if (f2fs_sb_has_casefold(sbi)) { 3957 f2fs_err(sbi, "Filesystem with casefold feature cannot be mounted without CONFIG_UNICODE"); 3958 return -EINVAL; 3959 } 3960 #endif 3961 return 0; 3962 } 3963 3964 static void f2fs_tuning_parameters(struct f2fs_sb_info *sbi) 3965 { 3966 struct f2fs_sm_info *sm_i = SM_I(sbi); 3967 3968 /* adjust parameters according to the volume size */ 3969 if (sm_i->main_segments <= SMALL_VOLUME_SEGMENTS) { 3970 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_REUSE; 3971 if (f2fs_block_unit_discard(sbi)) 3972 sm_i->dcc_info->discard_granularity = 1; 3973 sm_i->ipu_policy = 1 << F2FS_IPU_FORCE | 3974 1 << F2FS_IPU_HONOR_OPU_WRITE; 3975 } 3976 3977 sbi->readdir_ra = 1; 3978 } 3979 3980 static int f2fs_fill_super(struct super_block *sb, void *data, int silent) 3981 { 3982 struct f2fs_sb_info *sbi; 3983 struct f2fs_super_block *raw_super; 3984 struct inode *root; 3985 int err; 3986 bool skip_recovery = false, need_fsck = false; 3987 char *options = NULL; 3988 int recovery, i, valid_super_block; 3989 struct curseg_info *seg_i; 3990 int retry_cnt = 1; 3991 3992 try_onemore: 3993 err = -EINVAL; 3994 raw_super = NULL; 3995 valid_super_block = -1; 3996 recovery = 0; 3997 3998 /* allocate memory for f2fs-specific super block info */ 3999 sbi = kzalloc(sizeof(struct f2fs_sb_info), GFP_KERNEL); 4000 if (!sbi) 4001 return -ENOMEM; 4002 4003 sbi->sb = sb; 4004 4005 /* Load the checksum driver */ 4006 sbi->s_chksum_driver = crypto_alloc_shash("crc32", 0, 0); 4007 if (IS_ERR(sbi->s_chksum_driver)) { 4008 f2fs_err(sbi, "Cannot load crc32 driver."); 4009 err = PTR_ERR(sbi->s_chksum_driver); 4010 sbi->s_chksum_driver = NULL; 4011 goto free_sbi; 4012 } 4013 4014 /* set a block size */ 4015 if (unlikely(!sb_set_blocksize(sb, F2FS_BLKSIZE))) { 4016 f2fs_err(sbi, "unable to set blocksize"); 4017 goto free_sbi; 4018 } 4019 4020 err = read_raw_super_block(sbi, &raw_super, &valid_super_block, 4021 &recovery); 4022 if (err) 4023 goto free_sbi; 4024 4025 sb->s_fs_info = sbi; 4026 sbi->raw_super = raw_super; 4027 4028 /* precompute checksum seed for metadata */ 4029 if (f2fs_sb_has_inode_chksum(sbi)) 4030 sbi->s_chksum_seed = f2fs_chksum(sbi, ~0, raw_super->uuid, 4031 sizeof(raw_super->uuid)); 4032 4033 default_options(sbi); 4034 /* parse mount options */ 4035 options = kstrdup((const char *)data, GFP_KERNEL); 4036 if (data && !options) { 4037 err = -ENOMEM; 4038 goto free_sb_buf; 4039 } 4040 4041 err = parse_options(sb, options, false); 4042 if (err) 4043 goto free_options; 4044 4045 sb->s_maxbytes = max_file_blocks(NULL) << 4046 le32_to_cpu(raw_super->log_blocksize); 4047 sb->s_max_links = F2FS_LINK_MAX; 4048 4049 err = f2fs_setup_casefold(sbi); 4050 if (err) 4051 goto free_options; 4052 4053 #ifdef CONFIG_QUOTA 4054 sb->dq_op = &f2fs_quota_operations; 4055 sb->s_qcop = &f2fs_quotactl_ops; 4056 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ; 4057 4058 if (f2fs_sb_has_quota_ino(sbi)) { 4059 for (i = 0; i < MAXQUOTAS; i++) { 4060 if (f2fs_qf_ino(sbi->sb, i)) 4061 sbi->nquota_files++; 4062 } 4063 } 4064 #endif 4065 4066 sb->s_op = &f2fs_sops; 4067 #ifdef CONFIG_FS_ENCRYPTION 4068 sb->s_cop = &f2fs_cryptops; 4069 #endif 4070 #ifdef CONFIG_FS_VERITY 4071 sb->s_vop = &f2fs_verityops; 4072 #endif 4073 sb->s_xattr = f2fs_xattr_handlers; 4074 sb->s_export_op = &f2fs_export_ops; 4075 sb->s_magic = F2FS_SUPER_MAGIC; 4076 sb->s_time_gran = 1; 4077 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) | 4078 (test_opt(sbi, POSIX_ACL) ? SB_POSIXACL : 0); 4079 memcpy(&sb->s_uuid, raw_super->uuid, sizeof(raw_super->uuid)); 4080 sb->s_iflags |= SB_I_CGROUPWB; 4081 4082 /* init f2fs-specific super block info */ 4083 sbi->valid_super_block = valid_super_block; 4084 init_f2fs_rwsem(&sbi->gc_lock); 4085 mutex_init(&sbi->writepages); 4086 init_f2fs_rwsem(&sbi->cp_global_sem); 4087 init_f2fs_rwsem(&sbi->node_write); 4088 init_f2fs_rwsem(&sbi->node_change); 4089 4090 /* disallow all the data/node/meta page writes */ 4091 set_sbi_flag(sbi, SBI_POR_DOING); 4092 spin_lock_init(&sbi->stat_lock); 4093 4094 err = f2fs_init_write_merge_io(sbi); 4095 if (err) 4096 goto free_bio_info; 4097 4098 init_f2fs_rwsem(&sbi->cp_rwsem); 4099 init_f2fs_rwsem(&sbi->quota_sem); 4100 init_waitqueue_head(&sbi->cp_wait); 4101 init_sb_info(sbi); 4102 4103 err = f2fs_init_iostat(sbi); 4104 if (err) 4105 goto free_bio_info; 4106 4107 err = init_percpu_info(sbi); 4108 if (err) 4109 goto free_iostat; 4110 4111 if (F2FS_IO_ALIGNED(sbi)) { 4112 sbi->write_io_dummy = 4113 mempool_create_page_pool(2 * (F2FS_IO_SIZE(sbi) - 1), 0); 4114 if (!sbi->write_io_dummy) { 4115 err = -ENOMEM; 4116 goto free_percpu; 4117 } 4118 } 4119 4120 /* init per sbi slab cache */ 4121 err = f2fs_init_xattr_caches(sbi); 4122 if (err) 4123 goto free_io_dummy; 4124 err = f2fs_init_page_array_cache(sbi); 4125 if (err) 4126 goto free_xattr_cache; 4127 4128 /* get an inode for meta space */ 4129 sbi->meta_inode = f2fs_iget(sb, F2FS_META_INO(sbi)); 4130 if (IS_ERR(sbi->meta_inode)) { 4131 f2fs_err(sbi, "Failed to read F2FS meta data inode"); 4132 err = PTR_ERR(sbi->meta_inode); 4133 goto free_page_array_cache; 4134 } 4135 4136 err = f2fs_get_valid_checkpoint(sbi); 4137 if (err) { 4138 f2fs_err(sbi, "Failed to get valid F2FS checkpoint"); 4139 goto free_meta_inode; 4140 } 4141 4142 if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_QUOTA_NEED_FSCK_FLAG)) 4143 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR); 4144 if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_DISABLED_QUICK_FLAG)) { 4145 set_sbi_flag(sbi, SBI_CP_DISABLED_QUICK); 4146 sbi->interval_time[DISABLE_TIME] = DEF_DISABLE_QUICK_INTERVAL; 4147 } 4148 4149 if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_FSCK_FLAG)) 4150 set_sbi_flag(sbi, SBI_NEED_FSCK); 4151 4152 /* Initialize device list */ 4153 err = f2fs_scan_devices(sbi); 4154 if (err) { 4155 f2fs_err(sbi, "Failed to find devices"); 4156 goto free_devices; 4157 } 4158 4159 err = f2fs_init_post_read_wq(sbi); 4160 if (err) { 