1 // SPDX-License-Identifier: GPL-2.0 2 /* 3 * fs/f2fs/super.c 4 * 5 * Copyright (c) 2012 Samsung Electronics Co., Ltd. 6 * http://www.samsung.com/ 7 */ 8 #include <linux/module.h> 9 #include <linux/init.h> 10 #include <linux/fs.h> 11 #include <linux/fs_context.h> 12 #include <linux/statfs.h> 13 #include <linux/buffer_head.h> 14 #include <linux/backing-dev.h> 15 #include <linux/kthread.h> 16 #include <linux/parser.h> 17 #include <linux/mount.h> 18 #include <linux/seq_file.h> 19 #include <linux/proc_fs.h> 20 #include <linux/random.h> 21 #include <linux/exportfs.h> 22 #include <linux/blkdev.h> 23 #include <linux/quotaops.h> 24 #include <linux/f2fs_fs.h> 25 #include <linux/sysfs.h> 26 #include <linux/quota.h> 27 #include <linux/unicode.h> 28 #include <linux/part_stat.h> 29 #include <linux/zstd.h> 30 #include <linux/lz4.h> 31 #include <linux/cleancache.h> 32 33 #include "f2fs.h" 34 #include "node.h" 35 #include "segment.h" 36 #include "xattr.h" 37 #include "gc.h" 38 #include "iostat.h" 39 40 #define CREATE_TRACE_POINTS 41 #include <trace/events/f2fs.h> 42 43 static struct kmem_cache *f2fs_inode_cachep; 44 45 #ifdef CONFIG_F2FS_FAULT_INJECTION 46 47 const char *f2fs_fault_name[FAULT_MAX] = { 48 [FAULT_KMALLOC] = "kmalloc", 49 [FAULT_KVMALLOC] = "kvmalloc", 50 [FAULT_PAGE_ALLOC] = "page alloc", 51 [FAULT_PAGE_GET] = "page get", 52 [FAULT_ALLOC_NID] = "alloc nid", 53 [FAULT_ORPHAN] = "orphan", 54 [FAULT_BLOCK] = "no more block", 55 [FAULT_DIR_DEPTH] = "too big dir depth", 56 [FAULT_EVICT_INODE] = "evict_inode fail", 57 [FAULT_TRUNCATE] = "truncate fail", 58 [FAULT_READ_IO] = "read IO error", 59 [FAULT_CHECKPOINT] = "checkpoint error", 60 [FAULT_DISCARD] = "discard error", 61 [FAULT_WRITE_IO] = "write IO error", 62 [FAULT_SLAB_ALLOC] = "slab alloc", 63 [FAULT_DQUOT_INIT] = "dquot initialize", 64 [FAULT_LOCK_OP] = "lock_op", 65 [FAULT_BLKADDR] = "invalid blkaddr", 66 }; 67 f2fs_build_fault_attr(struct f2fs_sb_info * sbi,unsigned int rate,unsigned int type)68 void f2fs_build_fault_attr(struct f2fs_sb_info *sbi, unsigned int rate, 69 unsigned int type) 70 { 71 struct f2fs_fault_info *ffi = &F2FS_OPTION(sbi).fault_info; 72 73 if (rate) { 74 atomic_set(&ffi->inject_ops, 0); 75 ffi->inject_rate = rate; 76 } 77 78 if (type) 79 ffi->inject_type = type; 80 81 if (!rate && !type) 82 memset(ffi, 0, sizeof(struct f2fs_fault_info)); 83 } 84 #endif 85 86 /* f2fs-wide shrinker description */ 87 static struct shrinker f2fs_shrinker_info = { 88 .scan_objects = f2fs_shrink_scan, 89 .count_objects = f2fs_shrink_count, 90 .seeks = DEFAULT_SEEKS, 91 }; 92 93 enum { 94 Opt_gc_background, 95 Opt_disable_roll_forward, 96 Opt_norecovery, 97 Opt_discard, 98 Opt_nodiscard, 99 Opt_noheap, 100 Opt_heap, 101 Opt_user_xattr, 102 Opt_nouser_xattr, 103 Opt_acl, 104 Opt_noacl, 105 Opt_active_logs, 106 Opt_disable_ext_identify, 107 Opt_inline_xattr, 108 Opt_noinline_xattr, 109 Opt_inline_xattr_size, 110 Opt_inline_data, 111 Opt_inline_dentry, 112 Opt_noinline_dentry, 113 Opt_flush_merge, 114 Opt_noflush_merge, 115 Opt_barrier, 116 Opt_nobarrier, 117 Opt_fastboot, 118 Opt_extent_cache, 119 Opt_noextent_cache, 120 Opt_noinline_data, 121 Opt_data_flush, 122 Opt_reserve_root, 123 Opt_resgid, 124 Opt_resuid, 125 Opt_mode, 126 Opt_io_size_bits, 127 Opt_fault_injection, 128 Opt_fault_type, 129 Opt_lazytime, 130 Opt_nolazytime, 131 Opt_quota, 132 Opt_noquota, 133 Opt_usrquota, 134 Opt_grpquota, 135 Opt_prjquota, 136 Opt_usrjquota, 137 Opt_grpjquota, 138 Opt_prjjquota, 139 Opt_offusrjquota, 140 Opt_offgrpjquota, 141 Opt_offprjjquota, 142 Opt_jqfmt_vfsold, 143 Opt_jqfmt_vfsv0, 144 Opt_jqfmt_vfsv1, 145 Opt_alloc, 146 Opt_fsync, 147 Opt_test_dummy_encryption, 148 Opt_inlinecrypt, 149 Opt_checkpoint_disable, 150 Opt_checkpoint_disable_cap, 151 Opt_checkpoint_disable_cap_perc, 152 Opt_checkpoint_enable, 153 Opt_checkpoint_merge, 154 Opt_nocheckpoint_merge, 155 Opt_compress_algorithm, 156 Opt_compress_log_size, 157 Opt_compress_extension, 158 Opt_nocompress_extension, 159 Opt_compress_chksum, 160 Opt_compress_mode, 161 Opt_compress_cache, 162 Opt_atgc, 163 Opt_gc_merge, 164 Opt_nogc_merge, 165 Opt_discard_unit, 166 Opt_memory_mode, 167 Opt_age_extent_cache, 168 Opt_err, 169 }; 170 171 static match_table_t f2fs_tokens = { 172 {Opt_gc_background, "background_gc=%s"}, 173 {Opt_disable_roll_forward, "disable_roll_forward"}, 174 {Opt_norecovery, "norecovery"}, 175 {Opt_discard, "discard"}, 176 {Opt_nodiscard, "nodiscard"}, 177 {Opt_noheap, "no_heap"}, 178 {Opt_heap, "heap"}, 179 {Opt_user_xattr, "user_xattr"}, 180 {Opt_nouser_xattr, "nouser_xattr"}, 181 {Opt_acl, "acl"}, 182 {Opt_noacl, "noacl"}, 183 {Opt_active_logs, "active_logs=%u"}, 184 {Opt_disable_ext_identify, "disable_ext_identify"}, 185 {Opt_inline_xattr, "inline_xattr"}, 186 {Opt_noinline_xattr, "noinline_xattr"}, 187 {Opt_inline_xattr_size, "inline_xattr_size=%u"}, 188 {Opt_inline_data, "inline_data"}, 189 {Opt_inline_dentry, "inline_dentry"}, 190 {Opt_noinline_dentry, "noinline_dentry"}, 191 {Opt_flush_merge, "flush_merge"}, 192 {Opt_noflush_merge, "noflush_merge"}, 193 {Opt_barrier, "barrier"}, 194 {Opt_nobarrier, "nobarrier"}, 195 {Opt_fastboot, "fastboot"}, 196 {Opt_extent_cache, "extent_cache"}, 197 {Opt_noextent_cache, "noextent_cache"}, 198 {Opt_noinline_data, "noinline_data"}, 199 {Opt_data_flush, "data_flush"}, 200 {Opt_reserve_root, "reserve_root=%u"}, 201 {Opt_resgid, "resgid=%u"}, 202 {Opt_resuid, "resuid=%u"}, 203 {Opt_mode, "mode=%s"}, 204 {Opt_io_size_bits, "io_bits=%u"}, 205 {Opt_fault_injection, "fault_injection=%u"}, 206 {Opt_fault_type, "fault_type=%u"}, 207 {Opt_lazytime, "lazytime"}, 208 {Opt_nolazytime, "nolazytime"}, 209 {Opt_quota, "quota"}, 210 {Opt_noquota, "noquota"}, 211 {Opt_usrquota, "usrquota"}, 212 {Opt_grpquota, "grpquota"}, 213 {Opt_prjquota, "prjquota"}, 214 {Opt_usrjquota, "usrjquota=%s"}, 215 {Opt_grpjquota, "grpjquota=%s"}, 216 {Opt_prjjquota, "prjjquota=%s"}, 217 {Opt_offusrjquota, "usrjquota="}, 218 {Opt_offgrpjquota, "grpjquota="}, 219 {Opt_offprjjquota, "prjjquota="}, 220 {Opt_jqfmt_vfsold, "jqfmt=vfsold"}, 221 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"}, 222 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"}, 223 {Opt_alloc, "alloc_mode=%s"}, 224 {Opt_fsync, "fsync_mode=%s"}, 225 {Opt_test_dummy_encryption, "test_dummy_encryption=%s"}, 226 {Opt_test_dummy_encryption, "test_dummy_encryption"}, 227 {Opt_inlinecrypt, "inlinecrypt"}, 228 {Opt_checkpoint_disable, "checkpoint=disable"}, 229 {Opt_checkpoint_disable_cap, "checkpoint=disable:%u"}, 230 {Opt_checkpoint_disable_cap_perc, "checkpoint=disable:%u%%"}, 231 {Opt_checkpoint_enable, "checkpoint=enable"}, 232 {Opt_checkpoint_merge, "checkpoint_merge"}, 233 {Opt_nocheckpoint_merge, "nocheckpoint_merge"}, 234 {Opt_compress_algorithm, "compress_algorithm=%s"}, 235 {Opt_compress_log_size, "compress_log_size=%u"}, 236 {Opt_compress_extension, "compress_extension=%s"}, 237 {Opt_nocompress_extension, "nocompress_extension=%s"}, 238 {Opt_compress_chksum, "compress_chksum"}, 239 {Opt_compress_mode, "compress_mode=%s"}, 240 {Opt_compress_cache, "compress_cache"}, 241 {Opt_atgc, "atgc"}, 242 {Opt_gc_merge, "gc_merge"}, 243 {Opt_nogc_merge, "nogc_merge"}, 244 {Opt_discard_unit, "discard_unit=%s"}, 245 {Opt_memory_mode, "memory=%s"}, 246 {Opt_age_extent_cache, "age_extent_cache"}, 247 {Opt_err, NULL}, 248 }; 249 f2fs_printk(struct f2fs_sb_info * sbi,const char * fmt,...)250 void f2fs_printk(struct f2fs_sb_info *sbi, const char *fmt, ...) 251 { 252 struct va_format vaf; 253 va_list args; 254 int level; 255 256 va_start(args, fmt); 257 258 level = printk_get_level(fmt); 259 vaf.fmt = printk_skip_level(fmt); 260 vaf.va = &args; 261 printk("%c%cF2FS-fs (%s): %pV\n", 262 KERN_SOH_ASCII, level, sbi->sb->s_id, &vaf); 263 264 va_end(args); 265 } 266 267 #ifdef CONFIG_UNICODE 268 static const struct f2fs_sb_encodings { 269 __u16 magic; 270 char *name; 271 char *version; 272 } f2fs_sb_encoding_map[] = { 273 {F2FS_ENC_UTF8_12_1, "utf8", "12.1.0"}, 274 }; 275 f2fs_sb_read_encoding(const struct f2fs_super_block * sb,const struct f2fs_sb_encodings ** encoding,__u16 * flags)276 static int f2fs_sb_read_encoding(const struct f2fs_super_block *sb, 277 const struct f2fs_sb_encodings **encoding, 278 __u16 *flags) 279 { 280 __u16 magic = le16_to_cpu(sb->s_encoding); 281 int i; 282 283 for (i = 0; i < ARRAY_SIZE(f2fs_sb_encoding_map); i++) 284 if (magic == f2fs_sb_encoding_map[i].magic) 285 break; 286 287 if (i >= ARRAY_SIZE(f2fs_sb_encoding_map)) 288 return -EINVAL; 289 290 *encoding = &f2fs_sb_encoding_map[i]; 291 *flags = le16_to_cpu(sb->s_encoding_flags); 292 293 return 0; 294 } 295 296 struct kmem_cache *f2fs_cf_name_slab; f2fs_create_casefold_cache(void)297 static int __init f2fs_create_casefold_cache(void) 298 { 299 f2fs_cf_name_slab = f2fs_kmem_cache_create("f2fs_casefolded_name", 300 F2FS_NAME_LEN); 301 return f2fs_cf_name_slab ? 0 : -ENOMEM; 302 } 303 f2fs_destroy_casefold_cache(void)304 static void f2fs_destroy_casefold_cache(void) 305 { 306 kmem_cache_destroy(f2fs_cf_name_slab); 307 } 308 #else f2fs_create_casefold_cache(void)309 static int __init f2fs_create_casefold_cache(void) { return 0; } f2fs_destroy_casefold_cache(void)310 static void f2fs_destroy_casefold_cache(void) { } 311 #endif 312 limit_reserve_root(struct f2fs_sb_info * sbi)313 static inline void limit_reserve_root(struct f2fs_sb_info *sbi) 314 { 315 block_t limit = min((sbi->user_block_count >> 3), 316 sbi->user_block_count - sbi->reserved_blocks); 317 318 /* limit is 12.5% */ 319 if (test_opt(sbi, RESERVE_ROOT) && 320 F2FS_OPTION(sbi).root_reserved_blocks > limit) { 321 F2FS_OPTION(sbi).root_reserved_blocks = limit; 322 f2fs_info(sbi, "Reduce reserved blocks for root = %u", 323 F2FS_OPTION(sbi).root_reserved_blocks); 324 } 325 if (!test_opt(sbi, RESERVE_ROOT) && 326 (!uid_eq(F2FS_OPTION(sbi).s_resuid, 327 make_kuid(&init_user_ns, F2FS_DEF_RESUID)) || 328 !gid_eq(F2FS_OPTION(sbi).s_resgid, 329 make_kgid(&init_user_ns, F2FS_DEF_RESGID)))) 330 f2fs_info(sbi, "Ignore s_resuid=%u, s_resgid=%u w/o reserve_root", 331 from_kuid_munged(&init_user_ns, 332 F2FS_OPTION(sbi).s_resuid), 333 from_kgid_munged(&init_user_ns, 334 F2FS_OPTION(sbi).s_resgid)); 335 } 336 adjust_reserved_segment(struct f2fs_sb_info * sbi)337 static inline int adjust_reserved_segment(struct f2fs_sb_info *sbi) 338 { 339 unsigned int sec_blks = sbi->blocks_per_seg * sbi->segs_per_sec; 340 unsigned int avg_vblocks; 341 unsigned int wanted_reserved_segments; 342 block_t avail_user_block_count; 343 344 if (!F2FS_IO_ALIGNED(sbi)) 345 return 0; 346 347 /* average valid block count in section in worst case */ 348 avg_vblocks = sec_blks / F2FS_IO_SIZE(sbi); 349 350 /* 351 * we need enough free space when migrating one section in worst case 352 */ 353 wanted_reserved_segments = (F2FS_IO_SIZE(sbi) / avg_vblocks) * 354 reserved_segments(sbi); 355 wanted_reserved_segments -= reserved_segments(sbi); 356 357 avail_user_block_count = sbi->user_block_count - 358 sbi->current_reserved_blocks - 359 F2FS_OPTION(sbi).root_reserved_blocks; 360 361 if (wanted_reserved_segments * sbi->blocks_per_seg > 362 avail_user_block_count) { 363 f2fs_err(sbi, "IO align feature can't grab additional reserved segment: %u, available segments: %u", 364 wanted_reserved_segments, 365 avail_user_block_count >> sbi->log_blocks_per_seg); 366 return -ENOSPC; 367 } 368 369 SM_I(sbi)->additional_reserved_segments = wanted_reserved_segments; 370 371 f2fs_info(sbi, "IO align feature needs additional reserved segment: %u", 372 wanted_reserved_segments); 373 374 return 0; 375 } 376 adjust_unusable_cap_perc(struct f2fs_sb_info * sbi)377 static inline void adjust_unusable_cap_perc(struct f2fs_sb_info *sbi) 378 { 379 if (!F2FS_OPTION(sbi).unusable_cap_perc) 380 return; 381 382 if (F2FS_OPTION(sbi).unusable_cap_perc == 100) 383 F2FS_OPTION(sbi).unusable_cap = sbi->user_block_count; 384 else 385 F2FS_OPTION(sbi).unusable_cap = (sbi->user_block_count / 100) * 386 F2FS_OPTION(sbi).unusable_cap_perc; 387 388 f2fs_info(sbi, "Adjust unusable cap for checkpoint=disable = %u / %u%%", 389 F2FS_OPTION(sbi).unusable_cap, 390 F2FS_OPTION(sbi).unusable_cap_perc); 391 } 392 init_once(void * foo)393 static void init_once(void *foo) 394 { 395 struct f2fs_inode_info *fi = (struct f2fs_inode_info *) foo; 396 397 inode_init_once(&fi->vfs_inode); 398 } 399 400 #ifdef CONFIG_QUOTA 401 static const char * const quotatypes[] = INITQFNAMES; 402 #define QTYPE2NAME(t) (quotatypes[t]) f2fs_set_qf_name(struct super_block * sb,int qtype,substring_t * args)403 static int f2fs_set_qf_name(struct super_block *sb, int qtype, 404 substring_t *args) 405 { 406 struct f2fs_sb_info *sbi = F2FS_SB(sb); 407 char *qname; 408 int ret = -EINVAL; 409 410 if (sb_any_quota_loaded(sb) && !F2FS_OPTION(sbi).s_qf_names[qtype]) { 411 f2fs_err(sbi, "Cannot change journaled quota options when quota turned on"); 412 return -EINVAL; 413 } 414 if (f2fs_sb_has_quota_ino(sbi)) { 415 f2fs_info(sbi, "QUOTA feature is enabled, so ignore qf_name"); 416 return 0; 417 } 418 419 qname = match_strdup(args); 420 if (!qname) { 421 f2fs_err(sbi, "Not enough memory for storing quotafile name"); 422 return -ENOMEM; 423 } 424 if (F2FS_OPTION(sbi).s_qf_names[qtype]) { 425 if (strcmp(F2FS_OPTION(sbi).s_qf_names[qtype], qname) == 0) 426 ret = 0; 427 else 428 f2fs_err(sbi, "%s quota file already specified", 429 QTYPE2NAME(qtype)); 430 goto errout; 431 } 432 if (strchr(qname, '/')) { 433 f2fs_err(sbi, "quotafile must be on filesystem root"); 434 goto errout; 435 } 436 F2FS_OPTION(sbi).s_qf_names[qtype] = qname; 437 set_opt(sbi, QUOTA); 438 return 0; 439 errout: 440 kfree(qname); 441 return ret; 442 } 443 f2fs_clear_qf_name(struct super_block * sb,int qtype)444 static int f2fs_clear_qf_name(struct super_block *sb, int qtype) 445 { 446 struct f2fs_sb_info *sbi = F2FS_SB(sb); 447 448 if (sb_any_quota_loaded(sb) && F2FS_OPTION(sbi).s_qf_names[qtype]) { 449 f2fs_err(sbi, "Cannot change journaled quota options when quota turned on"); 450 return -EINVAL; 451 } 452 kfree(F2FS_OPTION(sbi).s_qf_names[qtype]); 453 F2FS_OPTION(sbi).s_qf_names[qtype] = NULL; 454 return 0; 455 } 456 f2fs_check_quota_options(struct f2fs_sb_info * sbi)457 static int f2fs_check_quota_options(struct f2fs_sb_info *sbi) 458 { 459 /* 460 * We do the test below only for project quotas. 'usrquota' and 461 * 'grpquota' mount options are allowed even without quota feature 462 * to support legacy quotas in quota files. 463 */ 464 if (test_opt(sbi, PRJQUOTA) && !f2fs_sb_has_project_quota(sbi)) { 465 f2fs_err(sbi, "Project quota feature not enabled. Cannot enable project quota enforcement."); 466 return -1; 467 } 468 if (F2FS_OPTION(sbi).s_qf_names[USRQUOTA] || 469 F2FS_OPTION(sbi).s_qf_names[GRPQUOTA] || 470 F2FS_OPTION(sbi).s_qf_names[PRJQUOTA]) { 471 if (test_opt(sbi, USRQUOTA) && 472 F2FS_OPTION(sbi).s_qf_names[USRQUOTA]) 473 clear_opt(sbi, USRQUOTA); 474 475 if (test_opt(sbi, GRPQUOTA) && 476 F2FS_OPTION(sbi).s_qf_names[GRPQUOTA]) 477 clear_opt(sbi, GRPQUOTA); 478 479 if (test_opt(sbi, PRJQUOTA) && 480 F2FS_OPTION(sbi).s_qf_names[PRJQUOTA]) 481 clear_opt(sbi, PRJQUOTA); 482 483 if (test_opt(sbi, GRPQUOTA) || test_opt(sbi, USRQUOTA) || 484 test_opt(sbi, PRJQUOTA)) { 485 f2fs_err(sbi, "old and new quota format mixing"); 486 return -1; 487 } 488 489 if (!F2FS_OPTION(sbi).s_jquota_fmt) { 490 f2fs_err(sbi, "journaled quota format not specified"); 491 return -1; 492 } 493 } 494 495 if (f2fs_sb_has_quota_ino(sbi) && F2FS_OPTION(sbi).s_jquota_fmt) { 496 f2fs_info(sbi, "QUOTA feature is enabled, so ignore jquota_fmt"); 497 F2FS_OPTION(sbi).s_jquota_fmt = 0; 498 } 499 return 0; 500 } 501 #endif 502 f2fs_set_test_dummy_encryption(struct super_block * sb,const char * opt,const substring_t * arg,bool is_remount)503 static int f2fs_set_test_dummy_encryption(struct super_block *sb, 504 const char *opt, 505 const substring_t *arg, 506 bool is_remount) 507 { 508 struct f2fs_sb_info *sbi = F2FS_SB(sb); 509 struct fs_parameter param = { 510 .type = fs_value_is_string, 511 .string = arg->from ? arg->from : "", 512 }; 513 struct fscrypt_dummy_policy *policy = 514 &F2FS_OPTION(sbi).dummy_enc_policy; 515 int err; 516 517 if (!IS_ENABLED(CONFIG_FS_ENCRYPTION)) { 518 f2fs_warn(sbi, "test_dummy_encryption option not supported"); 519 return -EINVAL; 520 } 521 522 if (!f2fs_sb_has_encrypt(sbi)) { 523 f2fs_err(sbi, "Encrypt feature is off"); 524 return -EINVAL; 525 } 526 527 /* 528 * This mount option is just for testing, and it's not worthwhile to 529 * implement the extra complexity (e.g. RCU protection) that would be 530 * needed to allow it to be set or changed during remount. We do allow 531 * it to be specified during remount, but only if there is no change. 