4161 f2fs_err(sbi, "Failed to initialize post read workqueue"); 4162 goto free_devices; 4163 } 4164 4165 sbi->total_valid_node_count = 4166 le32_to_cpu(sbi->ckpt->valid_node_count); 4167 percpu_counter_set(&sbi->total_valid_inode_count, 4168 le32_to_cpu(sbi->ckpt->valid_inode_count)); 4169 sbi->user_block_count = le64_to_cpu(sbi->ckpt->user_block_count); 4170 sbi->total_valid_block_count = 4171 le64_to_cpu(sbi->ckpt->valid_block_count); 4172 sbi->last_valid_block_count = sbi->total_valid_block_count; 4173 sbi->reserved_blocks = 0; 4174 sbi->current_reserved_blocks = 0; 4175 limit_reserve_root(sbi); 4176 adjust_unusable_cap_perc(sbi); 4177 4178 for (i = 0; i < NR_INODE_TYPE; i++) { 4179 INIT_LIST_HEAD(&sbi->inode_list[i]); 4180 spin_lock_init(&sbi->inode_lock[i]); 4181 } 4182 mutex_init(&sbi->flush_lock); 4183 4184 f2fs_init_extent_cache_info(sbi); 4185 4186 f2fs_init_ino_entry_info(sbi); 4187 4188 f2fs_init_fsync_node_info(sbi); 4189 4190 /* setup checkpoint request control and start checkpoint issue thread */ 4191 f2fs_init_ckpt_req_control(sbi); 4192 if (!f2fs_readonly(sb) && !test_opt(sbi, DISABLE_CHECKPOINT) && 4193 test_opt(sbi, MERGE_CHECKPOINT)) { 4194 err = f2fs_start_ckpt_thread(sbi); 4195 if (err) { 4196 f2fs_err(sbi, 4197 "Failed to start F2FS issue_checkpoint_thread (%d)", 4198 err); 4199 goto stop_ckpt_thread; 4200 } 4201 } 4202 4203 /* setup f2fs internal modules */ 4204 err = f2fs_build_segment_manager(sbi); 4205 if (err) { 4206 f2fs_err(sbi, "Failed to initialize F2FS segment manager (%d)", 4207 err); 4208 goto free_sm; 4209 } 4210 err = f2fs_build_node_manager(sbi); 4211 if (err) { 4212 f2fs_err(sbi, "Failed to initialize F2FS node manager (%d)", 4213 err); 4214 goto free_nm; 4215 } 4216 4217 err = adjust_reserved_segment(sbi); 4218 if (err) 4219 goto free_nm; 4220 4221 /* For write statistics */ 4222 sbi->sectors_written_start = f2fs_get_sectors_written(sbi); 4223 4224 /* Read accumulated write IO statistics if exists */ 4225 seg_i = CURSEG_I(sbi, CURSEG_HOT_NODE); 4226 if (__exist_node_summaries(sbi)) 4227 sbi->kbytes_written = 4228 le64_to_cpu(seg_i->journal->info.kbytes_written); 4229 4230 f2fs_build_gc_manager(sbi); 4231 4232 err = f2fs_build_stats(sbi); 4233 if (err) 4234 goto free_nm; 4235 4236 /* get an inode for node space */ 4237 sbi->node_inode = f2fs_iget(sb, F2FS_NODE_INO(sbi)); 4238 if (IS_ERR(sbi->node_inode)) { 4239 f2fs_err(sbi, "Failed to read node inode"); 4240 err = PTR_ERR(sbi->node_inode); 4241 goto free_stats; 4242 } 4243 4244 /* read root inode and dentry */ 4245 root = f2fs_iget(sb, F2FS_ROOT_INO(sbi)); 4246 if (IS_ERR(root)) { 4247 f2fs_err(sbi, "Failed to read root inode"); 4248 err = PTR_ERR(root); 4249 goto free_node_inode; 4250 } 4251 if (!S_ISDIR(root->i_mode) || !root->i_blocks || 4252 !root->i_size || !root->i_nlink) { 4253 iput(root); 4254 err = -EINVAL; 4255 goto free_node_inode; 4256 } 4257 4258 sb->s_root = d_make_root(root); /* allocate root dentry */ 4259 if (!sb->s_root) { 4260 err = -ENOMEM; 4261 goto free_node_inode; 4262 } 4263 4264 err = f2fs_init_compress_inode(sbi); 4265 if (err) 4266 goto free_root_inode; 4267 4268 err = f2fs_register_sysfs(sbi); 4269 if (err) 4270 goto free_compress_inode; 4271 4272 #ifdef CONFIG_QUOTA 4273 /* Enable quota usage during mount */ 4274 if (f2fs_sb_has_quota_ino(sbi) && !f2fs_readonly(sb)) { 4275 err = f2fs_enable_quotas(sb); 4276 if (err) 4277 f2fs_err(sbi, "Cannot turn on quotas: error %d", err); 4278 } 4279 #endif 4280 /* if there are any orphan inodes, free them */ 4281 err = f2fs_recover_orphan_inodes(sbi); 4282 if (err) 4283 goto free_meta; 4284 4285 if (unlikely(is_set_ckpt_flags(sbi, CP_DISABLED_FLAG))) 4286 goto reset_checkpoint; 4287 4288 /* recover fsynced data */ 4289 if (!test_opt(sbi, DISABLE_ROLL_FORWARD) && 4290 !test_opt(sbi, NORECOVERY)) { 4291 /* 4292 * mount should be failed, when device has readonly mode, and 4293 * previous checkpoint was not done by clean system shutdown. 4294 */ 4295 if (f2fs_hw_is_readonly(sbi)) { 4296 if (!is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG)) { 4297 err = f2fs_recover_fsync_data(sbi, true); 4298 if (err > 0) { 4299 err = -EROFS; 4300 f2fs_err(sbi, "Need to recover fsync data, but " 4301 "write access unavailable, please try " 4302 "mount w/ disable_roll_forward or norecovery"); 4303 } 4304 if (err < 0) 4305 goto free_meta; 4306 } 4307 f2fs_info(sbi, "write access unavailable, skipping recovery"); 4308 goto reset_checkpoint; 4309 } 4310 4311 if (need_fsck) 4312 set_sbi_flag(sbi, SBI_NEED_FSCK); 4313 4314 if (skip_recovery) 4315 goto reset_checkpoint; 4316 4317 err = f2fs_recover_fsync_data(sbi, false); 4318 if (err < 0) { 4319 if (err != -ENOMEM) 4320 skip_recovery = true; 4321 need_fsck = true; 4322 f2fs_err(sbi, "Cannot recover all fsync data errno=%d", 4323 err); 4324 goto free_meta; 4325 } 4326 } else { 4327 err = f2fs_recover_fsync_data(sbi, true); 4328 4329 if (!f2fs_readonly(sb) && err > 0) { 4330 err = -EINVAL; 4331 f2fs_err(sbi, "Need to recover fsync data"); 4332 goto free_meta; 4333 } 4334 } 4335 4336 /* 4337 * If the f2fs is not readonly and fsync data recovery succeeds, 4338 * check zoned block devices' write pointer consistency. 4339 */ 4340 if (!err && !f2fs_readonly(sb) && f2fs_sb_has_blkzoned(sbi)) { 4341 err = f2fs_check_write_pointer(sbi); 4342 if (err) 4343 goto free_meta; 4344 } 4345 4346 reset_checkpoint: 4347 f2fs_init_inmem_curseg(sbi); 4348 4349 /* f2fs_recover_fsync_data() cleared this already */ 4350 clear_sbi_flag(sbi, SBI_POR_DOING); 4351 4352 if (test_opt(sbi, DISABLE_CHECKPOINT)) { 4353 err = f2fs_disable_checkpoint(sbi); 4354 if (err) 4355 goto sync_free_meta; 4356 } else if (is_set_ckpt_flags(sbi, CP_DISABLED_FLAG)) { 4357 f2fs_enable_checkpoint(sbi); 4358 } 4359 4360 /* 4361 * If filesystem is not mounted as read-only then 4362 * do start the gc_thread. 4363 */ 4364 if ((F2FS_OPTION(sbi).bggc_mode != BGGC_MODE_OFF || 4365 test_opt(sbi, GC_MERGE)) && !f2fs_readonly(sb)) { 4366 /* After POR, we can run background GC thread.