532 */ 533 if (is_remount && !fscrypt_is_dummy_policy_set(policy)) { 534 f2fs_warn(sbi, "Can't set test_dummy_encryption on remount"); 535 return -EINVAL; 536 } 537 538 err = fscrypt_parse_test_dummy_encryption(¶m, policy); 539 if (err) { 540 if (err == -EEXIST) 541 f2fs_warn(sbi, 542 "Can't change test_dummy_encryption on remount"); 543 else if (err == -EINVAL) 544 f2fs_warn(sbi, "Value of option \"%s\" is unrecognized", 545 opt); 546 else 547 f2fs_warn(sbi, "Error processing option \"%s\" [%d]", 548 opt, err); 549 return -EINVAL; 550 } 551 f2fs_warn(sbi, "Test dummy encryption mode enabled"); 552 return 0; 553 } 554 555 #ifdef CONFIG_F2FS_FS_COMPRESSION is_compress_extension_exist(struct f2fs_sb_info * sbi,const char * new_ext,bool is_ext)556 static bool is_compress_extension_exist(struct f2fs_sb_info *sbi, 557 const char *new_ext, bool is_ext) 558 { 559 unsigned char (*ext)[F2FS_EXTENSION_LEN]; 560 int ext_cnt; 561 int i; 562 563 if (is_ext) { 564 ext = F2FS_OPTION(sbi).extensions; 565 ext_cnt = F2FS_OPTION(sbi).compress_ext_cnt; 566 } else { 567 ext = F2FS_OPTION(sbi).noextensions; 568 ext_cnt = F2FS_OPTION(sbi).nocompress_ext_cnt; 569 } 570 571 for (i = 0; i < ext_cnt; i++) { 572 if (!strcasecmp(new_ext, ext[i])) 573 return true; 574 } 575 576 return false; 577 } 578 579 /* 580 * 1. The same extension name cannot not appear in both compress and non-compress extension 581 * at the same time. 582 * 2. If the compress extension specifies all files, the types specified by the non-compress 583 * extension will be treated as special cases and will not be compressed. 584 * 3. Don't allow the non-compress extension specifies all files. 585 */ f2fs_test_compress_extension(struct f2fs_sb_info * sbi)586 static int f2fs_test_compress_extension(struct f2fs_sb_info *sbi) 587 { 588 unsigned char (*ext)[F2FS_EXTENSION_LEN]; 589 unsigned char (*noext)[F2FS_EXTENSION_LEN]; 590 int ext_cnt, noext_cnt, index = 0, no_index = 0; 591 592 ext = F2FS_OPTION(sbi).extensions; 593 ext_cnt = F2FS_OPTION(sbi).compress_ext_cnt; 594 noext = F2FS_OPTION(sbi).noextensions; 595 noext_cnt = F2FS_OPTION(sbi).nocompress_ext_cnt; 596 597 if (!noext_cnt) 598 return 0; 599 600 for (no_index = 0; no_index < noext_cnt; no_index++) { 601 if (!strcasecmp("*", noext[no_index])) { 602 f2fs_info(sbi, "Don't allow the nocompress extension specifies all files"); 603 return -EINVAL; 604 } 605 for (index = 0; index < ext_cnt; index++) { 606 if (!strcasecmp(ext[index], noext[no_index])) { 607 f2fs_info(sbi, "Don't allow the same extension %s appear in both compress and nocompress extension", 608 ext[index]); 609 return -EINVAL; 610 } 611 } 612 } 613 return 0; 614 } 615 616 #ifdef CONFIG_F2FS_FS_LZ4 f2fs_set_lz4hc_level(struct f2fs_sb_info * sbi,const char * str)617 static int f2fs_set_lz4hc_level(struct f2fs_sb_info *sbi, const char *str) 618 { 619 #ifdef CONFIG_F2FS_FS_LZ4HC 620 unsigned int level; 621 #endif 622 623 if (strlen(str) == 3) { 624 F2FS_OPTION(sbi).compress_level = 0; 625 return 0; 626 } 627 628 #ifdef CONFIG_F2FS_FS_LZ4HC 629 str += 3; 630 631 if (str[0] != ':') { 632 f2fs_info(sbi, "wrong format, e.g. <alg_name>:<compr_level>"); 633 return -EINVAL; 634 } 635 if (kstrtouint(str + 1, 10, &level)) 636 return -EINVAL; 637 638 if (level < LZ4HC_MIN_CLEVEL || level > LZ4HC_MAX_CLEVEL) { 639 f2fs_info(sbi, "invalid lz4hc compress level: %d", level); 640 return -EINVAL; 641 } 642 643 F2FS_OPTION(sbi).compress_level = level; 644 return 0; 645 #else 646 f2fs_info(sbi, "kernel doesn't support lz4hc compression"); 647 return -EINVAL; 648 #endif 649 } 650 #endif 651 652 #ifdef CONFIG_F2FS_FS_ZSTD f2fs_set_zstd_level(struct f2fs_sb_info * sbi,const char * str)653 static int f2fs_set_zstd_level(struct f2fs_sb_info *sbi, const char *str) 654 { 655 unsigned int level; 656 int len = 4; 657 658 if (strlen(str) == len) { 659 F2FS_OPTION(sbi).compress_level = 0; 660 return 0; 661 } 662 663 str += len; 664 665 if (str[0] != ':') { 666 f2fs_info(sbi, "wrong format, e.g. <alg_name>:<compr_level>"); 667 return -EINVAL; 668 } 669 if (kstrtouint(str + 1, 10, &level)) 670 return -EINVAL; 671 672 if (!level || level > ZSTD_maxCLevel()) { 673 f2fs_info(sbi, "invalid zstd compress level: %d", level); 674 return -EINVAL; 675 } 676 677 F2FS_OPTION(sbi).compress_level = level; 678 return 0; 679 } 680 #endif 681 #endif 682 parse_options(struct super_block * sb,char * options,bool is_remount)683 static int parse_options(struct super_block *sb, char *options, bool is_remount) 684 { 685 struct f2fs_sb_info *sbi = F2FS_SB(sb); 686 substring_t args[MAX_OPT_ARGS]; 687 #ifdef CONFIG_F2FS_FS_COMPRESSION 688 unsigned char (*ext)[F2FS_EXTENSION_LEN]; 689 unsigned char (*noext)[F2FS_EXTENSION_LEN]; 690 int ext_cnt, noext_cnt; 691 #endif 692 char *p, *name; 693 int arg = 0; 694 kuid_t uid; 695 kgid_t gid; 696 int ret; 697 698 if (!options) 699 goto default_check; 700 701 while ((p = strsep(&options, ",")) != NULL) { 702 int token; 703 704 if (!*p) 705 continue; 706 /* 707 * Initialize args struct so we know whether arg was 708 * found; some options take optional arguments. 709 */ 710 args[0].to = args[0].from = NULL; 711 token = match_token(p, f2fs_tokens, args); 712 713 switch (token) { 714 case Opt_gc_background: 715 name = match_strdup(&args[0]); 716 717 if (!name) 718 return -ENOMEM; 719 if (!strcmp(name, "on")) { 720 F2FS_OPTION(sbi).bggc_mode = BGGC_MODE_ON; 721 } else if (!strcmp(name, "off")) { 722 F2FS_OPTION(sbi).bggc_mode = BGGC_MODE_OFF; 723 } else if (!strcmp(name, "sync")) { 724 F2FS_OPTION(sbi).bggc_mode = BGGC_MODE_SYNC; 725 } else { 726 kfree(name); 727 return -EINVAL; 728 } 729 kfree(name); 730 break; 731 case Opt_disable_roll_forward: 732 set_opt(sbi, DISABLE_ROLL_FORWARD); 733 break; 734 case Opt_norecovery: 735 /* this option mounts f2fs with ro */ 736 set_opt(sbi, NORECOVERY); 737 if (!f2fs_readonly(sb)) 738 return -EINVAL; 739 break; 740 case Opt_discard: 741 if (!f2fs_hw_support_discard(sbi)) { 742 f2fs_warn(sbi, "device does not support discard"); 743 break; 744 } 745 set_opt(sbi, DISCARD); 746 break; 747 case Opt_nodiscard: 748 if (f2fs_hw_should_discard(sbi)) { 749 f2fs_warn(sbi, "discard is required for zoned block devices"); 750 return -EINVAL; 751 } 752 clear_opt(sbi, DISCARD); 753 break; 754 case Opt_noheap: 755 set_opt(sbi, NOHEAP); 756 break; 757 case Opt_heap: 758 clear_opt(sbi, NOHEAP); 759 break; 760 #ifdef CONFIG_F2FS_FS_XATTR 761 case Opt_user_xattr: 762 set_opt(sbi, XATTR_USER); 763 break; 764 case Opt_nouser_xattr: 765 clear_opt(sbi, XATTR_USER); 766 break; 767 case Opt_inline_xattr: 768 set_opt(sbi, INLINE_XATTR); 769 break; 770 case Opt_noinline_xattr: 771 clear_opt(sbi, INLINE_XATTR); 772 break; 773 case Opt_inline_xattr_size: 774 if (args->from && match_int(args, &arg)) 775 return -EINVAL; 776 set_opt(sbi, INLINE_XATTR_SIZE); 777 F2FS_OPTION(sbi).inline_xattr_size = arg; 778 break; 779 #else 780 case Opt_user_xattr: 781 f2fs_info(sbi, "user_xattr options not supported"); 782 break; 783 case Opt_nouser_xattr: 784 f2fs_info(sbi, "nouser_xattr options not supported"); 785 break; 786 case Opt_inline_xattr: 787 f2fs_info(sbi, "inline_xattr options not supported"); 788 break; 789 case Opt_noinline_xattr: 790 f2fs_info(sbi, "noinline_xattr options not supported"); 791 break; 792 #endif 793 #ifdef CONFIG_F2FS_FS_POSIX_ACL 794 case Opt_acl: 795 set_opt(sbi, POSIX_ACL); 796 break; 797 case Opt_noacl: 798 clear_opt(sbi, POSIX_ACL); 799 break; 800 #else 801 case Opt_acl: 802 f2fs_info(sbi, "acl options not supported"); 803 break; 804 case Opt_noacl: 805 f2fs_info(sbi, "noacl options not supported"); 806 break; 807 #endif 808 case Opt_active_logs: 809 if (args->from && match_int(args, &arg)) 810 return -EINVAL; 811 if (arg != 2 && arg != 4 && 812 arg != NR_CURSEG_PERSIST_TYPE) 813 return -EINVAL; 814 F2FS_OPTION(sbi).active_logs = arg; 815 break; 816 case Opt_disable_ext_identify: 817 set_opt(sbi, DISABLE_EXT_IDENTIFY); 818 break; 819 case Opt_inline_data: 820 set_opt(sbi, INLINE_DATA); 821 break; 822 case Opt_inline_dentry: 823 set_opt(sbi, INLINE_DENTRY); 824 break; 825 case Opt_noinline_dentry: 826 clear_opt(sbi, INLINE_DENTRY); 827 break; 828 case Opt_flush_merge: 829 set_opt(sbi, FLUSH_MERGE); 830 break; 831 case Opt_noflush_merge: 832 clear_opt(sbi, FLUSH_MERGE); 833 break; 834 case Opt_nobarrier: 835 set_opt(sbi, NOBARRIER); 836 break; 837 case Opt_barrier: 838 clear_opt(sbi, NOBARRIER); 839 break; 840 case Opt_fastboot: 841 set_opt(sbi, FASTBOOT); 842 break; 843 case Opt_extent_cache: 844 set_opt(sbi, READ_EXTENT_CACHE); 845 break; 846 case Opt_noextent_cache: 847 clear_opt(sbi, READ_EXTENT_CACHE); 848 break; 849 case Opt_noinline_data: 850 clear_opt(sbi, INLINE_DATA); 851 break; 852 case Opt_data_flush: 853 set_opt(sbi, DATA_FLUSH); 854 break; 855 case Opt_reserve_root: 856 if (args->from && match_int(args, &arg)) 857 return -EINVAL; 858 if (test_opt(sbi, RESERVE_ROOT)) { 859 f2fs_info(sbi, "Preserve previous reserve_root=%u", 860 F2FS_OPTION(sbi).root_reserved_blocks); 861 } else { 862 F2FS_OPTION(sbi).root_reserved_blocks = arg; 863 set_opt(sbi, RESERVE_ROOT); 864 } 865 break; 866 case Opt_resuid: 867 if (args->from && match_int(args, &arg)) 868 return -EINVAL; 869 uid = make_kuid(current_user_ns(), arg); 870 if (!uid_valid(uid)) { 871 f2fs_err(sbi, "Invalid uid value %d", arg); 872 return -EINVAL; 873 } 874 F2FS_OPTION(sbi).s_resuid = uid; 875 break; 876 case Opt_resgid: 877 if (args->from && match_int(args, &arg)) 878 return -EINVAL; 879 gid = make_kgid(current_user_ns(), arg); 880 if (!gid_valid(gid)) { 881 f2fs_err(sbi, "Invalid gid value %d", arg); 882 return -EINVAL; 883 } 884 F2FS_OPTION(sbi).s_resgid = gid; 885 break; 886 case Opt_mode: 887 name = match_strdup(&args[0]); 888 889 if (!name) 890 return -ENOMEM; 891 if (!strcmp(name, "adaptive")) { 892 if (f2fs_sb_has_blkzoned(sbi)) { 893 f2fs_warn(sbi, "adaptive mode is not allowed with zoned block device feature"); 894 kfree(name); 895 return -EINVAL; 896 } 897 F2FS_OPTION(sbi).fs_mode = FS_MODE_ADAPTIVE; 898 } else if (!strcmp(name, "lfs")) { 899 F2FS_OPTION(sbi).fs_mode = FS_MODE_LFS; 900 } else if (!strcmp(name, "fragment:segment")) { 901 F2FS_OPTION(sbi).fs_mode = FS_MODE_FRAGMENT_SEG; 902 } else if (!strcmp(name, "fragment:block")) { 903 F2FS_OPTION(sbi).fs_mode = FS_MODE_FRAGMENT_BLK; 904 } else { 905 kfree(name); 906 return -EINVAL; 907 } 908 kfree(name); 909 break; 910 case Opt_io_size_bits: 911 if (args->from && match_int(args, &arg)) 912 return -EINVAL; 913 if (arg <= 0 || arg > __ilog2_u32(BIO_MAX_VECS)) { 914 f2fs_warn(sbi, "Not support %ld, larger than %d", 915 BIT(arg), BIO_MAX_VECS); 916 return -EINVAL; 917 } 918 F2FS_OPTION(sbi).write_io_size_bits = arg; 919 break; 920 #ifdef CONFIG_F2FS_FAULT_INJECTION 921 case Opt_fault_injection: 922 if (args->from && match_int(args, &arg)) 923 return -EINVAL; 924 f2fs_build_fault_attr(sbi, arg, F2FS_ALL_FAULT_TYPE); 925 set_opt(sbi, FAULT_INJECTION); 926 break; 927 928 case Opt_fault_type: 929 if (args->from && match_int(args, &arg)) 930 return -EINVAL; 931 f2fs_build_fault_attr(sbi, 0, arg); 932 set_opt(sbi, FAULT_INJECTION); 933 break; 934 #else 935 case Opt_fault_injection: 936 f2fs_info(sbi, "fault_injection options not supported"); 937 break; 938 939 case Opt_fault_type: 940 f2fs_info(sbi, "fault_type options not supported"); 941 break; 942 #endif 943 case Opt_lazytime: 944 sb->s_flags |= SB_LAZYTIME; 945 break; 946 case Opt_nolazytime: 947 sb->s_flags &= ~SB_LAZYTIME; 948 break; 949 #ifdef CONFIG_QUOTA 950 case Opt_quota: 951 case Opt_usrquota: 952 set_opt(sbi, USRQUOTA); 953 break; 954 case Opt_grpquota: 955 set_opt(sbi, GRPQUOTA); 956 break; 957 case Opt_prjquota: 958 set_opt(sbi, PRJQUOTA); 959 break; 960 case Opt_usrjquota: 961 ret = f2fs_set_qf_name(sb, USRQUOTA, &args[0]); 962 if (ret) 963 return ret; 964 break; 965 case Opt_grpjquota: 966 ret = f2fs_set_qf_name(sb, GRPQUOTA, &args[0]); 967 if (ret) 968 return ret; 969 break; 970 case Opt_prjjquota: 971 ret = f2fs_set_qf_name(sb, PRJQUOTA, &args[0]); 972 if (ret) 973 return ret; 974 break; 975 case Opt_offusrjquota: 976 ret = f2fs_clear_qf_name(sb, USRQUOTA); 977 if (ret) 978 return ret; 979 break; 980 case Opt_offgrpjquota: 981 ret = f2fs_clear_qf_name(sb, GRPQUOTA); 982 if (ret) 983 return ret; 984 break; 985 case Opt_offprjjquota: 986 ret = f2fs_clear_qf_name(sb, PRJQUOTA); 987 if (ret) 988 return ret; 989 break; 990 case Opt_jqfmt_vfsold: 991 F2FS_OPTION(sbi).s_jquota_fmt = QFMT_VFS_OLD; 992 break; 993 case Opt_jqfmt_vfsv0: 994 F2FS_OPTION(sbi).s_jquota_fmt = QFMT_VFS_V0; 995 break; 996 case Opt_jqfmt_vfsv1: 997 F2FS_OPTION(sbi).s_jquota_fmt = QFMT_VFS_V1; 998 break; 999 case Opt_noquota: 1000 clear_opt(sbi, QUOTA); 1001 clear_opt(sbi, USRQUOTA); 1002 clear_opt(sbi, GRPQUOTA); 1003 clear_opt(sbi, PRJQUOTA); 1004 break; 1005 #else 1006 case Opt_quota: 1007 case Opt_usrquota: 1008 case Opt_grpquota: 1009 case Opt_prjquota: 1010 case Opt_usrjquota: 1011 case Opt_grpjquota: 1012 case Opt_prjjquota: 1013 case Opt_offusrjquota: 1014 case Opt_offgrpjquota: 1015 case Opt_offprjjquota: 1016 case Opt_jqfmt_vfsold: 1017 case Opt_jqfmt_vfsv0: 1018 case Opt_jqfmt_vfsv1: 1019 case Opt_noquota: 1020 f2fs_info(sbi, "quota operations not supported"); 1021 break; 1022 #endif 1023 case Opt_alloc: 1024 name = match_strdup(&args[0]); 1025 if (!name) 1026 return -ENOMEM; 1027 1028 if (!strcmp(name, "default")) { 1029 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_DEFAULT; 1030 } else if (!strcmp(name, "reuse")) { 1031 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_REUSE; 1032 } else { 1033 kfree(name); 1034 return -EINVAL; 1035 } 1036 kfree(name); 1037 break; 1038 case Opt_fsync: 1039 name = match_strdup(&args[0]); 1040 if (!name) 1041 return -ENOMEM; 1042 if (!strcmp(name, "posix")) { 1043 F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_POSIX; 1044 } else if (!strcmp(name, "strict")) { 1045 F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_STRICT; 1046 } else if (!strcmp(name, "nobarrier")) { 1047 F2FS_OPTION(sbi).fsync_mode = 1048 FSYNC_MODE_NOBARRIER; 1049 } else { 1050 kfree(name); 1051 return -EINVAL; 1052 } 1053 kfree(name); 1054 break; 1055 case Opt_test_dummy_encryption: 1056 ret = f2fs_set_test_dummy_encryption(sb, p, &args[0], 1057 is_remount); 1058 if (ret) 1059 return ret; 1060 break; 1061 case Opt_inlinecrypt: 1062 #ifdef CONFIG_FS_ENCRYPTION_INLINE_CRYPT 1063 sb->s_flags |= SB_INLINECRYPT; 1064 #else 1065 f2fs_info(sbi, "inline encryption not supported"); 1066 #endif 1067 break; 1068 case Opt_checkpoint_disable_cap_perc: 1069 if (args->from && match_int(args, &arg)) 1070 return -EINVAL; 1071 if (arg < 0 || arg > 100) 1072 return -EINVAL; 1073 F2FS_OPTION(sbi).unusable_cap_perc = arg; 1074 set_opt(sbi, DISABLE_CHECKPOINT); 1075 break; 1076 case Opt_checkpoint_disable_cap: 1077 if (args->from && match_int(args, &arg)) 1078 return -EINVAL; 1079 F2FS_OPTION(sbi).unusable_cap = arg; 1080 set_opt(sbi, DISABLE_CHECKPOINT); 1081 break; 1082 case Opt_checkpoint_disable: 1083 set_opt(sbi, DISABLE_CHECKPOINT); 1084 break; 1085 case Opt_checkpoint_enable: 1086 clear_opt(sbi, DISABLE_CHECKPOINT); 1087 break; 1088 case Opt_checkpoint_merge: 1089 set_opt(sbi, MERGE_CHECKPOINT); 1090 break; 1091 case Opt_nocheckpoint_merge: 1092 clear_opt(sbi, MERGE_CHECKPOINT); 1093 break; 1094 #ifdef CONFIG_F2FS_FS_COMPRESSION 1095 case Opt_compress_algorithm: 1096 if (!f2fs_sb_has_compression(sbi)) { 1097 f2fs_info(sbi, "Image doesn't support compression"); 1098 break; 1099 } 1100 name = match_strdup(&args[0]); 1101 if (!name) 1102 return -ENOMEM; 1103 if (!strcmp(name, "lzo")) { 1104 #ifdef CONFIG_F2FS_FS_LZO 1105 F2FS_OPTION(sbi).compress_level = 0; 1106 F2FS_OPTION(sbi).compress_algorithm = 1107 COMPRESS_LZO; 1108 #else 1109 f2fs_info(sbi, "kernel doesn't support lzo compression"); 1110 #endif 1111 } else if (!strncmp(name, "lz4", 3)) { 1112 #ifdef CONFIG_F2FS_FS_LZ4 1113 ret = f2fs_set_lz4hc_level(sbi, name); 1114 if (ret) { 1115 kfree(name); 1116 return -EINVAL; 1117 } 1118 F2FS_OPTION(sbi).compress_algorithm = 1119 COMPRESS_LZ4; 1120 #else 1121 f2fs_info(sbi, "kernel doesn't support lz4 compression"); 1122 #endif 1123 } else if (!strncmp(name, "zstd", 4)) { 1124 #ifdef CONFIG_F2FS_FS_ZSTD 1125 ret = f2fs_set_zstd_level(sbi, name); 1126 if (ret) { 1127 kfree(name); 1128 return -EINVAL; 1129 } 1130 F2FS_OPTION(sbi).compress_algorithm = 1131 COMPRESS_ZSTD; 1132 #else 1133 f2fs_info(sbi, "kernel doesn't support zstd compression"); 1134 #endif 1135 } else if (!strcmp(name, "lzo-rle")) { 1136 #ifdef CONFIG_F2FS_FS_LZORLE 1137 F2FS_OPTION(sbi).compress_level = 0; 1138 F2FS_OPTION(sbi).compress_algorithm = 1139 COMPRESS_LZORLE; 1140 #else 1141 f2fs_info(sbi, "kernel doesn't support lzorle compression"); 1142 #endif 1143 } else { 1144 kfree(name); 1145 return -EINVAL; 1146 } 1147 kfree(name); 1148 break; 1149 case Opt_compress_log_size: 1150 if (!f2fs_sb_has_compression(sbi)) { 1151 f2fs_info(sbi, "Image doesn't support compression"); 1152 break; 1153 } 1154 if (args->from && match_int(args, &arg)) 1155 return -EINVAL; 1156 if (arg < MIN_COMPRESS_LOG_SIZE || 1157 arg > MAX_COMPRESS_LOG_SIZE) { 1158 f2fs_err(sbi, 1159 "Compress cluster log size is out of range"); 1160 return -EINVAL; 1161 } 1162 F2FS_OPTION(sbi).compress_log_size = arg; 1163 break; 1164 case Opt_compress_extension: 1165 if (!f2fs_sb_has_compression(sbi)) { 1166 f2fs_info(sbi, "Image doesn't support compression"); 1167 break; 1168 } 1169 name = match_strdup(&args[0]); 1170 if (!name) 1171 return -ENOMEM; 1172 1173 ext = F2FS_OPTION(sbi).extensions; 1174 ext_cnt = F2FS_OPTION(sbi).compress_ext_cnt; 1175 1176 if (strlen(name) >= F2FS_EXTENSION_LEN || 1177 ext_cnt >= COMPRESS_EXT_NUM) { 1178 f2fs_err(sbi, 1179 "invalid extension length/number"); 1180 kfree(name); 1181 return -EINVAL; 1182 } 1183 1184 if (is_compress_extension_exist(sbi, name, true)) { 1185 kfree(name); 1186 break; 1187 } 1188 1189 strcpy(ext[ext_cnt], name); 1190 F2FS_OPTION(sbi).compress_ext_cnt++; 1191 kfree(name); 1192 break; 1193 case Opt_nocompress_extension: 1194 if (!f2fs_sb_has_compression(sbi)) { 1195 f2fs_info(sbi, "Image doesn't support compression"); 1196 break; 1197 } 1198 name = match_strdup(&args[0]); 1199 if (!name) 1200 return -ENOMEM; 1201 1202 noext = F2FS_OPTION(sbi).noextensions; 1203 noext_cnt = F2FS_OPTION(sbi).nocompress_ext_cnt; 1204 1205 if (strlen(name) >= F2FS_EXTENSION_LEN || 1206 noext_cnt >= COMPRESS_EXT_NUM) { 1207 f2fs_err(sbi, 1208 "invalid extension length/number"); 1209 kfree(name); 1210 return -EINVAL; 1211 } 1212 1213 if (is_compress_extension_exist(sbi, name, false)) { 1214 kfree(name); 1215 break; 1216 } 1217 1218 strcpy(noext[noext_cnt], name); 1219 F2FS_OPTION(sbi).nocompress_ext_cnt++; 1220 kfree(name); 1221 break; 1222 case Opt_compress_chksum: 1223 if (!f2fs_sb_has_compression(sbi)) { 1224 f2fs_info(sbi, "Image doesn't support compression"); 1225 break; 1226 } 1227 F2FS_OPTION(sbi).compress_chksum = true; 1228 break; 1229 case Opt_compress_mode: 1230 if (!f2fs_sb_has_compression(sbi)) { 1231 f2fs_info(sbi, "Image doesn't support compression"); 1232 break; 1233 } 1234 name = match_strdup(&args[0]); 1235 if (!name) 1236 return -ENOMEM; 1237 if (!strcmp(name, "fs")) { 1238 F2FS_OPTION(sbi).compress_mode = COMPR_MODE_FS; 1239 } else if (!strcmp(name, "user")) { 1240 F2FS_OPTION(sbi).compress_mode = COMPR_MODE_USER; 1241 } else { 1242 kfree(name); 1243 return -EINVAL; 1244 } 1245 kfree(name); 1246 break; 1247 case Opt_compress_cache: 1248 if (!f2fs_sb_has_compression(sbi)) { 1249 f2fs_info(sbi, "Image doesn't support compression"); 1250 break; 1251 } 1252 set_opt(sbi, COMPRESS_CACHE); 1253 break; 1254 #else 1255 case Opt_compress_algorithm: 1256 case Opt_compress_log_size: 1257 case Opt_compress_extension: 1258 case Opt_nocompress_extension: 1259 case Opt_compress_chksum: 1260 case Opt_compress_mode: 1261 case Opt_compress_cache: 1262 f2fs_info(sbi, "compression options not supported"); 1263 break; 1264 #endif 1265 case Opt_atgc: 1266 set_opt(sbi, ATGC); 1267 break; 1268 case Opt_gc_merge: 1269 set_opt(sbi, GC_MERGE); 1270 break; 1271 case Opt_nogc_merge: 1272 clear_opt(sbi, GC_MERGE); 1273 break; 1274 case Opt_discard_unit: 1275 name = match_strdup(&args[0]); 1276 if (!name) 1277 return -ENOMEM; 1278 if (!strcmp(name, "block")) { 1279 F2FS_OPTION(sbi).discard_unit = 1280 DISCARD_UNIT_BLOCK; 1281 } else if (!strcmp(name, "segment")) { 1282 F2FS_OPTION(sbi).discard_unit = 1283 DISCARD_UNIT_SEGMENT; 1284 } else if (!strcmp(name, "section")) { 1285 F2FS_OPTION(sbi).discard_unit = 1286 DISCARD_UNIT_SECTION; 1287 } else { 1288 kfree(name); 1289 return -EINVAL; 1290 } 1291 kfree(name); 1292 break; 1293 case Opt_memory_mode: 1294 name = match_strdup(&args[0]); 1295 if (!name) 1296 return -ENOMEM; 1297 if (!strcmp(name, "normal")) { 1298 F2FS_OPTION(sbi).memory_mode = 1299 MEMORY_MODE_NORMAL; 1300 } else if (!strcmp(name, "low")) { 1301 F2FS_OPTION(sbi).memory_mode = 1302 MEMORY_MODE_LOW; 1303 } else { 1304 kfree(name); 1305 return -EINVAL; 1306 } 1307 kfree(name); 1308 break; 1309 case Opt_age_extent_cache: 1310 set_opt(sbi, AGE_EXTENT_CACHE); 1311 break; 1312 default: 1313 f2fs_err(sbi, "Unrecognized mount option \"%s\" or missing value", 1314 p); 1315 return -EINVAL; 1316 } 1317 } 1318 default_check: 1319 #ifdef CONFIG_QUOTA 1320 if (f2fs_check_quota_options(sbi)) 1321 return -EINVAL; 1322 #else 1323 if (f2fs_sb_has_quota_ino(sbi) && !f2fs_readonly(sbi->sb)) { 1324 f2fs_info(sbi, "Filesystem with quota feature cannot be mounted RDWR without CONFIG_QUOTA"); 1325 return -EINVAL; 1326 } 1327 if (f2fs_sb_has_project_quota(sbi) && !f2fs_readonly(sbi->sb)) { 1328 f2fs_err(sbi, "Filesystem with project quota feature cannot be mounted RDWR without CONFIG_QUOTA"); 1329 return -EINVAL; 1330 } 1331 #endif 1332 #ifndef CONFIG_UNICODE 1333 if (f2fs_sb_has_casefold(sbi)) { 1334 f2fs_err(sbi, 1335 "Filesystem with casefold feature cannot be mounted without CONFIG_UNICODE"); 1336 return -EINVAL; 1337 } 1338 #endif 1339 /* 1340 * The BLKZONED feature indicates that the drive was formatted with 1341 * zone alignment optimization. This is optional for host-aware 1342 * devices, but mandatory for host-managed zoned block devices. 1343 */ 1344 if (f2fs_sb_has_blkzoned(sbi)) { 1345 #ifdef CONFIG_BLK_DEV_ZONED 1346 if (F2FS_OPTION(sbi).discard_unit != 1347 DISCARD_UNIT_SECTION) { 1348 f2fs_info(sbi, "Zoned block device doesn't need small discard, set discard_unit=section by default"); 1349 F2FS_OPTION(sbi).discard_unit = 1350 DISCARD_UNIT_SECTION; 1351 } 1352 #else 1353 f2fs_err(sbi, "Zoned block device support is not enabled"); 1354 return -EINVAL; 1355 #endif 1356 } 1357 1358 #ifdef CONFIG_F2FS_FS_COMPRESSION 1359 if (f2fs_test_compress_extension(sbi)) { 1360 f2fs_err(sbi, "invalid compress or nocompress extension"); 1361 return -EINVAL; 1362 } 1363 #endif 1364 1365 if (F2FS_IO_SIZE_BITS(sbi) && !f2fs_lfs_mode(sbi)) { 1366 f2fs_err(sbi, "Should set mode=lfs with %luKB-sized IO", 1367 F2FS_IO_SIZE_KB(sbi)); 1368 return -EINVAL; 1369 } 1370 1371 if (test_opt(sbi, INLINE_XATTR_SIZE)) { 1372 int min_size, max_size; 1373 1374 if (!f2fs_sb_has_extra_attr(sbi) || 1375 !f2fs_sb_has_flexible_inline_xattr(sbi)) { 1376 f2fs_err(sbi, "extra_attr or flexible_inline_xattr feature is off"); 1377 return -EINVAL; 1378 } 1379 if (!test_opt(sbi, INLINE_XATTR)) { 1380 f2fs_err(sbi, "inline_xattr_size option should be set with inline_xattr option"); 1381 return -EINVAL; 1382 } 1383 1384 min_size = sizeof(struct f2fs_xattr_header) / sizeof(__le32); 1385 max_size = MAX_INLINE_XATTR_SIZE; 1386 1387 if (F2FS_OPTION(sbi).inline_xattr_size < min_size || 1388 F2FS_OPTION(sbi).inline_xattr_size > max_size) { 1389 f2fs_err(sbi, "inline xattr size is out of range: %d ~ %d", 1390 min_size, max_size); 1391 return -EINVAL; 1392 } 1393 } 1394 1395 if (test_opt(sbi, DISABLE_CHECKPOINT) && f2fs_lfs_mode(sbi)) { 1396 f2fs_err(sbi, "LFS is not compatible with checkpoint=disable"); 1397 return -EINVAL; 1398 } 1399 1400 if (test_opt(sbi, ATGC) && f2fs_lfs_mode(sbi)) { 1401 f2fs_err(sbi, "LFS is not compatible with ATGC"); 1402 return -EINVAL; 1403 } 1404 1405 if (f2fs_is_readonly(sbi) && test_opt(sbi, FLUSH_MERGE)) { 1406 f2fs_err(sbi, "FLUSH_MERGE not compatible with readonly mode"); 1407 return -EINVAL; 1408 } 1409 1410 if (f2fs_sb_has_readonly(sbi) && !f2fs_readonly(sbi->sb)) { 1411 f2fs_err(sbi, "Allow to mount readonly mode only"); 1412 return -EROFS; 1413 } 1414 return 0; 1415 } 1416 f2fs_alloc_inode(struct super_block * sb)1417 static struct inode *f2fs_alloc_inode(struct super_block *sb) 1418 { 1419 struct f2fs_inode_info *fi; 1420 1421 if (time_to_inject(F2FS_SB(sb), FAULT_SLAB_ALLOC)) 1422 return NULL; 1423 1424 fi = f2fs_kmem_cache_alloc(f2fs_inode_cachep, 1425 GFP_F2FS_ZERO, false, F2FS_SB(sb)); 1426 if (!fi) 1427 return NULL; 1428 1429 init_once((void *) fi); 1430 1431 /* Initialize f2fs-specific inode info */ 1432 atomic_set(&fi->dirty_pages, 0); 1433 atomic_set(&fi->i_compr_blocks, 0); 1434 init_f2fs_rwsem(&fi->i_sem); 1435 spin_lock_init(&fi->i_size_lock); 1436 INIT_LIST_HEAD(&fi->dirty_list); 1437 INIT_LIST_HEAD(&fi->gdirty_list); 1438 init_f2fs_rwsem(&fi->i_gc_rwsem[READ]); 1439 init_f2fs_rwsem(&fi->i_gc_rwsem[WRITE]); 1440 init_f2fs_rwsem(&fi->i_xattr_sem); 1441 1442 /* Will be used by directory only */ 1443 fi->i_dir_level = F2FS_SB(sb)->dir_level; 1444 1445 return &fi->vfs_inode; 1446 } 1447 f2fs_drop_inode(struct inode * inode)1448 static int f2fs_drop_inode(struct inode *inode) 1449 { 1450 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1451 int ret; 1452 1453 /* 1454 * during filesystem shutdown, if checkpoint is disabled, 1455 * drop useless meta/node dirty pages. 1456 */ 1457 if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED))) { 1458 if (inode->i_ino == F2FS_NODE_INO(sbi) || 1459 inode->i_ino == F2FS_META_INO(sbi)) { 1460 trace_f2fs_drop_inode(inode, 1); 1461 return 1; 1462 } 1463 } 1464 1465 /* 1466 * This is to avoid a deadlock condition like below. 1467 * writeback_single_inode(inode) 1468 * - f2fs_write_data_page 1469 * - f2fs_gc -> iput -> evict 1470 * - inode_wait_for_writeback(inode) 1471 */ 1472 if ((!inode_unhashed(inode) && inode->i_state & I_SYNC)) { 1473 if (!inode->i_nlink && !is_bad_inode(inode)) { 1474 /* to avoid evict_inode call simultaneously */ 1475 atomic_inc(&inode->i_count); 1476 spin_unlock(&inode->i_lock); 1477 1478 /* should remain fi->extent_tree for writepage */ 1479 f2fs_destroy_extent_node(inode); 1480 1481 sb_start_intwrite(inode->i_sb); 1482 f2fs_i_size_write(inode, 0); 1483 1484 f2fs_submit_merged_write_cond(F2FS_I_SB(inode), 1485 inode, NULL, 0, DATA); 1486 truncate_inode_pages_final(inode->i_mapping); 1487 1488 if (F2FS_HAS_BLOCKS(inode)) 1489 f2fs_truncate(inode); 1490 1491 sb_end_intwrite(inode->i_sb); 1492 1493 spin_lock(&inode->i_lock); 1494 atomic_dec(&inode->i_count); 1495 } 1496 trace_f2fs_drop_inode(inode, 0); 1497 return 0; 1498 } 1499 ret = generic_drop_inode(inode); 1500 if (!ret) 1501 ret = fscrypt_drop_inode(inode); 1502 trace_f2fs_drop_inode(inode, ret); 1503 return ret; 1504 } 1505 f2fs_inode_dirtied(struct inode * inode,bool sync)1506 int f2fs_inode_dirtied(struct inode *inode, bool sync) 1507 { 1508 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1509 int ret = 0; 1510 1511 spin_lock(&sbi->inode_lock[DIRTY_META]); 1512 if (is_inode_flag_set(inode, FI_DIRTY_INODE)) { 1513 ret = 1; 1514 } else { 1515 set_inode_flag(inode, FI_DIRTY_INODE); 1516 stat_inc_dirty_inode(sbi, DIRTY_META); 1517 } 1518 if (sync && list_empty(&F2FS_I(inode)->gdirty_list)) { 1519 list_add_tail(&F2FS_I(inode)->gdirty_list, 1520 &sbi->inode_list[DIRTY_META]); 1521 inc_page_count(sbi, F2FS_DIRTY_IMETA); 1522 } 1523 spin_unlock(&sbi->inode_lock[DIRTY_META]); 1524 return ret; 1525 } 1526 f2fs_inode_synced(struct inode * inode)1527 void f2fs_inode_synced(struct inode *inode) 1528 { 1529 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1530 1531 spin_lock(&sbi->inode_lock[DIRTY_META]); 1532 if (!is_inode_flag_set(inode, FI_DIRTY_INODE)) { 1533 spin_unlock(&sbi->inode_lock[DIRTY_META]); 1534 return; 1535 } 1536 if (!list_empty(&F2FS_I(inode)->gdirty_list)) { 1537 list_del_init(&F2FS_I(inode)->gdirty_list); 1538 dec_page_count(sbi, F2FS_DIRTY_IMETA); 1539 } 1540 clear_inode_flag(inode, FI_DIRTY_INODE); 1541 clear_inode_flag(inode, FI_AUTO_RECOVER); 1542 stat_dec_dirty_inode(F2FS_I_SB(inode), DIRTY_META); 1543 spin_unlock(&sbi->inode_lock[DIRTY_META]); 1544 } 1545 1546 /* 1547 * f2fs_dirty_inode() is called from __mark_inode_dirty() 1548 * 1549 * We should call set_dirty_inode to write the dirty inode through write_inode. 1550 */ f2fs_dirty_inode(struct inode * inode,int flags)1551 static void f2fs_dirty_inode(struct inode *inode, int flags) 1552 { 1553 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1554 1555 if (inode->i_ino == F2FS_NODE_INO(sbi) || 1556 inode->i_ino == F2FS_META_INO(sbi)) 1557 return; 1558 1559 if (is_inode_flag_set(inode, FI_AUTO_RECOVER)) 1560 clear_inode_flag(inode, FI_AUTO_RECOVER); 1561 1562 f2fs_inode_dirtied(inode, false); 1563 } 1564 f2fs_free_inode(struct inode * inode)1565 static void f2fs_free_inode(struct inode *inode) 1566 { 1567 fscrypt_free_inode(inode); 1568 kmem_cache_free(f2fs_inode_cachep, F2FS_I(inode)); 1569 } 1570 destroy_percpu_info(struct f2fs_sb_info * sbi)1571 static void destroy_percpu_info(struct f2fs_sb_info *sbi) 1572 { 1573 percpu_counter_destroy(&sbi->total_valid_inode_count); 1574 percpu_counter_destroy(&sbi->rf_node_block_count); 1575 percpu_counter_destroy(&sbi->alloc_valid_block_count); 1576 } 1577 destroy_device_list(struct f2fs_sb_info * sbi)1578 static void destroy_device_list(struct f2fs_sb_info *sbi) 1579 { 1580 int i; 1581 1582 for (i = 0; i < sbi->s_ndevs; i++) { 1583 blkdev_put(FDEV(i).bdev, FMODE_EXCL); 1584 #ifdef CONFIG_BLK_DEV_ZONED 1585 kvfree(FDEV(i).blkz_seq); 1586 #endif 1587 } 1588 kvfree(sbi->devs); 1589 } 1590 f2fs_put_super(struct super_block * sb)1591 static void f2fs_put_super(struct super_block *sb) 1592 { 1593 struct f2fs_sb_info *sbi = F2FS_SB(sb); 1594 int i; 1595 bool done; 1596 1597 /* unregister procfs/sysfs entries in advance to avoid race case */ 1598 f2fs_unregister_sysfs(sbi); 1599 1600 f2fs_quota_off_umount(sb); 1601 1602 /* prevent remaining shrinker jobs */ 1603 mutex_lock(&sbi->umount_mutex); 1604 1605 /* 1606 * flush all issued checkpoints and stop checkpoint issue thread. 1607 * after then, all checkpoints should be done by each process context. 1608 */ 1609 f2fs_stop_ckpt_thread(sbi); 1610 1611 /* 1612 * We don't need to do checkpoint when superblock is clean. 1613 * But, the previous checkpoint was not done by umount, it needs to do 1614 * clean checkpoint again. 1615 */ 1616 if ((is_sbi_flag_set(sbi, SBI_IS_DIRTY) || 1617 !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG))) { 1618 struct cp_control cpc = { 1619 .reason = CP_UMOUNT, 1620 }; 1621 f2fs_write_checkpoint(sbi, &cpc); 1622 } 1623 1624 /* be sure to wait for any on-going discard commands */ 1625 done = f2fs_issue_discard_timeout(sbi); 1626 if (f2fs_realtime_discard_enable(sbi) && !sbi->discard_blks && done) { 1627 struct cp_control cpc = { 1628 .reason = CP_UMOUNT | CP_TRIMMED, 1629 }; 1630 f2fs_write_checkpoint(sbi, &cpc); 1631 } 1632 1633 /* 1634 * normally superblock is clean, so we need to release this. 1635 * In addition, EIO will skip do checkpoint, we need this as well. 1636 */ 1637 f2fs_release_ino_entry(sbi, true); 1638 1639 f2fs_leave_shrinker(sbi); 1640 mutex_unlock(&sbi->umount_mutex); 1641 1642 /* our cp_error case, we can wait for any writeback page */ 1643 f2fs_flush_merged_writes(sbi); 1644 1645 f2fs_wait_on_all_pages(sbi, F2FS_WB_CP_DATA); 1646 1647 f2fs_bug_on(sbi, sbi->fsync_node_num); 1648 1649 f2fs_destroy_compress_inode(sbi); 1650 1651 iput(sbi->node_inode); 1652 sbi->node_inode = NULL; 1653 1654 iput(sbi->meta_inode); 1655 sbi->meta_inode = NULL; 1656 1657 /* 1658 * iput() can update stat information, if f2fs_write_checkpoint() 1659 * above failed with error. 