*/ 4367 err = f2fs_start_gc_thread(sbi); 4368 if (err) 4369 goto sync_free_meta; 4370 } 4371 kvfree(options); 4372 4373 /* recover broken superblock */ 4374 if (recovery) { 4375 err = f2fs_commit_super(sbi, true); 4376 f2fs_info(sbi, "Try to recover %dth superblock, ret: %d", 4377 sbi->valid_super_block ? 1 : 2, err); 4378 } 4379 4380 f2fs_join_shrinker(sbi); 4381 4382 f2fs_tuning_parameters(sbi); 4383 4384 f2fs_notice(sbi, "Mounted with checkpoint version = %llx", 4385 cur_cp_version(F2FS_CKPT(sbi))); 4386 f2fs_update_time(sbi, CP_TIME); 4387 f2fs_update_time(sbi, REQ_TIME); 4388 clear_sbi_flag(sbi, SBI_CP_DISABLED_QUICK); 4389 return 0; 4390 4391 sync_free_meta: 4392 /* safe to flush all the data */ 4393 sync_filesystem(sbi->sb); 4394 retry_cnt = 0; 4395 4396 free_meta: 4397 #ifdef CONFIG_QUOTA 4398 f2fs_truncate_quota_inode_pages(sb); 4399 if (f2fs_sb_has_quota_ino(sbi) && !f2fs_readonly(sb)) 4400 f2fs_quota_off_umount(sbi->sb); 4401 #endif 4402 /* 4403 * Some dirty meta pages can be produced by f2fs_recover_orphan_inodes() 4404 * failed by EIO. Then, iput(node_inode) can trigger balance_fs_bg() 4405 * followed by f2fs_write_checkpoint() through f2fs_write_node_pages(), which 4406 * falls into an infinite loop in f2fs_sync_meta_pages(). 4407 */ 4408 truncate_inode_pages_final(META_MAPPING(sbi)); 4409 /* evict some inodes being cached by GC */ 4410 evict_inodes(sb); 4411 f2fs_unregister_sysfs(sbi); 4412 free_compress_inode: 4413 f2fs_destroy_compress_inode(sbi); 4414 free_root_inode: 4415 dput(sb->s_root); 4416 sb->s_root = NULL; 4417 free_node_inode: 4418 f2fs_release_ino_entry(sbi, true); 4419 truncate_inode_pages_final(NODE_MAPPING(sbi)); 4420 iput(sbi->node_inode); 4421 sbi->node_inode = NULL; 4422 free_stats: 4423 f2fs_destroy_stats(sbi); 4424 free_nm: 4425 /* stop discard thread before destroying node manager */ 4426 f2fs_stop_discard_thread(sbi); 4427 f2fs_destroy_node_manager(sbi); 4428 free_sm: 4429 f2fs_destroy_segment_manager(sbi); 4430 f2fs_destroy_post_read_wq(sbi); 4431 stop_ckpt_thread: 4432 f2fs_stop_ckpt_thread(sbi); 4433 free_devices: 4434 destroy_device_list(sbi); 4435 kvfree(sbi->ckpt); 4436 free_meta_inode: 4437 make_bad_inode(sbi->meta_inode); 4438 iput(sbi->meta_inode); 4439 sbi->meta_inode = NULL; 4440 free_page_array_cache: 4441 f2fs_destroy_page_array_cache(sbi); 4442 free_xattr_cache: 4443 f2fs_destroy_xattr_caches(sbi); 4444 free_io_dummy: 4445 mempool_destroy(sbi->write_io_dummy); 4446 free_percpu: 4447 destroy_percpu_info(sbi); 4448 free_iostat: 4449 f2fs_destroy_iostat(sbi); 4450 free_bio_info: 4451 for (i = 0; i < NR_PAGE_TYPE; i++) 4452 kvfree(sbi->write_io[i]); 4453 4454 #if IS_ENABLED(CONFIG_UNICODE) 4455 utf8_unload(sb->s_encoding); 4456 sb->s_encoding = NULL; 4457 #endif 4458 free_options: 4459 #ifdef CONFIG_QUOTA 4460 for (i = 0; i < MAXQUOTAS; i++) 4461 kfree(F2FS_OPTION(sbi).s_qf_names[i]); 4462 #endif 4463 fscrypt_free_dummy_policy(&F2FS_OPTION(sbi).dummy_enc_policy); 