1660 */ 1661 f2fs_destroy_stats(sbi); 1662 1663 /* destroy f2fs internal modules */ 1664 f2fs_destroy_node_manager(sbi); 1665 f2fs_destroy_segment_manager(sbi); 1666 1667 f2fs_destroy_post_read_wq(sbi); 1668 1669 kvfree(sbi->ckpt); 1670 1671 sb->s_fs_info = NULL; 1672 if (sbi->s_chksum_driver) 1673 crypto_free_shash(sbi->s_chksum_driver); 1674 kfree(sbi->raw_super); 1675 1676 destroy_device_list(sbi); 1677 f2fs_destroy_page_array_cache(sbi); 1678 f2fs_destroy_xattr_caches(sbi); 1679 mempool_destroy(sbi->write_io_dummy); 1680 #ifdef CONFIG_QUOTA 1681 for (i = 0; i < MAXQUOTAS; i++) 1682 kfree(F2FS_OPTION(sbi).s_qf_names[i]); 1683 #endif 1684 fscrypt_free_dummy_policy(&F2FS_OPTION(sbi).dummy_enc_policy); 1685 destroy_percpu_info(sbi); 1686 f2fs_destroy_iostat(sbi); 1687 for (i = 0; i < NR_PAGE_TYPE; i++) 1688 kvfree(sbi->write_io[i]); 1689 #ifdef CONFIG_UNICODE 1690 utf8_unload(sb->s_encoding); 1691 #endif 1692 kfree(sbi); 1693 } 1694 f2fs_sync_fs(struct super_block * sb,int sync)1695 int f2fs_sync_fs(struct super_block *sb, int sync) 1696 { 1697 struct f2fs_sb_info *sbi = F2FS_SB(sb); 1698 int err = 0; 1699 1700 if (unlikely(f2fs_cp_error(sbi))) 1701 return 0; 1702 if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED))) 1703 return 0; 1704 1705 trace_f2fs_sync_fs(sb, sync); 1706 1707 if (unlikely(is_sbi_flag_set(sbi, SBI_POR_DOING))) 1708 return -EAGAIN; 1709 1710 if (sync) 1711 err = f2fs_issue_checkpoint(sbi); 1712 1713 return err; 1714 } 1715 f2fs_freeze(struct super_block * sb)1716 static int f2fs_freeze(struct super_block *sb) 1717 { 1718 if (f2fs_readonly(sb)) 1719 return 0; 1720 1721 /* IO error happened before */ 1722 if (unlikely(f2fs_cp_error(F2FS_SB(sb)))) 1723 return -EIO; 1724 1725 /* must be clean, since sync_filesystem() was already called */ 1726 if (is_sbi_flag_set(F2FS_SB(sb), SBI_IS_DIRTY)) 1727 return -EINVAL; 1728 1729 /* Let's flush checkpoints and stop the thread. */ 1730 f2fs_flush_ckpt_thread(F2FS_SB(sb)); 1731 1732 /* to avoid deadlock on f2fs_evict_inode->SB_FREEZE_FS */ 1733 set_sbi_flag(F2FS_SB(sb), SBI_IS_FREEZING); 1734 return 0; 1735 } 1736 f2fs_unfreeze(struct super_block * sb)1737 static int f2fs_unfreeze(struct super_block *sb) 1738 { 1739 clear_sbi_flag(F2FS_SB(sb), SBI_IS_FREEZING); 1740 return 0; 1741 } 1742 1743 #ifdef CONFIG_QUOTA f2fs_statfs_project(struct super_block * sb,kprojid_t projid,struct kstatfs * buf)1744 static int f2fs_statfs_project(struct super_block *sb, 1745 kprojid_t projid, struct kstatfs *buf) 1746 { 1747 struct kqid qid; 1748 struct dquot *dquot; 1749 u64 limit; 1750 u64 curblock; 1751 1752 qid = make_kqid_projid(projid); 1753 dquot = dqget(sb, qid); 1754 if (IS_ERR(dquot)) 1755 return PTR_ERR(dquot); 1756 spin_lock(&dquot->dq_dqb_lock); 1757 1758 limit = min_not_zero(dquot->dq_dqb.dqb_bsoftlimit, 1759 dquot->dq_dqb.dqb_bhardlimit); 1760 if (limit) 1761 limit >>= sb->s_blocksize_bits; 1762 1763 if (limit && buf->f_blocks > limit) { 1764 curblock = (dquot->dq_dqb.dqb_curspace + 1765 dquot->dq_dqb.dqb_rsvspace) >> sb->s_blocksize_bits; 1766 buf->f_blocks = limit; 1767 buf->f_bfree = buf->f_bavail = 1768 (buf->f_blocks > curblock) ? 1769 (buf->f_blocks - curblock) : 0; 1770 } 1771 1772 limit = min_not_zero(dquot->dq_dqb.dqb_isoftlimit, 1773 dquot->dq_dqb.dqb_ihardlimit); 1774 1775 if (limit && buf->f_files > limit) { 1776 buf->f_files = limit; 1777 buf->f_ffree = 1778 (buf->f_files > dquot->dq_dqb.dqb_curinodes) ? 1779 (buf->f_files - dquot->dq_dqb.dqb_curinodes) : 0; 1780 } 1781 1782 spin_unlock(&dquot->dq_dqb_lock); 1783 dqput(dquot); 1784 return 0; 1785 } 1786 #endif 1787 f2fs_statfs(struct dentry * dentry,struct kstatfs * buf)1788 static int f2fs_statfs(struct dentry *dentry, struct kstatfs *buf) 1789 { 1790 struct super_block *sb = dentry->d_sb; 1791 struct f2fs_sb_info *sbi = F2FS_SB(sb); 1792 u64 id = huge_encode_dev(sb->s_bdev->bd_dev); 1793 block_t total_count, user_block_count, start_count; 1794 u64 avail_node_count; 1795 unsigned int total_valid_node_count; 1796 1797 total_count = le64_to_cpu(sbi->raw_super->block_count); 1798 start_count = le32_to_cpu(sbi->raw_super->segment0_blkaddr); 1799 buf->f_type = F2FS_SUPER_MAGIC; 1800 buf->f_bsize = sbi->blocksize; 1801 1802 buf->f_blocks = total_count - start_count; 1803 1804 spin_lock(&sbi->stat_lock); 1805 1806 user_block_count = sbi->user_block_count; 1807 total_valid_node_count = valid_node_count(sbi); 1808 avail_node_count = sbi->total_node_count - F2FS_RESERVED_NODE_NUM; 1809 buf->f_bfree = user_block_count - valid_user_blocks(sbi) - 1810 sbi->current_reserved_blocks; 1811 1812 if (unlikely(buf->f_bfree <= sbi->unusable_block_count)) 1813 buf->f_bfree = 0; 1814 else 1815 buf->f_bfree -= sbi->unusable_block_count; 1816 spin_unlock(&sbi->stat_lock); 1817 1818 if (buf->f_bfree > F2FS_OPTION(sbi).root_reserved_blocks) 1819 buf->f_bavail = buf->f_bfree - 1820 F2FS_OPTION(sbi).root_reserved_blocks; 1821 else 1822 buf->f_bavail = 0; 1823 1824 if (avail_node_count > user_block_count) { 1825 buf->f_files = user_block_count; 1826 buf->f_ffree = buf->f_bavail; 1827 } else { 1828 buf->f_files = avail_node_count; 1829 buf->f_ffree = min(avail_node_count - total_valid_node_count, 1830 buf->f_bavail); 1831 } 1832 1833 buf->f_namelen = F2FS_NAME_LEN; 1834 buf->f_fsid = u64_to_fsid(id); 1835 1836 #ifdef CONFIG_QUOTA 1837 if (is_inode_flag_set(dentry->d_inode, FI_PROJ_INHERIT) && 1838 sb_has_quota_limits_enabled(sb, PRJQUOTA)) { 1839 f2fs_statfs_project(sb, F2FS_I(dentry->d_inode)->i_projid, buf); 1840 } 1841 #endif 1842 return 0; 1843 } 1844 f2fs_show_quota_options(struct seq_file * seq,struct super_block * sb)1845 static inline void f2fs_show_quota_options(struct seq_file *seq, 1846 struct super_block *sb) 1847 { 1848 #ifdef CONFIG_QUOTA 1849 struct f2fs_sb_info *sbi = F2FS_SB(sb); 1850 1851 if (F2FS_OPTION(sbi).s_jquota_fmt) { 1852 char *fmtname = ""; 1853 1854 switch (F2FS_OPTION(sbi).s_jquota_fmt) { 1855 case QFMT_VFS_OLD: 1856 fmtname = "vfsold"; 1857 break; 1858 case QFMT_VFS_V0: 1859 fmtname = "vfsv0"; 1860 break; 1861 case QFMT_VFS_V1: 1862 fmtname = "vfsv1"; 1863 break; 1864 } 1865 seq_printf(seq, ",jqfmt=%s", fmtname); 1866 } 1867 1868 if (F2FS_OPTION(sbi).s_qf_names[USRQUOTA]) 1869 seq_show_option(seq, "usrjquota", 1870 F2FS_OPTION(sbi).s_qf_names[USRQUOTA]); 1871 1872 if (F2FS_OPTION(sbi).s_qf_names[GRPQUOTA]) 1873 seq_show_option(seq, "grpjquota", 1874 F2FS_OPTION(sbi).s_qf_names[GRPQUOTA]); 1875 1876 if (F2FS_OPTION(sbi).s_qf_names[PRJQUOTA]) 1877 seq_show_option(seq, "prjjquota", 1878 F2FS_OPTION(sbi).s_qf_names[PRJQUOTA]); 1879 #endif 1880 } 1881 1882 #ifdef CONFIG_F2FS_FS_COMPRESSION f2fs_show_compress_options(struct seq_file * seq,struct super_block * sb)1883 static inline void f2fs_show_compress_options(struct seq_file *seq, 1884 struct super_block *sb) 1885 { 1886 struct f2fs_sb_info *sbi = F2FS_SB(sb); 1887 char *algtype = ""; 1888 int i; 1889 1890 if (!f2fs_sb_has_compression(sbi)) 1891 return; 1892 1893 switch (F2FS_OPTION(sbi).compress_algorithm) { 1894 case COMPRESS_LZO: 1895 algtype = "lzo"; 1896 break; 1897 case COMPRESS_LZ4: 1898 algtype = "lz4"; 1899 break; 1900 case COMPRESS_ZSTD: 1901 algtype = "zstd"; 1902 break; 1903 case COMPRESS_LZORLE: 1904 algtype = "lzo-rle"; 1905 break; 1906 } 1907 seq_printf(seq, ",compress_algorithm=%s", algtype); 1908 1909 if (F2FS_OPTION(sbi).compress_level) 1910 seq_printf(seq, ":%d", F2FS_OPTION(sbi).compress_level); 1911 1912 seq_printf(seq, ",compress_log_size=%u", 1913 F2FS_OPTION(sbi).compress_log_size); 1914 1915 for (i = 0; i < F2FS_OPTION(sbi).compress_ext_cnt; i++) { 1916 seq_printf(seq, ",compress_extension=%s", 1917 F2FS_OPTION(sbi).extensions[i]); 1918 } 1919 1920 for (i = 0; i < F2FS_OPTION(sbi).nocompress_ext_cnt; i++) { 1921 seq_printf(seq, ",nocompress_extension=%s", 1922 F2FS_OPTION(sbi).noextensions[i]); 1923 } 1924 1925 if (F2FS_OPTION(sbi).compress_chksum) 1926 seq_puts(seq, ",compress_chksum"); 1927 1928 if (F2FS_OPTION(sbi).compress_mode == COMPR_MODE_FS) 1929 seq_printf(seq, ",compress_mode=%s", "fs"); 1930 else if (F2FS_OPTION(sbi).compress_mode == COMPR_MODE_USER) 1931 seq_printf(seq, ",compress_mode=%s", "user"); 1932 1933 if (test_opt(sbi, COMPRESS_CACHE)) 1934 seq_puts(seq, ",compress_cache"); 1935 } 1936 #endif 1937 f2fs_show_options(struct seq_file * seq,struct dentry * root)1938 static int f2fs_show_options(struct seq_file *seq, struct dentry *root) 1939 { 1940 struct f2fs_sb_info *sbi = F2FS_SB(root->d_sb); 1941 1942 if (F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_SYNC) 1943 seq_printf(seq, ",background_gc=%s", "sync"); 1944 else if (F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_ON) 1945 seq_printf(seq, ",background_gc=%s", "on"); 1946 else if (F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_OFF) 1947 seq_printf(seq, ",background_gc=%s", "off"); 1948 1949 if (test_opt(sbi, GC_MERGE)) 1950 seq_puts(seq, ",gc_merge"); 1951 else 1952 seq_puts(seq, ",nogc_merge"); 1953 1954 if (test_opt(sbi, DISABLE_ROLL_FORWARD)) 1955 seq_puts(seq, ",disable_roll_forward"); 1956 if (test_opt(sbi, NORECOVERY)) 1957 seq_puts(seq, ",norecovery"); 1958 if (test_opt(sbi, DISCARD)) { 1959 seq_puts(seq, ",discard"); 1960 if (F2FS_OPTION(sbi).discard_unit == DISCARD_UNIT_BLOCK) 1961 seq_printf(seq, ",discard_unit=%s", "block"); 1962 else if (F2FS_OPTION(sbi).discard_unit == DISCARD_UNIT_SEGMENT) 1963 seq_printf(seq, ",discard_unit=%s", "segment"); 1964 else if (F2FS_OPTION(sbi).discard_unit == DISCARD_UNIT_SECTION) 1965 seq_printf(seq, ",discard_unit=%s", "section"); 1966 } else { 1967 seq_puts(seq, ",nodiscard"); 1968 } 1969 if (test_opt(sbi, NOHEAP)) 1970 seq_puts(seq, ",no_heap"); 1971 else 1972 seq_puts(seq, ",heap"); 1973 #ifdef CONFIG_F2FS_FS_XATTR 1974 if (test_opt(sbi, XATTR_USER)) 1975 seq_puts(seq, ",user_xattr"); 1976 else 1977 seq_puts(seq, ",nouser_xattr"); 1978 if (test_opt(sbi, INLINE_XATTR)) 1979 seq_puts(seq, ",inline_xattr"); 1980 else 1981 seq_puts(seq, ",noinline_xattr"); 1982 if (test_opt(sbi, INLINE_XATTR_SIZE)) 1983 seq_printf(seq, ",inline_xattr_size=%u", 1984 F2FS_OPTION(sbi).inline_xattr_size); 1985 #endif 1986 #ifdef CONFIG_F2FS_FS_POSIX_ACL 1987 if (test_opt(sbi, POSIX_ACL)) 1988 seq_puts(seq, ",acl"); 1989 else 1990 seq_puts(seq, ",noacl"); 1991 #endif 1992 if (test_opt(sbi, DISABLE_EXT_IDENTIFY)) 1993 seq_puts(seq, ",disable_ext_identify"); 1994 if (test_opt(sbi, INLINE_DATA)) 1995 seq_puts(seq, ",inline_data"); 1996 else 1997 seq_puts(seq, ",noinline_data"); 1998 if (test_opt(sbi, INLINE_DENTRY)) 1999 seq_puts(seq, ",inline_dentry"); 2000 else 2001 seq_puts(seq, ",noinline_dentry"); 2002 if (test_opt(sbi, FLUSH_MERGE)) 2003 seq_puts(seq, ",flush_merge"); 2004 else 2005 seq_puts(seq, ",noflush_merge"); 2006 if (test_opt(sbi, NOBARRIER)) 2007 seq_puts(seq, ",nobarrier"); 2008 else 2009 seq_puts(seq, ",barrier"); 2010 if (test_opt(sbi, FASTBOOT)) 2011 seq_puts(seq, ",fastboot"); 2012 if (test_opt(sbi, READ_EXTENT_CACHE)) 2013 seq_puts(seq, ",extent_cache"); 2014 else 2015 seq_puts(seq, ",noextent_cache"); 2016 if (test_opt(sbi, AGE_EXTENT_CACHE)) 2017 seq_puts(seq, ",age_extent_cache"); 2018 if (test_opt(sbi, DATA_FLUSH)) 2019 seq_puts(seq, ",data_flush"); 2020 2021 seq_puts(seq, ",mode="); 2022 if (F2FS_OPTION(sbi).fs_mode == FS_MODE_ADAPTIVE) 2023 seq_puts(seq, "adaptive"); 2024 else if (F2FS_OPTION(sbi).fs_mode == FS_MODE_LFS) 2025 seq_puts(seq, "lfs"); 2026 else if (F2FS_OPTION(sbi).fs_mode == FS_MODE_FRAGMENT_SEG) 2027 seq_puts(seq, "fragment:segment"); 2028 else if (F2FS_OPTION(sbi).fs_mode == FS_MODE_FRAGMENT_BLK) 2029 seq_puts(seq, "fragment:block"); 2030 seq_printf(seq, ",active_logs=%u", F2FS_OPTION(sbi).active_logs); 2031 if (test_opt(sbi, RESERVE_ROOT)) 2032 seq_printf(seq, ",reserve_root=%u,resuid=%u,resgid=%u", 2033 F2FS_OPTION(sbi).root_reserved_blocks, 2034 from_kuid_munged(&init_user_ns, 2035 F2FS_OPTION(sbi).s_resuid), 2036 from_kgid_munged(&init_user_ns, 2037 F2FS_OPTION(sbi).s_resgid)); 2038 if (F2FS_IO_SIZE_BITS(sbi)) 2039 seq_printf(seq, ",io_bits=%u", 2040 F2FS_OPTION(sbi).write_io_size_bits); 2041 #ifdef CONFIG_F2FS_FAULT_INJECTION 2042 if (test_opt(sbi, FAULT_INJECTION)) { 2043 seq_printf(seq, ",fault_injection=%u", 2044 F2FS_OPTION(sbi).fault_info.inject_rate); 2045 seq_printf(seq, ",fault_type=%u", 2046 F2FS_OPTION(sbi).fault_info.inject_type); 2047 } 2048 #endif 2049 #ifdef CONFIG_QUOTA 2050 if (test_opt(sbi, QUOTA)) 2051 seq_puts(seq, ",quota"); 2052 if (test_opt(sbi, USRQUOTA)) 2053 seq_puts(seq, ",usrquota"); 2054 if (test_opt(sbi, GRPQUOTA)) 2055 seq_puts(seq, ",grpquota"); 2056 if (test_opt(sbi, PRJQUOTA)) 2057 seq_puts(seq, ",prjquota"); 2058 #endif 2059 f2fs_show_quota_options(seq, sbi->sb); 2060 2061 fscrypt_show_test_dummy_encryption(seq, ',', sbi->sb); 2062 2063 if (sbi->sb->s_flags & SB_INLINECRYPT) 2064 seq_puts(seq, ",inlinecrypt"); 2065 2066 if (F2FS_OPTION(sbi).alloc_mode == ALLOC_MODE_DEFAULT) 2067 seq_printf(seq, ",alloc_mode=%s", "default"); 2068 else if (F2FS_OPTION(sbi).alloc_mode == ALLOC_MODE_REUSE) 2069 seq_printf(seq, ",alloc_mode=%s", "reuse"); 2070 2071 if (test_opt(sbi, DISABLE_CHECKPOINT)) 2072 seq_printf(seq, ",checkpoint=disable:%u", 2073 F2FS_OPTION(sbi).unusable_cap); 2074 if (test_opt(sbi, MERGE_CHECKPOINT)) 2075 seq_puts(seq, ",checkpoint_merge"); 2076 else 2077 seq_puts(seq, ",nocheckpoint_merge"); 2078 if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_POSIX) 2079 seq_printf(seq, ",fsync_mode=%s", "posix"); 2080 else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_STRICT) 2081 seq_printf(seq, ",fsync_mode=%s", "strict"); 2082 else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_NOBARRIER) 2083 seq_printf(seq, ",fsync_mode=%s", "nobarrier"); 2084 2085 #ifdef CONFIG_F2FS_FS_COMPRESSION 2086 f2fs_show_compress_options(seq, sbi->sb); 2087 #endif 2088 2089 if (test_opt(sbi, ATGC)) 2090 seq_puts(seq, ",atgc"); 2091 2092 if (F2FS_OPTION(sbi).memory_mode == MEMORY_MODE_NORMAL) 2093 seq_printf(seq, ",memory=%s", "normal"); 2094 else if (F2FS_OPTION(sbi).memory_mode == MEMORY_MODE_LOW) 2095 seq_printf(seq, ",memory=%s", "low"); 2096 2097 return 0; 2098 } 2099 default_options(struct f2fs_sb_info * sbi)2100 static void default_options(struct f2fs_sb_info *sbi) 2101 { 2102 /* init some FS parameters */ 2103 if (f2fs_sb_has_readonly(sbi)) 2104 F2FS_OPTION(sbi).active_logs = NR_CURSEG_RO_TYPE; 2105 else 2106 F2FS_OPTION(sbi).active_logs = NR_CURSEG_PERSIST_TYPE; 2107 2108 F2FS_OPTION(sbi).inline_xattr_size = DEFAULT_INLINE_XATTR_ADDRS; 2109 if (le32_to_cpu(F2FS_RAW_SUPER(sbi)->segment_count_main) <= 2110 SMALL_VOLUME_SEGMENTS) 2111 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_REUSE; 2112 else 2113 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_DEFAULT; 2114 F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_POSIX; 2115 F2FS_OPTION(sbi).s_resuid = make_kuid(&init_user_ns, F2FS_DEF_RESUID); 2116 F2FS_OPTION(sbi).s_resgid = make_kgid(&init_user_ns, F2FS_DEF_RESGID); 2117 if (f2fs_sb_has_compression(sbi)) { 2118 F2FS_OPTION(sbi).compress_algorithm = COMPRESS_LZ4; 2119 F2FS_OPTION(sbi).compress_log_size = MIN_COMPRESS_LOG_SIZE; 2120 F2FS_OPTION(sbi).compress_ext_cnt = 0; 2121 F2FS_OPTION(sbi).compress_mode = COMPR_MODE_FS; 2122 } 2123 F2FS_OPTION(sbi).bggc_mode = BGGC_MODE_ON; 2124 F2FS_OPTION(sbi).memory_mode = MEMORY_MODE_NORMAL; 2125 2126 sbi->sb->s_flags &= ~SB_INLINECRYPT; 2127 2128 set_opt(sbi, INLINE_XATTR); 2129 set_opt(sbi, INLINE_DATA); 2130 set_opt(sbi, INLINE_DENTRY); 2131 set_opt(sbi, READ_EXTENT_CACHE); 