4464 kvfree(options); 4465 free_sb_buf: 4466 kfree(raw_super); 4467 free_sbi: 4468 if (sbi->s_chksum_driver) 4469 crypto_free_shash(sbi->s_chksum_driver); 4470 kfree(sbi); 4471 4472 /* give only one another chance */ 4473 if (retry_cnt > 0 && skip_recovery) { 4474 retry_cnt--; 4475 shrink_dcache_sb(sb); 4476 goto try_onemore; 4477 } 4478 return err; 4479 } 4480 4481 static struct dentry *f2fs_mount(struct file_system_type *fs_type, int flags, 4482 const char *dev_name, void *data) 4483 { 4484 return mount_bdev(fs_type, flags, dev_name, data, f2fs_fill_super); 4485 } 4486 4487 static void kill_f2fs_super(struct super_block *sb) 4488 { 4489 if (sb->s_root) { 4490 struct f2fs_sb_info *sbi = F2FS_SB(sb); 4491 4492 set_sbi_flag(sbi, SBI_IS_CLOSE); 4493 f2fs_stop_gc_thread(sbi); 4494 f2fs_stop_discard_thread(sbi); 4495 4496 #ifdef CONFIG_F2FS_FS_COMPRESSION 4497 /* 4498 * latter evict_inode() can bypass checking and invalidating 4499 * compress inode cache. 4500 */ 4501 if (test_opt(sbi, COMPRESS_CACHE)) 4502 truncate_inode_pages_final(COMPRESS_MAPPING(sbi)); 4503 #endif 4504 4505 if (is_sbi_flag_set(sbi, SBI_IS_DIRTY) || 4506 !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG)) { 4507 struct cp_control cpc = { 4508 .reason = CP_UMOUNT, 4509 }; 4510 f2fs_write_checkpoint(sbi, &cpc); 4511 } 4512 4513 if (is_sbi_flag_set(sbi, SBI_IS_RECOVERED) && f2fs_readonly(sb)) 4514 sb->s_flags &= ~SB_RDONLY; 4515 } 4516 kill_block_super(sb); 4517 } 4518 4519 static struct file_system_type f2fs_fs_type = { 4520 .owner = THIS_MODULE, 4521 .name = "f2fs", 4522 .mount = f2fs_mount, 4523 .kill_sb = kill_f2fs_super, 4524 .fs_flags = FS_REQUIRES_DEV | FS_ALLOW_IDMAP, 4525 }; 4526 MODULE_ALIAS_FS("f2fs"); 4527 4528 static int __init init_inodecache(void) 4529 { 4530 f2fs_inode_cachep = kmem_cache_create("f2fs_inode_cache", 4531 sizeof(struct f2fs_inode_info), 0, 4532 SLAB_RECLAIM_ACCOUNT|SLAB_ACCOUNT, NULL); 4533 if (!f2fs_inode_cachep) 4534 return -ENOMEM; 4535 return 0; 4536 } 4537 4538 static void destroy_inodecache(void) 4539 { 4540 /* 4541 * Make sure all delayed rcu free inodes are flushed before we 4542 * destroy cache. 4543 */ 4544 rcu_barrier(); 4545 kmem_cache_destroy(f2fs_inode_cachep); 4546 } 4547 4548 static int __init init_f2fs_fs(void) 4549 { 4550 int err; 4551 4552 if (PAGE_SIZE != F2FS_BLKSIZE) { 4553 printk("F2FS not supported on PAGE_SIZE(%lu) != %d\n", 4554 PAGE_SIZE, F2FS_BLKSIZE); 4555 return -EINVAL; 4556 } 4557 4558 err = init_inodecache(); 4559 if (err) 4560 goto fail; 4561 err = f2fs_create_node_manager_caches(); 4562 if (err) 4563 goto free_inodecache; 4564 err = f2fs_create_segment_manager_caches(); 4565 if (err) 4566 goto free_node_manager_caches; 4567 err = f2fs_create_checkpoint_caches(); 4568 if (err) 4569 goto free_segment_manager_caches; 4570 err = f2fs_create_recovery_cache(); 4571 if (err) 4572 goto free_checkpoint_caches; 4573 err = f2fs_create_extent_cache(); 4574 if (err) 4575 goto free_recovery_cache; 4576 err = f2fs_create_garbage_collection_cache(); 4577 if (err) 4578 goto free_extent_cache; 4579 err = f2fs_init_sysfs(); 4580 if (err) 4581 goto free_garbage_collection_cache; 4582 err = register_shrinker(&f2fs_shrinker_info); 4583 if (err) 4584 goto free_sysfs; 4585 err = register_filesystem(&f2fs_fs_type); 4586 if (err) 4587 goto free_shrinker; 4588 f2fs_create_root_stats(); 4589 err = f2fs_init_post_read_processing(); 4590 if (err) 4591 goto free_root_stats; 4592 err = f2fs_init_iostat_processing(); 4593 if (err) 4594 goto free_post_read; 4595 err = f2fs_init_bio_entry_cache(); 4596 if (err) 4597 goto free_iostat; 4598 err = f2fs_init_bioset(); 4599 if (err) 4600 goto free_bio_enrty_cache; 4601 err = f2fs_init_compress_mempool(); 4602 if (err) 4603 goto free_bioset; 4604 err = f2fs_init_compress_cache(); 4605 if (err) 4606 goto free_compress_mempool; 4607 err = f2fs_create_casefold_cache(); 4608 if (err) 4609 goto free_compress_cache; 4610 return 0; 4611 free_compress_cache: 4612 f2fs_destroy_compress_cache(); 4613 free_compress_mempool: 4614 f2fs_destroy_compress_mempool(); 4615 free_bioset: 4616 f2fs_destroy_bioset(); 4617 free_bio_enrty_cache: 4618 f2fs_destroy_bio_entry_cache(); 4619 free_iostat: 4620 f2fs_destroy_iostat_processing(); 4621 free_post_read: 4622 f2fs_destroy_post_read_processing(); 4623 free_root_stats: 4624 f2fs_destroy_root_stats(); 4625 unregister_filesystem(&f2fs_fs_type); 4626 free_shrinker: 4627 unregister_shrinker(&f2fs_shrinker_info); 4628 free_sysfs: 4629 f2fs_exit_sysfs(); 4630 free_garbage_collection_cache: 4631 f2fs_destroy_garbage_collection_cache(); 4632 free_extent_cache: 4633 f2fs_destroy_extent_cache(); 4634 free_recovery_cache: 4635 f2fs_destroy_recovery_cache(); 4636 free_checkpoint_caches: 4637 f2fs_destroy_checkpoint_caches(); 4638 free_segment_manager_caches: 4639 f2fs_destroy_segment_manager_caches(); 4640 free_node_manager_caches: 4641 f2fs_destroy_node_manager_caches(); 4642 free_inodecache: 4643 destroy_inodecache(); 4644 fail: 4645 return err; 4646 } 4647 4648 static void __exit exit_f2fs_fs(void) 4649 { 4650 f2fs_destroy_casefold_cache(); 4651 f2fs_destroy_compress_cache(); 4652 f2fs_destroy_compress_mempool(); 4653 f2fs_destroy_bioset(); 4654 f2fs_destroy_bio_entry_cache(); 4655 f2fs_destroy_iostat_processing(); 4656 f2fs_destroy_post_read_processing(); 4657 f2fs_destroy_root_stats(); 4658 unregister_filesystem(&f2fs_fs_type); 4659 unregister_shrinker(&f2fs_shrinker_info); 4660 f2fs_exit_sysfs(); 4661 f2fs_destroy_garbage_collection_cache(); 4662 f2fs_destroy_extent_cache(); 4663 f2fs_destroy_recovery_cache(); 4664 f2fs_destroy_checkpoint_caches(); 4665 f2fs_destroy_segment_manager_caches(); 4666 f2fs_destroy_node_manager_caches(); 4667 destroy_inodecache(); 4668 } 4669 4670 module_init(init_f2fs_fs) 4671 module_exit(exit_f2fs_fs) 4672 4673 MODULE_AUTHOR("Samsung Electronics's Praesto Team"); 4674 MODULE_DESCRIPTION("Flash Friendly File System"); 4675 MODULE_LICENSE("GPL"); 4676 MODULE_SOFTDEP("pre: crc32"); 4677 4678