2132 set_opt(sbi, NOHEAP); 2133 clear_opt(sbi, DISABLE_CHECKPOINT); 2134 set_opt(sbi, MERGE_CHECKPOINT); 2135 F2FS_OPTION(sbi).unusable_cap = 0; 2136 sbi->sb->s_flags |= SB_LAZYTIME; 2137 if (!f2fs_is_readonly(sbi)) 2138 set_opt(sbi, FLUSH_MERGE); 2139 if (f2fs_hw_support_discard(sbi) || f2fs_hw_should_discard(sbi)) 2140 set_opt(sbi, DISCARD); 2141 if (f2fs_sb_has_blkzoned(sbi)) { 2142 F2FS_OPTION(sbi).fs_mode = FS_MODE_LFS; 2143 F2FS_OPTION(sbi).discard_unit = DISCARD_UNIT_SECTION; 2144 } else { 2145 F2FS_OPTION(sbi).fs_mode = FS_MODE_ADAPTIVE; 2146 F2FS_OPTION(sbi).discard_unit = DISCARD_UNIT_BLOCK; 2147 } 2148 2149 #ifdef CONFIG_F2FS_FS_XATTR 2150 set_opt(sbi, XATTR_USER); 2151 #endif 2152 #ifdef CONFIG_F2FS_FS_POSIX_ACL 2153 set_opt(sbi, POSIX_ACL); 2154 #endif 2155 2156 f2fs_build_fault_attr(sbi, 0, 0); 2157 } 2158 2159 #ifdef CONFIG_QUOTA 2160 static int f2fs_enable_quotas(struct super_block *sb); 2161 #endif 2162 f2fs_disable_checkpoint(struct f2fs_sb_info * sbi)2163 static int f2fs_disable_checkpoint(struct f2fs_sb_info *sbi) 2164 { 2165 unsigned int s_flags = sbi->sb->s_flags; 2166 struct cp_control cpc; 2167 unsigned int gc_mode = sbi->gc_mode; 2168 int err = 0; 2169 int ret; 2170 block_t unusable; 2171 2172 if (s_flags & SB_RDONLY) { 2173 f2fs_err(sbi, "checkpoint=disable on readonly fs"); 2174 return -EINVAL; 2175 } 2176 sbi->sb->s_flags |= SB_ACTIVE; 2177 2178 /* check if we need more GC first */ 2179 unusable = f2fs_get_unusable_blocks(sbi); 2180 if (!f2fs_disable_cp_again(sbi, unusable)) 2181 goto skip_gc; 2182 2183 f2fs_update_time(sbi, DISABLE_TIME); 2184 2185 sbi->gc_mode = GC_URGENT_HIGH; 2186 2187 while (!f2fs_time_over(sbi, DISABLE_TIME)) { 2188 struct f2fs_gc_control gc_control = { 2189 .victim_segno = NULL_SEGNO, 2190 .init_gc_type = FG_GC, 2191 .should_migrate_blocks = false, 2192 .err_gc_skipped = true, 2193 .nr_free_secs = 1 }; 2194 2195 f2fs_down_write(&sbi->gc_lock); 2196 err = f2fs_gc(sbi, &gc_control); 2197 if (err == -ENODATA) { 2198 err = 0; 2199 break; 2200 } 2201 if (err && err != -EAGAIN) 2202 break; 2203 } 2204 2205 ret = sync_filesystem(sbi->sb); 2206 if (ret || err) { 2207 err = ret ? ret : err; 2208 goto restore_flag; 2209 } 2210 2211 unusable = f2fs_get_unusable_blocks(sbi); 2212 if (f2fs_disable_cp_again(sbi, unusable)) { 2213 err = -EAGAIN; 2214 goto restore_flag; 2215 } 2216 2217 skip_gc: 2218 f2fs_down_write(&sbi->gc_lock); 2219 cpc.reason = CP_PAUSE; 2220 set_sbi_flag(sbi, SBI_CP_DISABLED); 2221 err = f2fs_write_checkpoint(sbi, &cpc); 2222 if (err) 2223 goto out_unlock; 2224 2225 spin_lock(&sbi->stat_lock); 2226 sbi->unusable_block_count = unusable; 2227 spin_unlock(&sbi->stat_lock); 2228 2229 out_unlock: 2230 f2fs_up_write(&sbi->gc_lock); 2231 restore_flag: 2232 sbi->gc_mode = gc_mode; 2233 sbi->sb->s_flags = s_flags; /* Restore SB_RDONLY status */ 2234 return err; 2235 } 2236 f2fs_enable_checkpoint(struct f2fs_sb_info * sbi)2237 static void f2fs_enable_checkpoint(struct f2fs_sb_info *sbi) 2238 { 2239 int retry = DEFAULT_RETRY_IO_COUNT; 2240 2241 /* we should flush all the data to keep data consistency */ 2242 do { 2243 sync_inodes_sb(sbi->sb); 2244 f2fs_io_schedule_timeout(DEFAULT_IO_TIMEOUT); 2245 } while (get_pages(sbi, F2FS_DIRTY_DATA) && retry--); 2246 2247 if (unlikely(retry < 0)) 2248 f2fs_warn(sbi, "checkpoint=enable has some unwritten data."); 2249 2250 f2fs_down_write(&sbi->gc_lock); 2251 f2fs_dirty_to_prefree(sbi); 2252 2253 clear_sbi_flag(sbi, SBI_CP_DISABLED); 2254 set_sbi_flag(sbi, SBI_IS_DIRTY); 2255 f2fs_up_write(&sbi->gc_lock); 2256 2257 f2fs_sync_fs(sbi->sb, 1); 2258 2259 /* Let's ensure there's no pending checkpoint anymore */ 2260 f2fs_flush_ckpt_thread(sbi); 2261 } 2262 f2fs_remount(struct super_block * sb,int * flags,char * data)2263 static int f2fs_remount(struct super_block *sb, int *flags, char *data) 2264 { 2265 struct f2fs_sb_info *sbi = F2FS_SB(sb); 2266 struct f2fs_mount_info org_mount_opt; 2267 unsigned long old_sb_flags; 2268 int err; 2269 bool need_restart_gc = false, need_stop_gc = false; 2270 bool need_restart_ckpt = false, need_stop_ckpt = false; 2271 bool need_restart_flush = false, need_stop_flush = false; 2272 bool need_restart_discard = false, need_stop_discard = false; 2273 bool no_read_extent_cache = !test_opt(sbi, READ_EXTENT_CACHE); 2274 bool no_age_extent_cache = !test_opt(sbi, AGE_EXTENT_CACHE); 2275 bool enable_checkpoint = !test_opt(sbi, DISABLE_CHECKPOINT); 2276 bool no_io_align = !F2FS_IO_ALIGNED(sbi); 2277 bool no_atgc = !test_opt(sbi, ATGC); 2278 bool no_discard = !test_opt(sbi, DISCARD); 2279 bool no_compress_cache = !test_opt(sbi, COMPRESS_CACHE); 2280 bool block_unit_discard = f2fs_block_unit_discard(sbi); 2281 #ifdef CONFIG_QUOTA 2282 int i, j; 2283 #endif 2284 2285 /* 2286 * Save the old mount options in case we 2287 * need to restore them. 2288 */ 2289 org_mount_opt = sbi->mount_opt; 2290 old_sb_flags = sb->s_flags; 2291 2292 #ifdef CONFIG_QUOTA 2293 org_mount_opt.s_jquota_fmt = F2FS_OPTION(sbi).s_jquota_fmt; 2294 for (i = 0; i < MAXQUOTAS; i++) { 2295 if (F2FS_OPTION(sbi).s_qf_names[i]) { 2296 org_mount_opt.s_qf_names[i] = 2297 kstrdup(F2FS_OPTION(sbi).s_qf_names[i], 2298 GFP_KERNEL); 2299 if (!org_mount_opt.s_qf_names[i]) { 2300 for (j = 0; j < i; j++) 2301 kfree(org_mount_opt.s_qf_names[j]); 2302 return -ENOMEM; 2303 } 2304 } else { 2305 org_mount_opt.s_qf_names[i] = NULL; 2306 } 2307 } 2308 #endif 2309 2310 /* recover superblocks we couldn't write due to previous RO mount */ 2311 if (!(*flags & SB_RDONLY) && is_sbi_flag_set(sbi, SBI_NEED_SB_WRITE)) { 2312 err = f2fs_commit_super(sbi, false); 2313 f2fs_info(sbi, "Try to recover all the superblocks, ret: %d", 2314 err); 2315 if (!err) 2316 clear_sbi_flag(sbi, SBI_NEED_SB_WRITE); 2317 } 2318 2319 default_options(sbi); 2320 2321 /* parse mount options */ 2322 err = parse_options(sb, data, true); 2323 if (err) 2324 goto restore_opts; 2325 2326 /* 2327 * Previous and new state of filesystem is RO, 2328 * so skip checking GC and FLUSH_MERGE conditions. 2329 */ 2330 if (f2fs_readonly(sb) && (*flags & SB_RDONLY)) 2331 goto skip; 2332 2333 if (f2fs_dev_is_readonly(sbi) && !(*flags & SB_RDONLY)) { 2334 err = -EROFS; 2335 goto restore_opts; 2336 } 2337 2338 #ifdef CONFIG_QUOTA 2339 if (!f2fs_readonly(sb) && (*flags & SB_RDONLY)) { 2340 err = dquot_suspend(sb, -1); 2341 if (err < 0) 2342 goto restore_opts; 2343 } else if (f2fs_readonly(sb) && !(*flags & SB_RDONLY)) { 2344 /* dquot_resume needs RW */ 2345 sb->s_flags &= ~SB_RDONLY; 2346 if (sb_any_quota_suspended(sb)) { 2347 dquot_resume(sb, -1); 2348 } else if (f2fs_sb_has_quota_ino(sbi)) { 2349 err = f2fs_enable_quotas(sb); 2350 if (err) 2351 goto restore_opts; 2352 } 2353 } 2354 #endif 2355 if (f2fs_lfs_mode(sbi) && !IS_F2FS_IPU_DISABLE(sbi)) { 2356 err = -EINVAL; 2357 f2fs_warn(sbi, "LFS is not compatible with IPU"); 2358 goto restore_opts; 2359 } 2360 2361 /* disallow enable atgc dynamically */ 2362 if (no_atgc == !!test_opt(sbi, ATGC)) { 2363 err = -EINVAL; 2364 f2fs_warn(sbi, "switch atgc option is not allowed"); 2365 goto restore_opts; 2366 } 2367 2368 /* disallow enable/disable extent_cache dynamically */ 2369 if (no_read_extent_cache == !!test_opt(sbi, READ_EXTENT_CACHE)) { 2370 err = -EINVAL; 2371 f2fs_warn(sbi, "switch extent_cache option is not allowed"); 2372 goto restore_opts; 2373 } 2374 /* disallow enable/disable age extent_cache dynamically */ 2375 if (no_age_extent_cache == !!test_opt(sbi, AGE_EXTENT_CACHE)) { 2376 err = -EINVAL; 2377 f2fs_warn(sbi, "switch age_extent_cache option is not allowed"); 2378 goto restore_opts; 2379 } 2380 2381 if (no_io_align == !!F2FS_IO_ALIGNED(sbi)) { 2382 err = -EINVAL; 2383 f2fs_warn(sbi, "switch io_bits option is not allowed"); 2384 goto restore_opts; 2385 } 2386 2387 if (no_compress_cache == !!test_opt(sbi, COMPRESS_CACHE)) { 2388 err = -EINVAL; 2389 f2fs_warn(sbi, "switch compress_cache option is not allowed"); 2390 goto restore_opts; 2391 } 2392 2393 if (block_unit_discard != f2fs_block_unit_discard(sbi)) { 2394 err = -EINVAL; 2395 f2fs_warn(sbi, "switch discard_unit option is not allowed"); 2396 goto restore_opts; 2397 } 2398 2399 if ((*flags & SB_RDONLY) && test_opt(sbi, DISABLE_CHECKPOINT)) { 2400 err = -EINVAL; 2401 f2fs_warn(sbi, "disabling checkpoint not compatible with read-only"); 2402 goto restore_opts; 2403 } 2404 2405 /* 2406 * We stop the GC thread if FS is mounted as RO 2407 * or if background_gc = off is passed in mount 2408 * option. Also sync the filesystem. 2409 */ 2410 if ((*flags & SB_RDONLY) || 2411 (F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_OFF && 2412 !test_opt(sbi, GC_MERGE))) { 2413 if (sbi->gc_thread) { 2414 f2fs_stop_gc_thread(sbi); 2415 need_restart_gc = true; 2416 } 2417 } else if (!sbi->gc_thread) { 2418 err = f2fs_start_gc_thread(sbi); 2419 if (err) 2420 goto restore_opts; 2421 need_stop_gc = true; 2422 } 2423 2424 if (*flags & SB_RDONLY) { 2425 sync_inodes_sb(sb); 2426 2427 set_sbi_flag(sbi, SBI_IS_DIRTY); 2428 set_sbi_flag(sbi, SBI_IS_CLOSE); 2429 f2fs_sync_fs(sb, 1); 2430 clear_sbi_flag(sbi, SBI_IS_CLOSE); 2431 } 2432 2433 if ((*flags & SB_RDONLY) || test_opt(sbi, DISABLE_CHECKPOINT) || 2434 !test_opt(sbi, MERGE_CHECKPOINT)) { 2435 f2fs_stop_ckpt_thread(sbi); 2436 need_restart_ckpt = true; 2437 } else { 2438 /* Flush if the prevous checkpoint, if exists. */ 2439 f2fs_flush_ckpt_thread(sbi); 2440 2441 err = f2fs_start_ckpt_thread(sbi); 2442 if (err) { 2443 f2fs_err(sbi, 2444 "Failed to start F2FS issue_checkpoint_thread (%d)", 2445 err); 2446 goto restore_gc; 2447 } 2448 need_stop_ckpt = true; 2449 } 2450 2451 /* 2452 * We stop issue flush thread if FS is mounted as RO 2453 * or if flush_merge is not passed in mount option. 2454 */ 2455 if ((*flags & SB_RDONLY) || !test_opt(sbi, FLUSH_MERGE)) { 2456 clear_opt(sbi, FLUSH_MERGE); 2457 f2fs_destroy_flush_cmd_control(sbi, false); 2458 need_restart_flush = true; 2459 } else { 2460 err = f2fs_create_flush_cmd_control(sbi); 2461 if (err) 2462 goto restore_ckpt; 2463 need_stop_flush = true; 2464 } 2465 2466 if (no_discard == !!test_opt(sbi, DISCARD)) { 2467 if (test_opt(sbi, DISCARD)) { 2468 err = f2fs_start_discard_thread(sbi); 2469 if (err) 2470 goto restore_flush; 2471 need_stop_discard = true; 2472 } else { 2473 f2fs_stop_discard_thread(sbi); 2474 f2fs_issue_discard_timeout(sbi); 2475 need_restart_discard = true; 2476 } 2477 } 2478 2479 if (enable_checkpoint == !!test_opt(sbi, DISABLE_CHECKPOINT)) { 2480 if (test_opt(sbi, DISABLE_CHECKPOINT)) { 2481 err = f2fs_disable_checkpoint(sbi); 2482 if (err) 2483 goto restore_discard; 2484 } else { 2485 f2fs_enable_checkpoint(sbi); 2486 } 2487 } 2488 2489 skip: 2490 #ifdef CONFIG_QUOTA 2491 /* Release old quota file names */ 2492 for (i = 0; i < MAXQUOTAS; i++) 2493 kfree(org_mount_opt.s_qf_names[i]); 2494 #endif 2495 /* Update the POSIXACL Flag */ 2496 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) | 2497 (test_opt(sbi, POSIX_ACL) ? SB_POSIXACL : 0); 2498 2499 limit_reserve_root(sbi); 2500 adjust_unusable_cap_perc(sbi); 2501 *flags = (*flags & ~SB_LAZYTIME) | (sb->s_flags & SB_LAZYTIME); 2502 return 0; 2503 restore_discard: 2504 if (need_restart_discard) { 2505 if (f2fs_start_discard_thread(sbi)) 2506 f2fs_warn(sbi, "discard has been stopped"); 2507 } else if (need_stop_discard) { 2508 f2fs_stop_discard_thread(sbi); 2509 } 2510 restore_flush: 2511 if (need_restart_flush) { 2512 if (f2fs_create_flush_cmd_control(sbi)) 2513 f2fs_warn(sbi, "background flush thread has stopped"); 2514 } else if (need_stop_flush) { 2515 clear_opt(sbi, FLUSH_MERGE); 2516 f2fs_destroy_flush_cmd_control(sbi, false); 2517 } 2518 restore_ckpt: 2519 if (need_restart_ckpt) { 2520 if (f2fs_start_ckpt_thread(sbi)) 2521 f2fs_warn(sbi, "background ckpt thread has stopped"); 2522 } else if (need_stop_ckpt) { 2523 f2fs_stop_ckpt_thread(sbi); 2524 } 2525 restore_gc: 2526 if (need_restart_gc) { 2527 if (f2fs_start_gc_thread(sbi)) 2528 f2fs_warn(sbi, "background gc thread has stopped"); 2529 } else if (need_stop_gc) { 2530 f2fs_stop_gc_thread(sbi); 2531 } 2532 restore_opts: 2533 #ifdef CONFIG_QUOTA 2534 F2FS_OPTION(sbi).s_jquota_fmt = org_mount_opt.s_jquota_fmt; 2535 for (i = 0; i < MAXQUOTAS; i++) { 2536 kfree(F2FS_OPTION(sbi).s_qf_names[i]); 2537 F2FS_OPTION(sbi).s_qf_names[i] = org_mount_opt.s_qf_names[i]; 2538 } 2539 #endif 2540 sbi->mount_opt = org_mount_opt; 2541 sb->s_flags = old_sb_flags; 2542 return err; 2543 } 2544 2545 #ifdef CONFIG_QUOTA f2fs_need_recovery(struct f2fs_sb_info * sbi)2546 static bool f2fs_need_recovery(struct f2fs_sb_info *sbi) 2547 { 2548 /* need to recovery orphan */ 2549 if (is_set_ckpt_flags(sbi, CP_ORPHAN_PRESENT_FLAG)) 2550 return true; 2551 /* need to recovery data */ 2552 if (test_opt(sbi, DISABLE_ROLL_FORWARD)) 2553 return false; 2554 if (test_opt(sbi, NORECOVERY)) 2555 return false; 2556 return !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG); 2557 } 2558 f2fs_recover_quota_begin(struct f2fs_sb_info * sbi)2559 static bool f2fs_recover_quota_begin(struct f2fs_sb_info *sbi) 2560 { 2561 bool readonly = f2fs_readonly(sbi->sb); 2562 2563 if (!f2fs_need_recovery(sbi)) 2564 return false; 2565 2566 /* it doesn't need to check f2fs_sb_has_readonly() */ 2567 if (f2fs_hw_is_readonly(sbi)) 2568 return false; 2569 2570 if (readonly) { 2571 sbi->sb->s_flags &= ~SB_RDONLY; 2572 set_sbi_flag(sbi, SBI_IS_WRITABLE); 2573 } 2574 2575 /* 2576 * Turn on quotas which were not enabled for read-only mounts if 2577 * filesystem has quota feature, so that they are updated correctly. 2578 */ 2579 return f2fs_enable_quota_files(sbi, readonly); 2580 } 2581 f2fs_recover_quota_end(struct f2fs_sb_info * sbi,bool quota_enabled)2582 static void f2fs_recover_quota_end(struct f2fs_sb_info *sbi, 2583 bool quota_enabled) 2584 { 2585 if (quota_enabled) 2586 f2fs_quota_off_umount(sbi->sb); 2587 2588 if (is_sbi_flag_set(sbi, SBI_IS_WRITABLE)) { 2589 clear_sbi_flag(sbi, SBI_IS_WRITABLE); 2590 sbi->sb->s_flags |= SB_RDONLY; 2591 } 2592 } 2593 2594 /* Read data from quotafile */ f2fs_quota_read(struct super_block * sb,int type,char * data,size_t len,loff_t off)2595 static ssize_t f2fs_quota_read(struct super_block *sb, int type, char *data, 2596 size_t len, loff_t off) 2597 { 2598 struct inode *inode = sb_dqopt(sb)->files[type]; 2599 struct address_space *mapping = inode->i_mapping; 2600 block_t blkidx = F2FS_BYTES_TO_BLK(off); 2601 int offset = off & (sb->s_blocksize - 1); 2602 int tocopy; 2603 size_t toread; 2604 loff_t i_size = i_size_read(inode); 2605 struct page *page; 2606 2607 if (off > i_size) 2608 return 0; 2609 2610 if (off + len > i_size) 2611 len = i_size - off; 2612 toread = len; 2613 while (toread > 0) { 2614 tocopy = min_t(unsigned long, sb->s_blocksize - offset, toread); 2615 repeat: 2616 page = read_cache_page_gfp(mapping, blkidx, GFP_NOFS); 2617 if (IS_ERR(page)) { 2618 if (PTR_ERR(page) == -ENOMEM) { 2619 congestion_wait(BLK_RW_ASYNC, 2620 DEFAULT_IO_TIMEOUT); 2621 goto repeat; 2622 } 2623 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR); 2624 return PTR_ERR(page); 2625 } 2626 2627 lock_page(page); 2628 2629 if (unlikely(page->mapping != mapping)) { 2630 f2fs_put_page(page, 1); 2631 goto repeat; 2632 } 2633 if (unlikely(!PageUptodate(page))) { 2634 f2fs_put_page(page, 1); 2635 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR); 2636 return -EIO; 2637 } 2638 2639 memcpy_from_page(data, page, offset, tocopy); 2640 f2fs_put_page(page, 1); 2641 2642 offset = 0; 2643 toread -= tocopy; 2644 data += tocopy; 2645 blkidx++; 2646 } 2647 return len; 2648 } 2649 2650 /* Write to quotafile */ f2fs_quota_write(struct super_block * sb,int type,const char * data,size_t len,loff_t off)2651 static ssize_t f2fs_quota_write(struct super_block *sb, int type, 2652 const char *data, size_t len, loff_t off) 2653 { 2654 struct inode *inode = sb_dqopt(sb)->files[type]; 2655 struct address_space *mapping = inode->i_mapping; 2656 const struct address_space_operations *a_ops = mapping->a_ops; 2657 int offset = off & (sb->s_blocksize - 1); 2658 size_t towrite = len; 2659 struct page *page; 2660 void *fsdata = NULL; 2661 int err = 0; 2662 int tocopy; 2663 2664 while (towrite > 0) { 2665 tocopy = min_t(unsigned long, sb->s_blocksize - offset, 2666 towrite); 2667 retry: 2668 err = a_ops->write_begin(NULL, mapping, off, tocopy, 0, 2669 &page, &fsdata); 2670 if (unlikely(err)) { 2671 if (err == -ENOMEM) { 2672 f2fs_io_schedule_timeout(DEFAULT_IO_TIMEOUT); 2673 goto retry; 2674 } 2675 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR); 2676 break; 2677 } 2678 2679 memcpy_to_page(page, offset, data, tocopy); 2680 2681 a_ops->write_end(NULL, mapping, off, tocopy, tocopy, 2682 page, fsdata); 2683 offset = 0; 2684 towrite -= tocopy; 2685 off += tocopy; 2686 data += tocopy; 2687 cond_resched(); 2688 } 2689 2690 if (len == towrite) 2691 return err; 2692 inode->i_mtime = inode->i_ctime = current_time(inode); 2693 f2fs_mark_inode_dirty_sync(inode, false); 2694 return len - towrite; 2695 } 2696 f2fs_dquot_initialize(struct inode * inode)2697 int f2fs_dquot_initialize(struct inode *inode) 2698 { 2699 if (time_to_inject(F2FS_I_SB(inode), FAULT_DQUOT_INIT)) 2700 return -ESRCH; 2701 2702 return dquot_initialize(inode); 2703 } 2704 f2fs_get_dquots(struct inode * inode)2705 static struct dquot **f2fs_get_dquots(struct inode *inode) 2706 { 2707 return F2FS_I(inode)->i_dquot; 2708 } 2709 f2fs_get_reserved_space(struct inode * inode)2710 static qsize_t *f2fs_get_reserved_space(struct inode *inode) 2711 { 2712 return &F2FS_I(inode)->i_reserved_quota; 2713 } 2714 f2fs_quota_on_mount(struct f2fs_sb_info * sbi,int type)2715 static int f2fs_quota_on_mount(struct f2fs_sb_info *sbi, int type) 2716 { 2717 if (is_set_ckpt_flags(sbi, CP_QUOTA_NEED_FSCK_FLAG)) { 2718 f2fs_err(sbi, "quota sysfile may be corrupted, skip loading it"); 2719 return 0; 2720 } 2721 2722 return dquot_quota_on_mount(sbi->sb, F2FS_OPTION(sbi).s_qf_names[type], 2723 F2FS_OPTION(sbi).s_jquota_fmt, type); 2724 } 2725 f2fs_enable_quota_files(struct f2fs_sb_info * sbi,bool rdonly)2726 int f2fs_enable_quota_files(struct f2fs_sb_info *sbi, bool rdonly) 2727 { 2728 int enabled = 0; 2729 int i, err; 2730 2731 if (f2fs_sb_has_quota_ino(sbi) && rdonly) { 2732 err = f2fs_enable_quotas(sbi->sb); 2733 if (err) { 2734 f2fs_err(sbi, "Cannot turn on quota_ino: %d", err); 2735 return 0; 2736 } 2737 return 1; 2738 } 2739 2740 for (i = 0; i < MAXQUOTAS; i++) { 2741 if (F2FS_OPTION(sbi).s_qf_names[i]) { 2742 err = f2fs_quota_on_mount(sbi, i); 2743 if (!err) { 2744 enabled = 1; 2745 continue; 2746 } 2747 f2fs_err(sbi, "Cannot turn on quotas: %d on %d", 2748 err, i); 2749 } 2750 } 2751 return enabled; 2752 } 2753 f2fs_quota_enable(struct super_block * sb,int type,int format_id,unsigned int flags)2754 static int f2fs_quota_enable(struct super_block *sb, int type, int format_id, 2755 unsigned int flags) 2756 { 2757 struct inode *qf_inode; 2758 unsigned long qf_inum; 2759 int err; 2760 2761 BUG_ON(!f2fs_sb_has_quota_ino(F2FS_SB(sb))); 2762 2763 qf_inum = f2fs_qf_ino(sb, type); 2764 if (!qf_inum) 2765 return -EPERM; 2766 2767 qf_inode = f2fs_iget(sb, qf_inum); 2768 if (IS_ERR(qf_inode)) { 2769 f2fs_err(F2FS_SB(sb), "Bad quota inode %u:%lu", type, qf_inum); 2770 return PTR_ERR(qf_inode); 2771 } 2772 2773 /* Don't account quota for quota files to avoid recursion */ 2774 qf_inode->i_flags |= S_NOQUOTA; 2775 err = dquot_load_quota_inode(qf_inode, type, format_id, flags); 2776 iput(qf_inode); 2777 return err; 2778 } 2779 f2fs_enable_quotas(struct super_block * sb)2780 static int f2fs_enable_quotas(struct super_block *sb) 2781 { 2782 struct f2fs_sb_info *sbi = F2FS_SB(sb); 2783 int type, err = 0; 2784 unsigned long qf_inum; 2785 bool quota_mopt[MAXQUOTAS] = { 2786 test_opt(sbi, USRQUOTA), 2787 test_opt(sbi, GRPQUOTA), 2788 test_opt(sbi, PRJQUOTA), 2789 }; 2790 2791 if (is_set_ckpt_flags(F2FS_SB(sb), CP_QUOTA_NEED_FSCK_FLAG)) { 2792 f2fs_err(sbi, "quota file may be corrupted, skip loading it"); 2793 return 0; 2794 } 2795 2796 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE; 2797 2798 for (type = 0; type < MAXQUOTAS; type++) { 2799 qf_inum = f2fs_qf_ino(sb, type); 2800 if (qf_inum) { 2801 err = f2fs_quota_enable(sb, type, QFMT_VFS_V1, 2802 DQUOT_USAGE_ENABLED | 2803 (quota_mopt[type] ? DQUOT_LIMITS_ENABLED : 0)); 2804 if (err) { 2805 f2fs_err(sbi, "Failed to enable quota tracking (type=%d, err=%d). Please run fsck to fix.", 2806 type, err); 2807 for (type--; type >= 0; type--) 2808 dquot_quota_off(sb, type); 2809 set_sbi_flag(F2FS_SB(sb), 2810 SBI_QUOTA_NEED_REPAIR); 2811 return err; 2812 } 2813 } 2814 } 2815 return 0; 2816 } 2817 f2fs_quota_sync_file(struct f2fs_sb_info * sbi,int type)2818 static int f2fs_quota_sync_file(struct f2fs_sb_info *sbi, int type) 2819 { 2820 struct quota_info *dqopt = sb_dqopt(sbi->sb); 2821 struct address_space *mapping = dqopt->files[type]->i_mapping; 2822 int ret = 0; 2823 2824 ret = dquot_writeback_dquots(sbi->sb, type); 2825 if (ret) 2826 goto out; 2827 2828 ret = filemap_fdatawrite(mapping); 2829 if (ret) 2830 goto out; 2831 2832 /* if we are using journalled quota */ 2833 if (is_journalled_quota(sbi)) 2834 goto out; 2835 2836 ret = filemap_fdatawait(mapping); 2837 2838 truncate_inode_pages(&dqopt->files[type]->i_data, 0); 2839 out: 2840 if (ret) 2841 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR); 2842 return ret; 2843 } 2844 f2fs_quota_sync(struct super_block * sb,int type)2845 int f2fs_quota_sync(struct super_block *sb, int type) 2846 { 2847 struct f2fs_sb_info *sbi = F2FS_SB(sb); 2848 struct quota_info *dqopt = sb_dqopt(sb); 2849 int cnt; 2850 int ret = 0; 2851 2852 /* 2853 * Now when everything is written we can discard the pagecache so 2854 * that userspace sees the changes. 2855 */ 2856 for (cnt = 0; cnt < MAXQUOTAS; cnt++) { 2857 2858 if (type != -1 && cnt != type) 2859 continue; 2860 2861 if (!sb_has_quota_active(sb, cnt)) 2862 continue; 2863 2864 if (!f2fs_sb_has_quota_ino(sbi)) 2865 inode_lock(dqopt->files[cnt]); 2866 2867 /* 2868 * do_quotactl 2869 * f2fs_quota_sync 2870 * f2fs_down_read(quota_sem) 2871 * dquot_writeback_dquots() 2872 * f2fs_dquot_commit 2873 * block_operation 2874 * f2fs_down_read(quota_sem) 2875 */ 2876 f2fs_lock_op(sbi); 2877 f2fs_down_read(&sbi->quota_sem); 2878 2879 ret = f2fs_quota_sync_file(sbi, cnt); 2880 2881 f2fs_up_read(&sbi->quota_sem); 2882 f2fs_unlock_op(sbi); 2883 2884 if (!f2fs_sb_has_quota_ino(sbi)) 2885 inode_unlock(dqopt->files[cnt]); 2886 2887 if (ret) 2888 break; 2889 } 2890 return ret; 2891 } 2892 f2fs_quota_on(struct super_block * sb,int type,int format_id,const struct path * path)2893 static int f2fs_quota_on(struct super_block *sb, int type, int format_id, 2894 const struct path *path) 2895 { 2896 struct inode *inode; 2897 int err; 2898 2899 /* if quota sysfile exists, deny enabling quota with specific file */ 2900 if (f2fs_sb_has_quota_ino(F2FS_SB(sb))) { 2901 f2fs_err(F2FS_SB(sb), "quota sysfile already exists"); 2902 return -EBUSY; 2903 } 2904 2905 err = f2fs_quota_sync(sb, type); 2906 if (err) 2907 return err; 2908 2909 err = dquot_quota_on(sb, type, format_id, path); 2910 if (err) 2911 return err; 2912 2913 inode = d_inode(path->dentry); 2914 2915 inode_lock(inode); 2916 F2FS_I(inode)->i_flags |= F2FS_NOATIME_FL | F2FS_IMMUTABLE_FL; 2917 f2fs_set_inode_flags(inode); 2918 inode_unlock(inode); 2919 f2fs_mark_inode_dirty_sync(inode, false); 2920 2921 return 0; 2922 } 2923 __f2fs_quota_off(struct super_block * sb,int type)2924 static int __f2fs_quota_off(struct super_block *sb, int type) 2925 { 2926 struct inode *inode = sb_dqopt(sb)->files[type]; 2927 int err; 2928 2929 if (!inode || !igrab(inode)) 2930 return dquot_quota_off(sb, type); 2931 2932 err = f2fs_quota_sync(sb, type); 2933 if (err) 2934 goto out_put; 2935 2936 err = dquot_quota_off(sb, type); 2937 if (err || f2fs_sb_has_quota_ino(F2FS_SB(sb))) 2938 goto out_put; 2939 2940 inode_lock(inode); 2941 F2FS_I(inode)->i_flags &= ~(F2FS_NOATIME_FL | F2FS_IMMUTABLE_FL); 2942 f2fs_set_inode_flags(inode); 2943 inode_unlock(inode); 2944 f2fs_mark_inode_dirty_sync(inode, false); 2945 out_put: 2946 iput(inode); 2947 return err; 2948 } 2949 f2fs_quota_off(struct super_block * sb,int type)2950 static int f2fs_quota_off(struct super_block *sb, int type) 2951 { 2952 struct f2fs_sb_info *sbi = F2FS_SB(sb); 2953 int err; 2954 2955 err = __f2fs_quota_off(sb, type); 2956 2957 /* 2958 * quotactl can shutdown journalled quota, result in inconsistence 2959 * between quota record and fs data by following updates, tag the 2960 * flag to let fsck be aware of it. 2961 */ 2962 if (is_journalled_quota(sbi)) 2963 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR); 2964 return err; 2965 } 2966 f2fs_quota_off_umount(struct super_block * sb)2967 void f2fs_quota_off_umount(struct super_block *sb) 2968 { 2969 int type; 2970 int err; 2971 2972 for (type = 0; type < MAXQUOTAS; type++) { 2973 err = __f2fs_quota_off(sb, type); 2974 if (err) { 2975 int ret = dquot_quota_off(sb, type); 2976 2977 f2fs_err(F2FS_SB(sb), "Fail to turn off disk quota (type: %d, err: %d, ret:%d), Please run fsck to fix it.", 2978 type, err, ret); 2979 set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR); 2980 } 2981 } 2982 /* 2983 * In case of checkpoint=disable, we must flush quota blocks. 2984 * This can cause NULL exception for node_inode in end_io, since 2985 * put_super already dropped it. 2986 */ 2987 sync_filesystem(sb); 2988 } 2989 f2fs_truncate_quota_inode_pages(struct super_block * sb)2990 static void f2fs_truncate_quota_inode_pages(struct super_block *sb) 2991 { 2992 struct quota_info *dqopt = sb_dqopt(sb); 2993 int type; 2994 2995 for (type = 0; type < MAXQUOTAS; type++) { 2996 if (!dqopt->files[type]) 2997 continue; 2998 f2fs_inode_synced(dqopt->files[type]); 2999 } 3000 } 3001 f2fs_dquot_commit(struct dquot * dquot)3002 static int f2fs_dquot_commit(struct dquot *dquot) 3003 { 3004 struct f2fs_sb_info *sbi = F2FS_SB(dquot->dq_sb); 3005 int ret; 3006 3007 f2fs_down_read_nested(&sbi->quota_sem, SINGLE_DEPTH_NESTING); 3008 ret = dquot_commit(dquot); 3009 if (ret < 0) 3010 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR); 3011 f2fs_up_read(&sbi->quota_sem); 3012 return ret; 3013 } 3014 f2fs_dquot_acquire(struct dquot * dquot)3015 static int f2fs_dquot_acquire(struct dquot *dquot) 3016 { 3017 struct f2fs_sb_info *sbi = F2FS_SB(dquot->dq_sb); 3018 int ret; 3019 3020 f2fs_down_read(&sbi->quota_sem); 3021 ret = dquot_acquire(dquot); 3022 if (ret < 0) 3023 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR); 3024 f2fs_up_read(&sbi->quota_sem); 3025 return ret; 3026 } 3027 f2fs_dquot_release(struct dquot * dquot)3028 static int f2fs_dquot_release(struct dquot *dquot) 3029 { 3030 struct f2fs_sb_info *sbi = F2FS_SB(dquot->dq_sb); 3031 int ret = dquot_release(dquot); 3032 3033 if (ret < 0) 3034 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR); 3035 return ret; 3036 } 3037 f2fs_dquot_mark_dquot_dirty(struct dquot * dquot)3038 static int f2fs_dquot_mark_dquot_dirty(struct dquot *dquot) 3039 { 3040 struct super_block *sb = dquot->dq_sb; 3041 struct f2fs_sb_info *sbi = F2FS_SB(sb); 3042 int ret = dquot_mark_dquot_dirty(dquot); 3043 3044 /* if we are using journalled quota */ 3045 if (is_journalled_quota(sbi)) 3046 set_sbi_flag(sbi, SBI_QUOTA_NEED_FLUSH); 3047 3048 return ret; 3049 } 3050 f2fs_dquot_commit_info(struct super_block * sb,int type)3051 static int f2fs_dquot_commit_info(struct super_block *sb, int type) 3052 { 3053 struct f2fs_sb_info *sbi = F2FS_SB(sb); 3054 int ret = dquot_commit_info(sb, type); 3055 3056 if (ret < 0) 3057 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR); 3058 return ret; 3059 } 3060 f2fs_get_projid(struct inode * inode,kprojid_t * projid)3061 static int f2fs_get_projid(struct inode *inode, kprojid_t *projid) 3062 { 3063 *projid = F2FS_I(inode)->i_projid; 3064 return 0; 3065 } 3066 3067 static const struct dquot_operations f2fs_quota_operations = { 3068 .get_reserved_space = f2fs_get_reserved_space, 3069 .write_dquot = f2fs_dquot_commit, 3070 .acquire_dquot = f2fs_dquot_acquire, 3071 .release_dquot = f2fs_dquot_release, 3072 .mark_dirty = f2fs_dquot_mark_dquot_dirty, 3073 .write_info = f2fs_dquot_commit_info, 3074 .alloc_dquot = dquot_alloc, 3075 .destroy_dquot = dquot_destroy, 3076 .get_projid = f2fs_get_projid, 3077 .get_next_id = dquot_get_next_id, 3078 }; 3079 3080 static const struct quotactl_ops f2fs_quotactl_ops = { 3081 .quota_on = f2fs_quota_on, 3082 .quota_off = f2fs_quota_off, 3083 .quota_sync = f2fs_quota_sync, 3084 .get_state = dquot_get_state, 3085 .set_info = dquot_set_dqinfo, 3086 .get_dqblk = dquot_get_dqblk, 3087 .set_dqblk = dquot_set_dqblk, 3088 .get_nextdqblk = dquot_get_next_dqblk, 3089 }; 3090 #else f2fs_dquot_initialize(struct inode * inode)3091 int f2fs_dquot_initialize(struct inode *inode) 3092 { 3093 return 0; 3094 } 3095 f2fs_quota_sync(struct super_block * sb,int type)3096 int f2fs_quota_sync(struct super_block *sb, int type) 3097 { 3098 return 0; 3099 } 3100 f2fs_quota_off_umount(struct super_block * sb)3101 void f2fs_quota_off_umount(struct super_block *sb) 3102 { 3103 } 3104 #endif 3105 3106 static const struct super_operations f2fs_sops = { 3107 .alloc_inode = f2fs_alloc_inode, 3108 .free_inode = f2fs_free_inode, 3109 .drop_inode = f2fs_drop_inode, 3110 .write_inode = f2fs_write_inode, 3111 .dirty_inode = f2fs_dirty_inode, 3112 .show_options = f2fs_show_options, 3113 #ifdef CONFIG_QUOTA 3114 .quota_read = f2fs_quota_read, 3115 .quota_write = f2fs_quota_write, 3116 .get_dquots = f2fs_get_dquots, 3117 #endif 3118 .evict_inode = f2fs_evict_inode, 3119 .put_super = f2fs_put_super, 3120 .sync_fs = f2fs_sync_fs, 3121 .freeze_fs = f2fs_freeze, 3122 .unfreeze_fs = f2fs_unfreeze, 3123 .statfs = f2fs_statfs, 3124 .remount_fs = f2fs_remount, 3125 }; 3126 3127 #ifdef CONFIG_FS_ENCRYPTION f2fs_get_context(struct inode * inode,void * ctx,size_t len)3128 static int f2fs_get_context(struct inode *inode, void *ctx, size_t len) 3129 { 3130 return f2fs_getxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION, 3131 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT, 3132 ctx, len, NULL); 3133 } 3134 f2fs_set_context(struct inode * inode,const void * ctx,size_t len,void * fs_data)3135 static int f2fs_set_context(struct inode *inode, const void *ctx, size_t len, 3136 void *fs_data) 3137 { 3138 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 3139 3140 /* 3141 * Encrypting the root directory is not allowed because fsck 3142 * expects lost+found directory to exist and remain unencrypted 3143 * if LOST_FOUND feature is enabled. 3144 * 3145 */ 3146 if (f2fs_sb_has_lost_found(sbi) && 3147 inode->i_ino == F2FS_ROOT_INO(sbi)) 3148 return -EPERM; 3149 3150 return f2fs_setxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION, 3151 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT, 3152 ctx, len, fs_data, XATTR_CREATE); 3153 } 3154 f2fs_get_dummy_policy(struct super_block * sb)3155 static const union fscrypt_policy *f2fs_get_dummy_policy(struct super_block *sb) 3156 { 3157 return F2FS_OPTION(F2FS_SB(sb)).dummy_enc_policy.policy; 3158 } 3159 f2fs_has_stable_inodes(struct super_block * sb)3160 static bool f2fs_has_stable_inodes(struct super_block *sb) 3161 { 3162 return true; 3163 } 3164 f2fs_get_ino_and_lblk_bits(struct super_block * sb,int * ino_bits_ret,int * lblk_bits_ret)3165 static void f2fs_get_ino_and_lblk_bits(struct super_block *sb, 3166 int *ino_bits_ret, int *lblk_bits_ret) 3167 { 3168 *ino_bits_ret = 8 * sizeof(nid_t); 3169 *lblk_bits_ret = 8 * sizeof(block_t); 3170 } 3171 f2fs_get_devices(struct super_block * sb,unsigned int * num_devs)3172 static struct block_device **f2fs_get_devices(struct super_block *sb, 3173 unsigned int *num_devs) 3174 { 3175 struct f2fs_sb_info *sbi = F2FS_SB(sb); 3176 struct block_device **devs; 3177 int i; 3178 3179 if (!f2fs_is_multi_device(sbi)) 3180 return NULL; 3181 3182 devs = kmalloc_array(sbi->s_ndevs, sizeof(*devs), GFP_KERNEL); 3183 if (!devs) 3184 return ERR_PTR(-ENOMEM); 3185 3186 for (i = 0; i < sbi->s_ndevs; i++) 3187 devs[i] = FDEV(i).bdev; 3188 *num_devs = sbi->s_ndevs; 3189 return devs; 3190 } 3191 3192 static const struct fscrypt_operations f2fs_cryptops = { 3193 .flags = FS_CFLG_SUPPORTS_SUBBLOCK_DATA_UNITS, 3194 .key_prefix = "f2fs:", 3195 .get_context = f2fs_get_context, 3196 .set_context = f2fs_set_context, 3197 .get_dummy_policy = f2fs_get_dummy_policy, 3198 .empty_dir = f2fs_empty_dir, 3199 .has_stable_inodes = f2fs_has_stable_inodes, 3200 .get_ino_and_lblk_bits = f2fs_get_ino_and_lblk_bits, 3201 .get_devices = f2fs_get_devices, 3202 }; 3203 #endif 3204 f2fs_nfs_get_inode(struct super_block * sb,u64 ino,u32 generation)3205 static struct inode *f2fs_nfs_get_inode(struct super_block *sb, 3206 u64 ino, u32 generation) 3207 { 3208 struct f2fs_sb_info *sbi = F2FS_SB(sb); 3209 struct inode *inode; 3210 3211 if (f2fs_check_nid_range(sbi, ino)) 3212 return ERR_PTR(-ESTALE); 3213 3214 /* 3215 * f2fs_iget isn't quite right if the inode is currently unallocated! 3216 * However f2fs_iget currently does appropriate checks to handle stale 3217 * inodes so everything is OK. 3218 */ 3219 inode = f2fs_iget(sb, ino); 3220 if (IS_ERR(inode)) 3221 return ERR_CAST(inode); 3222 if (unlikely(generation && inode->i_generation != generation)) { 3223 /* we didn't find the right inode.. */ 3224 iput(inode); 3225 return ERR_PTR(-ESTALE); 3226 } 3227 return inode; 3228 } 3229 f2fs_fh_to_dentry(struct super_block * sb,struct fid * fid,int fh_len,int fh_type)3230 static struct dentry *f2fs_fh_to_dentry(struct super_block *sb, struct fid *fid, 3231 int fh_len, int fh_type) 3232 { 3233 return generic_fh_to_dentry(sb, fid, fh_len, fh_type, 3234 f2fs_nfs_get_inode); 3235 } 3236 f2fs_fh_to_parent(struct super_block * sb,struct fid * fid,int fh_len,int fh_type)3237 static struct dentry *f2fs_fh_to_parent(struct super_block *sb, struct fid *fid, 3238 int fh_len, int fh_type) 3239 { 3240 return generic_fh_to_parent(sb, fid, fh_len, fh_type, 3241 f2fs_nfs_get_inode); 3242 } 3243 3244 static const struct export_operations f2fs_export_ops = { 3245 .fh_to_dentry = f2fs_fh_to_dentry, 3246 .fh_to_parent = f2fs_fh_to_parent, 3247 .get_parent = f2fs_get_parent, 3248 }; 3249 max_file_blocks(struct inode * inode)3250 loff_t max_file_blocks(struct inode *inode) 3251 { 3252 loff_t result = 0; 3253 loff_t leaf_count; 3254 3255 /* 3256 * note: previously, result is equal to (DEF_ADDRS_PER_INODE - 3257 * DEFAULT_INLINE_XATTR_ADDRS), but now f2fs try to reserve more 3258 * space in inode.i_addr, it will be more safe to reassign 3259 * result as zero. 3260 */ 3261 3262 if (inode && f2fs_compressed_file(inode)) 3263 leaf_count = ADDRS_PER_BLOCK(inode); 3264 else 3265 leaf_count = DEF_ADDRS_PER_BLOCK; 3266 3267 /* two direct node blocks */ 3268 result += (leaf_count * 2); 3269 3270 /* two indirect node blocks */ 3271 leaf_count *= NIDS_PER_BLOCK; 3272 result += (leaf_count * 2); 3273 3274 /* one double indirect node block */ 3275 leaf_count *= NIDS_PER_BLOCK; 3276 result += leaf_count; 3277 3278 /* 3279 * For compatibility with FSCRYPT_POLICY_FLAG_IV_INO_LBLK_{64,32} with 3280 * a 4K crypto data unit, we must restrict the max filesize to what can 3281 * fit within U32_MAX + 1 data units. 3282 */ 3283 3284 result = min(result, (((loff_t)U32_MAX + 1) * 4096) >> F2FS_BLKSIZE_BITS); 3285 3286 return result; 3287 } 3288 __f2fs_commit_super(struct buffer_head * bh,struct f2fs_super_block * super)3289 static int __f2fs_commit_super(struct buffer_head *bh, 3290 struct f2fs_super_block *super) 3291 { 3292 lock_buffer(bh); 3293 if (super) 3294 memcpy(bh->b_data + F2FS_SUPER_OFFSET, super, sizeof(*super)); 3295 set_buffer_dirty(bh); 3296 unlock_buffer(bh); 3297 3298 /* it's rare case, we can do fua all the time */ 3299 return __sync_dirty_buffer(bh, REQ_SYNC | REQ_PREFLUSH | REQ_FUA); 3300 } 3301 sanity_check_area_boundary(struct f2fs_sb_info * sbi,struct buffer_head * bh)3302 static inline bool sanity_check_area_boundary(struct f2fs_sb_info *sbi, 3303 struct buffer_head *bh) 3304 { 3305 struct f2fs_super_block *raw_super = (struct f2fs_super_block *) 3306 (bh->b_data + F2FS_SUPER_OFFSET); 3307 struct super_block *sb = sbi->sb; 3308 u32 segment0_blkaddr = le32_to_cpu(raw_super->segment0_blkaddr); 3309 u32 cp_blkaddr = le32_to_cpu(raw_super->cp_blkaddr); 3310 u32 sit_blkaddr = le32_to_cpu(raw_super->sit_blkaddr); 3311 u32 nat_blkaddr = le32_to_cpu(raw_super->nat_blkaddr); 3312 u32 ssa_blkaddr = le32_to_cpu(raw_super->ssa_blkaddr); 3313 u32 main_blkaddr = le32_to_cpu(raw_super->main_blkaddr); 3314 u32 segment_count_ckpt = le32_to_cpu(raw_super->segment_count_ckpt); 3315 u32 segment_count_sit = le32_to_cpu(raw_super->segment_count_sit); 3316 u32 segment_count_nat = le32_to_cpu(raw_super->segment_count_nat); 3317 u32 segment_count_ssa = le32_to_cpu(raw_super->segment_count_ssa); 3318 u32 segment_count_main = le32_to_cpu(raw_super->segment_count_main); 3319 u32 segment_count = le32_to_cpu(raw_super->segment_count); 3320 u32 log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg); 3321 u64 main_end_blkaddr = main_blkaddr + 3322 (segment_count_main << log_blocks_per_seg); 3323 u64 seg_end_blkaddr = segment0_blkaddr + 3324 (segment_count << log_blocks_per_seg); 3325 3326 if (segment0_blkaddr != cp_blkaddr) { 3327 f2fs_info(sbi, "Mismatch start address, segment0(%u) cp_blkaddr(%u)", 3328 segment0_blkaddr, cp_blkaddr); 3329 return true; 3330 } 3331 3332 if (cp_blkaddr + (segment_count_ckpt << log_blocks_per_seg) != 3333 sit_blkaddr) { 3334 f2fs_info(sbi, "Wrong CP boundary, start(%u) end(%u) blocks(%u)", 3335 cp_blkaddr, sit_blkaddr, 3336 segment_count_ckpt << log_blocks_per_seg); 3337 return true; 3338 } 3339 3340 if (sit_blkaddr + (segment_count_sit << log_blocks_per_seg) != 3341 nat_blkaddr) { 3342 f2fs_info(sbi, "Wrong SIT boundary, start(%u) end(%u) blocks(%u)", 3343 sit_blkaddr, nat_blkaddr, 3344 segment_count_sit << log_blocks_per_seg); 3345 return true; 3346 } 3347 3348 if (nat_blkaddr + (segment_count_nat << log_blocks_per_seg) != 3349 ssa_blkaddr) { 3350 f2fs_info(sbi, "Wrong NAT boundary, start(%u) end(%u) blocks(%u)", 3351 nat_blkaddr, ssa_blkaddr, 3352 segment_count_nat << log_blocks_per_seg); 3353 return true; 3354 } 3355 3356 if (ssa_blkaddr + (segment_count_ssa << log_blocks_per_seg) != 3357 main_blkaddr) { 3358 f2fs_info(sbi, "Wrong SSA boundary, start(%u) end(%u) blocks(%u)", 3359 ssa_blkaddr, main_blkaddr, 3360 segment_count_ssa << log_blocks_per_seg); 3361 return true; 3362 } 3363 3364 if (main_end_blkaddr > seg_end_blkaddr) { 3365 f2fs_info(sbi, "Wrong MAIN_AREA boundary, start(%u) end(%llu) block(%u)", 3366 main_blkaddr, seg_end_blkaddr, 3367 segment_count_main << log_blocks_per_seg); 3368 return true; 3369 } else if (main_end_blkaddr < seg_end_blkaddr) { 3370 int err = 0; 3371 char *res; 3372 3373 /* fix in-memory information all the time */ 3374 raw_super->segment_count = cpu_to_le32((main_end_blkaddr - 3375 segment0_blkaddr) >> log_blocks_per_seg); 3376 3377 if (f2fs_readonly(sb) || f2fs_hw_is_readonly(sbi)) { 3378 set_sbi_flag(sbi, SBI_NEED_SB_WRITE); 3379 res = "internally"; 3380 } else { 3381 err = __f2fs_commit_super(bh, NULL); 3382 res = err ? "failed" : "done"; 3383 } 3384 f2fs_info(sbi, "Fix alignment : %s, start(%u) end(%llu) block(%u)", 3385 res, main_blkaddr, seg_end_blkaddr, 3386 segment_count_main << log_blocks_per_seg); 3387 if (err) 3388 return true; 3389 } 3390 return false; 3391 } 3392 sanity_check_raw_super(struct f2fs_sb_info * sbi,struct buffer_head * bh)3393 static int sanity_check_raw_super(struct f2fs_sb_info *sbi, 3394 struct buffer_head *bh) 3395 { 3396 block_t segment_count, segs_per_sec, secs_per_zone, segment_count_main; 3397 block_t total_sections, blocks_per_seg; 3398 struct f2fs_super_block *raw_super = (struct f2fs_super_block *) 3399 (bh->b_data + F2FS_SUPER_OFFSET); 3400 size_t crc_offset = 0; 3401 __u32 crc = 0; 3402 3403 if (le32_to_cpu(raw_super->magic) != F2FS_SUPER_MAGIC) { 3404 f2fs_info(sbi, "Magic Mismatch, valid(0x%x) - read(0x%x)", 3405 F2FS_SUPER_MAGIC, le32_to_cpu(raw_super->magic)); 3406 return -EINVAL; 3407 } 3408 3409 /* Check checksum_offset and crc in superblock */ 3410 if (__F2FS_HAS_FEATURE(raw_super, F2FS_FEATURE_SB_CHKSUM)) { 3411 crc_offset = le32_to_cpu(raw_super->checksum_offset); 3412 if (crc_offset != 3413 offsetof(struct f2fs_super_block, crc)) { 3414 f2fs_info(sbi, "Invalid SB checksum offset: %zu", 3415 crc_offset); 3416 return -EFSCORRUPTED; 3417 } 3418 crc = le32_to_cpu(raw_super->crc); 3419 if (!f2fs_crc_valid(sbi, crc, raw_super, crc_offset)) { 3420 f2fs_info(sbi, "Invalid SB checksum value: %u", crc); 3421 return -EFSCORRUPTED; 3422 } 3423 } 3424 3425 /* Currently, support only 4KB block size */ 3426 if (le32_to_cpu(raw_super->log_blocksize) != F2FS_BLKSIZE_BITS) { 3427 f2fs_info(sbi, "Invalid log_blocksize (%u), supports only %u", 3428 le32_to_cpu(raw_super->log_blocksize), 3429 F2FS_BLKSIZE_BITS); 3430 return -EFSCORRUPTED; 3431 } 3432 3433 /* check log blocks per segment */ 3434 if (le32_to_cpu(raw_super->log_blocks_per_seg) != 9) { 3435 f2fs_info(sbi, "Invalid log blocks per segment (%u)", 3436 le32_to_cpu(raw_super->log_blocks_per_seg)); 3437 return -EFSCORRUPTED; 3438 } 3439 3440 /* Currently, support 512/1024/2048/4096/16K bytes sector size */ 3441 if (le32_to_cpu(raw_super->log_sectorsize) > 3442 F2FS_MAX_LOG_SECTOR_SIZE || 3443 le32_to_cpu(raw_super->log_sectorsize) < 3444 F2FS_MIN_LOG_SECTOR_SIZE) { 3445 f2fs_info(sbi, "Invalid log sectorsize (%u)", 3446 le32_to_cpu(raw_super->log_sectorsize)); 3447 return -EFSCORRUPTED; 3448 } 3449 if (le32_to_cpu(raw_super->log_sectors_per_block) + 3450 le32_to_cpu(raw_super->log_sectorsize) != 3451 F2FS_MAX_LOG_SECTOR_SIZE) { 3452 f2fs_info(sbi, "Invalid log sectors per block(%u) log sectorsize(%u)", 3453 le32_to_cpu(raw_super->log_sectors_per_block), 3454 le32_to_cpu(raw_super->log_sectorsize)); 3455 return -EFSCORRUPTED; 3456 } 3457 3458 segment_count = le32_to_cpu(raw_super->segment_count); 3459 segment_count_main = le32_to_cpu(raw_super->segment_count_main); 3460 segs_per_sec = le32_to_cpu(raw_super->segs_per_sec); 3461 secs_per_zone = le32_to_cpu(raw_super->secs_per_zone); 3462 total_sections = le32_to_cpu(raw_super->section_count); 3463 3464 /* blocks_per_seg should be 512, given the above check */ 3465 blocks_per_seg = BIT(le32_to_cpu(raw_super->log_blocks_per_seg)); 3466 3467 if (segment_count > F2FS_MAX_SEGMENT || 3468 segment_count < F2FS_MIN_SEGMENTS) { 3469 f2fs_info(sbi, "Invalid segment count (%u)", segment_count); 3470 return -EFSCORRUPTED; 3471 } 3472 3473 if (total_sections > segment_count_main || total_sections < 1 || 3474 segs_per_sec > segment_count || !segs_per_sec) { 3475 f2fs_info(sbi, "Invalid segment/section count (%u, %u x %u)", 3476 segment_count, total_sections, segs_per_sec); 3477 return -EFSCORRUPTED; 3478 } 3479 3480 if (segment_count_main != total_sections * segs_per_sec) { 3481 f2fs_info(sbi, "Invalid segment/section count (%u != %u * %u)", 3482 segment_count_main, total_sections, segs_per_sec); 3483 return -EFSCORRUPTED; 3484 } 3485 3486 if ((segment_count / segs_per_sec) < total_sections) { 3487 f2fs_info(sbi, "Small segment_count (%u < %u * %u)", 3488 segment_count, segs_per_sec, total_sections); 3489 return -EFSCORRUPTED; 3490 } 3491 3492 if (segment_count > (le64_to_cpu(raw_super->block_count) >> 9)) { 3493 f2fs_info(sbi, "Wrong segment_count / block_count (%u > %llu)", 3494 segment_count, le64_to_cpu(raw_super->block_count)); 3495 return -EFSCORRUPTED; 3496 } 3497 3498 if (RDEV(0).path[0]) { 3499 block_t dev_seg_count = le32_to_cpu(RDEV(0).total_segments); 3500 int i = 1; 3501 3502 while (i < MAX_DEVICES && RDEV(i).path[0]) { 3503 dev_seg_count += le32_to_cpu(RDEV(i).total_segments); 3504 i++; 3505 } 3506 if (segment_count != dev_seg_count) { 3507 f2fs_info(sbi, "Segment count (%u) mismatch with total segments from devices (%u)", 3508 segment_count, dev_seg_count); 3509 return -EFSCORRUPTED; 3510 } 3511 } else { 3512 if (__F2FS_HAS_FEATURE(raw_super, F2FS_FEATURE_BLKZONED) && 3513 !bdev_is_zoned(sbi->sb->s_bdev)) { 3514 f2fs_info(sbi, "Zoned block device path is missing"); 3515 return -EFSCORRUPTED; 3516 } 3517 } 3518 3519 if (secs_per_zone > total_sections || !secs_per_zone) { 3520 f2fs_info(sbi, "Wrong secs_per_zone / total_sections (%u, %u)", 3521 secs_per_zone, total_sections); 3522 return -EFSCORRUPTED; 3523 } 3524 if (le32_to_cpu(raw_super->extension_count) > F2FS_MAX_EXTENSION || 3525 raw_super->hot_ext_count > F2FS_MAX_EXTENSION || 3526 (le32_to_cpu(raw_super->extension_count) + 3527 raw_super->hot_ext_count) > F2FS_MAX_EXTENSION) { 3528 f2fs_info(sbi, "Corrupted extension count (%u + %u > %u)", 3529 le32_to_cpu(raw_super->extension_count), 3530 raw_super->hot_ext_count, 3531 F2FS_MAX_EXTENSION); 3532 return -EFSCORRUPTED; 3533 } 3534 3535 if (le32_to_cpu(raw_super->cp_payload) >= 3536 (blocks_per_seg - F2FS_CP_PACKS - 3537 NR_CURSEG_PERSIST_TYPE)) { 3538 f2fs_info(sbi, "Insane cp_payload (%u >= %u)", 3539 le32_to_cpu(raw_super->cp_payload), 3540 blocks_per_seg - F2FS_CP_PACKS - 3541 NR_CURSEG_PERSIST_TYPE); 3542 return -EFSCORRUPTED; 3543 } 3544 3545 /* check reserved ino info */ 3546 if (le32_to_cpu(raw_super->node_ino) != 1 || 3547 le32_to_cpu(raw_super->meta_ino) != 2 || 3548 le32_to_cpu(raw_super->root_ino) != 3) { 3549 f2fs_info(sbi, "Invalid Fs Meta Ino: node(%u) meta(%u) root(%u)", 3550 le32_to_cpu(raw_super->node_ino), 3551 le32_to_cpu(raw_super->meta_ino), 3552 le32_to_cpu(raw_super->root_ino)); 3553 return -EFSCORRUPTED; 3554 } 3555 3556 /* check CP/SIT/NAT/SSA/MAIN_AREA area boundary */ 3557 if (sanity_check_area_boundary(sbi, bh)) 3558 return -EFSCORRUPTED; 3559 3560 return 0; 3561 } 3562 f2fs_sanity_check_ckpt(struct f2fs_sb_info * sbi)3563 int f2fs_sanity_check_ckpt(struct f2fs_sb_info *sbi) 3564 { 3565 unsigned int total, fsmeta; 3566 struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi); 3567 struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi); 3568 unsigned int ovp_segments, reserved_segments; 3569 unsigned int main_segs, blocks_per_seg; 3570 unsigned int sit_segs, nat_segs; 3571 unsigned int sit_bitmap_size, nat_bitmap_size; 3572 unsigned int log_blocks_per_seg; 3573 unsigned int segment_count_main; 3574 unsigned int cp_pack_start_sum, cp_payload; 3575 block_t user_block_count, valid_user_blocks; 3576 block_t avail_node_count, valid_node_count; 3577 unsigned int nat_blocks, nat_bits_bytes, nat_bits_blocks; 3578 int i, j; 3579 3580 total = le32_to_cpu(raw_super->segment_count); 3581 fsmeta = le32_to_cpu(raw_super->segment_count_ckpt); 3582 sit_segs = le32_to_cpu(raw_super->segment_count_sit); 3583 fsmeta += sit_segs; 3584 nat_segs = le32_to_cpu(raw_super->segment_count_nat); 3585 fsmeta += nat_segs; 3586 fsmeta += le32_to_cpu(ckpt->rsvd_segment_count); 3587 fsmeta += le32_to_cpu(raw_super->segment_count_ssa); 3588 3589 if (unlikely(fsmeta >= total)) 3590 return 1; 3591 3592 ovp_segments = le32_to_cpu(ckpt->overprov_segment_count); 3593 reserved_segments = le32_to_cpu(ckpt->rsvd_segment_count); 3594 3595 if (!f2fs_sb_has_readonly(sbi) && 3596 unlikely(fsmeta < F2FS_MIN_META_SEGMENTS || 3597 ovp_segments == 0 || reserved_segments == 0)) { 3598 f2fs_err(sbi, "Wrong layout: check mkfs.f2fs version"); 3599 return 1; 3600 } 3601 user_block_count = le64_to_cpu(ckpt->user_block_count); 3602 segment_count_main = le32_to_cpu(raw_super->segment_count_main) + 3603 (f2fs_sb_has_readonly(sbi) ? 1 : 0); 3604 log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg); 3605 if (!user_block_count || user_block_count >= 3606 segment_count_main << log_blocks_per_seg) { 3607 f2fs_err(sbi, "Wrong user_block_count: %u", 3608 user_block_count); 3609 return 1; 3610 } 3611 3612 valid_user_blocks = le64_to_cpu(ckpt->valid_block_count); 3613 if (valid_user_blocks > user_block_count) { 3614 f2fs_err(sbi, "Wrong valid_user_blocks: %u, user_block_count: %u", 3615 valid_user_blocks, user_block_count); 3616 return 1; 3617 } 3618 3619 valid_node_count = le32_to_cpu(ckpt->valid_node_count); 3620 avail_node_count = sbi->total_node_count - F2FS_RESERVED_NODE_NUM; 3621 if (valid_node_count > avail_node_count) { 3622 f2fs_err(sbi, "Wrong valid_node_count: %u, avail_node_count: %u", 3623 valid_node_count, avail_node_count); 3624 return 1; 3625 } 3626 3627 main_segs = le32_to_cpu(raw_super->segment_count_main); 3628 blocks_per_seg = sbi->blocks_per_seg; 3629 3630 for (i = 0; i < NR_CURSEG_NODE_TYPE; i++) { 3631 if (le32_to_cpu(ckpt->cur_node_segno[i]) >= main_segs || 3632 le16_to_cpu(ckpt->cur_node_blkoff[i]) >= blocks_per_seg) 3633 return 1; 3634 3635 if (f2fs_sb_has_readonly(sbi)) 3636 goto check_data; 3637 3638 for (j = i + 1; j < NR_CURSEG_NODE_TYPE; j++) { 3639 if (le32_to_cpu(ckpt->cur_node_segno[i]) == 3640 le32_to_cpu(ckpt->cur_node_segno[j])) { 3641 f2fs_err(sbi, "Node segment (%u, %u) has the same segno: %u", 3642 i, j, 3643 le32_to_cpu(ckpt->cur_node_segno[i])); 3644 return 1; 3645 } 3646 } 3647 } 3648 check_data: 3649 for (i = 0; i < NR_CURSEG_DATA_TYPE; i++) { 3650 if (le32_to_cpu(ckpt->cur_data_segno[i]) >= main_segs || 3651 le16_to_cpu(ckpt->cur_data_blkoff[i]) >= blocks_per_seg) 3652 return 1; 3653 3654 if (f2fs_sb_has_readonly(sbi)) 3655 goto skip_cross; 3656 3657 for (j = i + 1; j < NR_CURSEG_DATA_TYPE; j++) { 3658 if (le32_to_cpu(ckpt->cur_data_segno[i]) == 3659 le32_to_cpu(ckpt->cur_data_segno[j])) { 3660 f2fs_err(sbi, "Data segment (%u, %u) has the same segno: %u", 3661 i, j, 3662 le32_to_cpu(ckpt->cur_data_segno[i])); 3663 return 1; 3664 } 3665 } 3666 } 3667 for (i = 0; i < NR_CURSEG_NODE_TYPE; i++) { 3668 for (j = 0; j < NR_CURSEG_DATA_TYPE; j++) { 3669 if (le32_to_cpu(ckpt->cur_node_segno[i]) == 3670 le32_to_cpu(ckpt->cur_data_segno[j])) { 3671 f2fs_err(sbi, "Node segment (%u) and Data segment (%u) has the same segno: %u", 3672 i, j, 3673 le32_to_cpu(ckpt->cur_node_segno[i])); 3674 return 1; 3675 } 3676 } 3677 } 3678 skip_cross: 3679 sit_bitmap_size = le32_to_cpu(ckpt->sit_ver_bitmap_bytesize); 3680 nat_bitmap_size = le32_to_cpu(ckpt->nat_ver_bitmap_bytesize); 3681 3682 if (sit_bitmap_size != ((sit_segs / 2) << log_blocks_per_seg) / 8 || 3683 nat_bitmap_size != ((nat_segs / 2) << log_blocks_per_seg) / 8) { 3684 f2fs_err(sbi, "Wrong bitmap size: sit: %u, nat:%u", 3685 sit_bitmap_size, nat_bitmap_size); 3686 return 1; 3687 } 3688 3689 cp_pack_start_sum = __start_sum_addr(sbi); 3690 cp_payload = __cp_payload(sbi); 3691 if (cp_pack_start_sum < cp_payload + 1 || 3692 cp_pack_start_sum > blocks_per_seg - 1 - 3693 NR_CURSEG_PERSIST_TYPE) { 3694 f2fs_err(sbi, "Wrong cp_pack_start_sum: %u", 3695 cp_pack_start_sum); 3696 return 1; 3697 } 3698 3699 if (__is_set_ckpt_flags(ckpt, CP_LARGE_NAT_BITMAP_FLAG) && 3700 le32_to_cpu(ckpt->checksum_offset) != CP_MIN_CHKSUM_OFFSET) { 3701 f2fs_warn(sbi, "using deprecated layout of large_nat_bitmap, " 3702 "please run fsck v1.13.0 or higher to repair, chksum_offset: %u, " 3703 "fixed with patch: \"f2fs-tools: relocate chksum_offset for large_nat_bitmap feature\"", 3704 le32_to_cpu(ckpt->checksum_offset)); 3705 return 1; 3706 } 3707 3708 nat_blocks = nat_segs << log_blocks_per_seg; 3709 nat_bits_bytes = nat_blocks / BITS_PER_BYTE; 3710 nat_bits_blocks = F2FS_BLK_ALIGN((nat_bits_bytes << 1) + 8); 3711 if (__is_set_ckpt_flags(ckpt, CP_NAT_BITS_FLAG) && 3712 (cp_payload + F2FS_CP_PACKS + 3713 NR_CURSEG_PERSIST_TYPE + nat_bits_blocks >= blocks_per_seg)) { 3714 f2fs_warn(sbi, "Insane cp_payload: %u, nat_bits_blocks: %u)", 3715 cp_payload, nat_bits_blocks); 3716 return 1; 3717 } 3718 3719 if (unlikely(f2fs_cp_error(sbi))) { 3720 f2fs_err(sbi, "A bug case: need to run fsck"); 3721 return 1; 3722 } 3723 return 0; 3724 } 3725 init_sb_info(struct f2fs_sb_info * sbi)3726 static void init_sb_info(struct f2fs_sb_info *sbi) 3727 { 3728 struct f2fs_super_block *raw_super = sbi->raw_super; 3729 int i; 3730 3731 sbi->log_sectors_per_block = 3732 le32_to_cpu(raw_super->log_sectors_per_block); 3733 sbi->log_blocksize = le32_to_cpu(raw_super->log_blocksize); 3734 sbi->blocksize = BIT(sbi->log_blocksize); 3735 sbi->log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg); 3736 sbi->blocks_per_seg = BIT(sbi->log_blocks_per_seg); 3737 sbi->segs_per_sec = le32_to_cpu(raw_super->segs_per_sec); 3738 sbi->secs_per_zone = le32_to_cpu(raw_super->secs_per_zone); 3739 sbi->total_sections = le32_to_cpu(raw_super->section_count); 3740 sbi->total_node_count = 3741 (le32_to_cpu(raw_super->segment_count_nat) / 2) 3742 * sbi->blocks_per_seg * NAT_ENTRY_PER_BLOCK; 3743 F2FS_ROOT_INO(sbi) = le32_to_cpu(raw_super->root_ino); 3744 F2FS_NODE_INO(sbi) = le32_to_cpu(raw_super->node_ino); 3745 F2FS_META_INO(sbi) = le32_to_cpu(raw_super->meta_ino); 3746 sbi->cur_victim_sec = NULL_SECNO; 3747 sbi->gc_mode = GC_NORMAL; 3748 sbi->next_victim_seg[BG_GC] = NULL_SEGNO; 3749 sbi->next_victim_seg[FG_GC] = NULL_SEGNO; 3750 sbi->max_victim_search = DEF_MAX_VICTIM_SEARCH; 3751 sbi->migration_granularity = sbi->segs_per_sec; 3752 sbi->seq_file_ra_mul = MIN_RA_MUL; 3753 sbi->max_fragment_chunk = DEF_FRAGMENT_SIZE; 3754 sbi->max_fragment_hole = DEF_FRAGMENT_SIZE; 3755 spin_lock_init(&sbi->gc_remaining_trials_lock); 3756 atomic64_set(&sbi->current_atomic_write, 0); 3757 3758 sbi->dir_level = DEF_DIR_LEVEL; 3759 sbi->interval_time[CP_TIME] = DEF_CP_INTERVAL; 3760 sbi->interval_time[REQ_TIME] = DEF_IDLE_INTERVAL; 3761 sbi->interval_time[DISCARD_TIME] = DEF_IDLE_INTERVAL; 3762 sbi->interval_time[GC_TIME] = DEF_IDLE_INTERVAL; 3763 sbi->interval_time[DISABLE_TIME] = DEF_DISABLE_INTERVAL; 3764 sbi->interval_time[UMOUNT_DISCARD_TIMEOUT] = 3765 DEF_UMOUNT_DISCARD_TIMEOUT; 3766 clear_sbi_flag(sbi, SBI_NEED_FSCK); 3767 3768 for (i = 0; i < NR_COUNT_TYPE; i++) 3769 atomic_set(&sbi->nr_pages[i], 0); 3770 3771 for (i = 0; i < META; i++) 3772 atomic_set(&sbi->wb_sync_req[i], 0); 3773 3774 INIT_LIST_HEAD(&sbi->s_list); 3775 mutex_init(&sbi->umount_mutex); 3776 init_f2fs_rwsem(&sbi->io_order_lock); 3777 spin_lock_init(&sbi->cp_lock); 3778 3779 sbi->dirty_device = 0; 3780 spin_lock_init(&sbi->dev_lock); 3781 3782 init_f2fs_rwsem(&sbi->sb_lock); 3783 init_f2fs_rwsem(&sbi->pin_sem); 3784 } 3785 init_percpu_info(struct f2fs_sb_info * sbi)3786 static int init_percpu_info(struct f2fs_sb_info *sbi) 3787 { 3788 int err; 3789 3790 err = percpu_counter_init(&sbi->alloc_valid_block_count, 0, GFP_KERNEL); 3791 if (err) 3792 return err; 3793 3794 err = percpu_counter_init(&sbi->rf_node_block_count, 0, GFP_KERNEL); 3795 if (err) 3796 goto err_valid_block; 3797 3798 err = percpu_counter_init(&sbi->total_valid_inode_count, 0, 3799 GFP_KERNEL); 3800 if (err) 3801 goto err_node_block; 3802 return 0; 3803 3804 err_node_block: 3805 percpu_counter_destroy(&sbi->rf_node_block_count); 3806 err_valid_block: 3807 percpu_counter_destroy(&sbi->alloc_valid_block_count); 3808 return err; 3809 } 3810 3811 #ifdef CONFIG_BLK_DEV_ZONED 3812 3813 struct f2fs_report_zones_args { 3814 struct f2fs_sb_info *sbi; 3815 struct f2fs_dev_info *dev; 3816 }; 3817 f2fs_report_zone_cb(struct blk_zone * zone,unsigned int idx,void * data)3818 static int f2fs_report_zone_cb(struct blk_zone *zone, unsigned int idx, 3819 void *data) 3820 { 3821 struct f2fs_report_zones_args *rz_args = data; 3822 block_t unusable_blocks = (zone->len - zone->capacity) >> 3823 F2FS_LOG_SECTORS_PER_BLOCK; 3824 3825 if (zone->type == BLK_ZONE_TYPE_CONVENTIONAL) 3826 return 0; 3827 3828 set_bit(idx, rz_args->dev->blkz_seq); 3829 if (!rz_args->sbi->unusable_blocks_per_sec) { 3830 rz_args->sbi->unusable_blocks_per_sec = unusable_blocks; 3831 return 0; 3832 } 3833 if (rz_args->sbi->unusable_blocks_per_sec != unusable_blocks) { 3834 f2fs_err(rz_args->sbi, "F2FS supports single zone capacity\n"); 3835 return -EINVAL; 3836 } 3837 return 0; 3838 } 3839 init_blkz_info(struct f2fs_sb_info * sbi,int devi)3840 static int init_blkz_info(struct f2fs_sb_info *sbi, int devi) 3841 { 3842 struct block_device *bdev = FDEV(devi).bdev; 3843 sector_t nr_sectors = bdev_nr_sectors(bdev); 3844 struct f2fs_report_zones_args rep_zone_arg; 3845 u64 zone_sectors; 3846 int ret; 3847 3848 if (!f2fs_sb_has_blkzoned(sbi)) 3849 return 0; 3850 3851 zone_sectors = bdev_zone_sectors(bdev); 3852 3853 if (sbi->blocks_per_blkz && sbi->blocks_per_blkz != 3854 SECTOR_TO_BLOCK(zone_sectors)) 3855 return -EINVAL; 3856 sbi->blocks_per_blkz = SECTOR_TO_BLOCK(zone_sectors); 3857 FDEV(devi).nr_blkz = div_u64(SECTOR_TO_BLOCK(nr_sectors), 3858 sbi->blocks_per_blkz); 3859 if (!bdev_is_zone_start(bdev, nr_sectors)) 3860 FDEV(devi).nr_blkz++; 3861 3862 FDEV(devi).blkz_seq = f2fs_kvzalloc(sbi, 3863 BITS_TO_LONGS(FDEV(devi).nr_blkz) 3864 * sizeof(unsigned long), 3865 GFP_KERNEL); 3866 if (!FDEV(devi).blkz_seq) 3867 return -ENOMEM; 3868 3869 rep_zone_arg.sbi = sbi; 3870 rep_zone_arg.dev = &FDEV(devi); 3871 3872 ret = blkdev_report_zones(bdev, 0, BLK_ALL_ZONES, f2fs_report_zone_cb, 3873 &rep_zone_arg); 3874 if (ret < 0) 3875 return ret; 3876 return 0; 3877 } 3878 #endif 3879 3880 /* 3881 * Read f2fs raw super block. 3882 * Because we have two copies of super block, so read both of them 3883 * to get the first valid one. If any one of them is broken, we pass 3884 * them recovery flag back to the caller. 3885 */ read_raw_super_block(struct f2fs_sb_info * sbi,struct f2fs_super_block ** raw_super,int * valid_super_block,int * recovery)3886 static int read_raw_super_block(struct f2fs_sb_info *sbi, 3887 struct f2fs_super_block **raw_super, 3888 int *valid_super_block, int *recovery) 3889 { 3890 struct super_block *sb = sbi->sb; 3891 int block; 3892 struct buffer_head *bh; 3893 struct f2fs_super_block *super; 3894 int err = 0; 3895 3896 super = kzalloc(sizeof(struct f2fs_super_block), GFP_KERNEL); 3897 if (!super) 3898 return -ENOMEM; 3899 3900 for (block = 0; block < 2; block++) { 3901 bh = sb_bread(sb, block); 3902 if (!bh) { 3903 f2fs_err(sbi, "Unable to read %dth superblock", 3904 block + 1); 3905 err = -EIO; 3906 *recovery = 1; 3907 continue; 3908 } 3909 3910 /* sanity checking of raw super */ 3911 err = sanity_check_raw_super(sbi, bh); 3912 if (err) { 3913 f2fs_err(sbi, "Can't find valid F2FS filesystem in %dth superblock", 3914 block + 1); 3915 brelse(bh); 3916 *recovery = 1; 3917 continue; 3918 } 3919 3920 if (!*raw_super) { 3921 memcpy(super, bh->b_data + F2FS_SUPER_OFFSET, 3922 sizeof(*super)); 3923 *valid_super_block = block; 3924 *raw_super = super; 3925 } 3926 brelse(bh); 3927 } 3928 3929 /* No valid superblock */ 3930 if (!*raw_super) 3931 kfree(super); 3932 else 3933 err = 0; 3934 3935 return err; 3936 } 3937 f2fs_commit_super(struct f2fs_sb_info * sbi,bool recover)3938 int f2fs_commit_super(struct f2fs_sb_info *sbi, bool recover) 3939 { 3940 struct buffer_head *bh; 3941 __u32 crc = 0; 3942 int err; 3943 3944 if ((recover && f2fs_readonly(sbi->sb)) || 3945 f2fs_hw_is_readonly(sbi)) { 3946 set_sbi_flag(sbi, SBI_NEED_SB_WRITE); 3947 return -EROFS; 3948 } 3949 3950 /* we should update superblock crc here */ 3951 if (!recover && f2fs_sb_has_sb_chksum(sbi)) { 3952 crc = f2fs_crc32(sbi, F2FS_RAW_SUPER(sbi), 3953 offsetof(struct f2fs_super_block, crc)); 3954 F2FS_RAW_SUPER(sbi)->crc = cpu_to_le32(crc); 3955 } 3956 3957 /* write back-up superblock first */ 3958 bh = sb_bread(sbi->sb, sbi->valid_super_block ? 0 : 1); 3959 if (!bh) 3960 return -EIO; 3961 err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi)); 3962 brelse(bh); 3963 3964 /* if we are in recovery path, skip writing valid superblock */ 3965 if (recover || err) 3966 return err; 3967 3968 /* write current valid superblock */ 3969 bh = sb_bread(sbi->sb, sbi->valid_super_block); 3970 if (!bh) 3971 return -EIO; 3972 err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi)); 3973 brelse(bh); 3974 return err; 3975 } 3976 f2fs_handle_stop(struct f2fs_sb_info * sbi,unsigned char reason)3977 void f2fs_handle_stop(struct f2fs_sb_info *sbi, unsigned char reason) 3978 { 3979 struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi); 3980 int err; 3981 3982 f2fs_down_write(&sbi->sb_lock); 3983 3984 if (raw_super->s_stop_reason[reason] < GENMASK(BITS_PER_BYTE - 1, 0)) 3985 raw_super->s_stop_reason[reason]++; 3986 3987 err = f2fs_commit_super(sbi, false); 3988 if (err) 3989 f2fs_err(sbi, "f2fs_commit_super fails to record reason:%u err:%d", 3990 reason, err); 3991 f2fs_up_write(&sbi->sb_lock); 3992 } 3993 f2fs_save_errors(struct f2fs_sb_info * sbi,unsigned char flag)3994 void f2fs_save_errors(struct f2fs_sb_info *sbi, unsigned char flag) 3995 { 3996 spin_lock(&sbi->error_lock); 3997 if (!test_bit(flag, (unsigned long *)sbi->errors)) { 3998 set_bit(flag, (unsigned long *)sbi->errors); 3999 sbi->error_dirty = true; 4000 } 4001 spin_unlock(&sbi->error_lock); 4002 } 4003 f2fs_update_errors(struct f2fs_sb_info * sbi)4004 static bool f2fs_update_errors(struct f2fs_sb_info *sbi) 4005 { 4006 bool need_update = false; 4007 4008 spin_lock(&sbi->error_lock); 4009 if (sbi->error_dirty) { 4010 memcpy(F2FS_RAW_SUPER(sbi)->s_errors, sbi->errors, 4011 MAX_F2FS_ERRORS); 4012 sbi->error_dirty = false; 4013 need_update = true; 4014 } 4015 spin_unlock(&sbi->error_lock); 4016 4017 return need_update; 4018 } 4019 f2fs_handle_error(struct f2fs_sb_info * sbi,unsigned char error)4020 void f2fs_handle_error(struct f2fs_sb_info *sbi, unsigned char error) 4021 { 4022 int err; 4023 4024 f2fs_save_errors(sbi, error); 4025 4026 f2fs_down_write(&sbi->sb_lock); 4027 4028 if (!f2fs_update_errors(sbi)) 4029 goto out_unlock; 4030 4031 err = f2fs_commit_super(sbi, false); 4032 if (err) 4033 f2fs_err(sbi, "f2fs_commit_super fails to record errors:%u, err:%d", 4034 error, err); 4035 out_unlock: 4036 f2fs_up_write(&sbi->sb_lock); 4037 } 4038 f2fs_scan_devices(struct f2fs_sb_info * sbi)4039 static int f2fs_scan_devices(struct f2fs_sb_info *sbi) 4040 { 4041 struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi); 4042 unsigned int max_devices = MAX_DEVICES; 4043 unsigned int logical_blksize; 4044 int i; 4045 4046 /* Initialize single device information */ 4047 if (!RDEV(0).path[0]) { 4048 if (!bdev_is_zoned(sbi->sb->s_bdev)) 4049 return 0; 4050 max_devices = 1; 4051 } 4052 4053 /* 4054 * Initialize multiple devices information, or single 4055 * zoned block device information. 4056 */ 4057 sbi->devs = f2fs_kzalloc(sbi, 4058 array_size(max_devices, 4059 sizeof(struct f2fs_dev_info)), 4060 GFP_KERNEL); 4061 if (!sbi->devs) 4062 return -ENOMEM; 4063 4064 logical_blksize = bdev_logical_block_size(sbi->sb->s_bdev); 4065 sbi->aligned_blksize = true; 4066 4067 for (i = 0; i < max_devices; i++) { 4068 4069 if (i > 0 && !RDEV(i).path[0]) 4070 break; 4071 4072 if (max_devices == 1) { 4073 /* Single zoned block device mount */ 4074 FDEV(0).bdev = 4075 blkdev_get_by_dev(sbi->sb->s_bdev->bd_dev, 4076 sbi->sb->s_mode, sbi->sb->s_type); 4077 } else { 4078 /* Multi-device mount */ 4079 memcpy(FDEV(i).path, RDEV(i).path, MAX_PATH_LEN); 4080 FDEV(i).total_segments = 4081 le32_to_cpu(RDEV(i).total_segments); 4082 if (i == 0) { 4083 FDEV(i).start_blk = 0; 4084 FDEV(i).end_blk = FDEV(i).start_blk + 4085 (FDEV(i).total_segments << 4086 sbi->log_blocks_per_seg) - 1 + 4087 le32_to_cpu(raw_super->segment0_blkaddr); 4088 } else { 4089 FDEV(i).start_blk = FDEV(i - 1).end_blk + 1; 4090 FDEV(i).end_blk = FDEV(i).start_blk + 4091 (FDEV(i).total_segments << 4092 sbi->log_blocks_per_seg) - 1; 4093 } 4094 FDEV(i).bdev = blkdev_get_by_path(FDEV(i).path, 4095 sbi->sb->s_mode, sbi->sb->s_type); 4096 } 4097 if (IS_ERR(FDEV(i).bdev)) 4098 return PTR_ERR(FDEV(i).bdev); 4099 4100 /* to release errored devices */ 4101 sbi->s_ndevs = i + 1; 4102 4103 if (logical_blksize != bdev_logical_block_size(FDEV(i).bdev)) 4104 sbi->aligned_blksize = false; 4105 4106 #ifdef CONFIG_BLK_DEV_ZONED 4107 if (bdev_zoned_model(FDEV(i).bdev) == BLK_ZONED_HM && 4108 !f2fs_sb_has_blkzoned(sbi)) { 4109 f2fs_err(sbi, "Zoned block device feature not enabled"); 4110 return -EINVAL; 4111 } 4112 if (bdev_zoned_model(FDEV(i).bdev) != BLK_ZONED_NONE) { 4113 if (init_blkz_info(sbi, i)) { 4114 f2fs_err(sbi, "Failed to initialize F2FS blkzone information"); 4115 return -EINVAL; 4116 } 4117 if (max_devices == 1) 4118 break; 4119 f2fs_info(sbi, "Mount Device [%2d]: %20s, %8u, %8x - %8x (zone: %s)", 4120 i, FDEV(i).path, 4121 FDEV(i).total_segments, 4122 FDEV(i).start_blk, FDEV(i).end_blk, 4123 bdev_zoned_model(FDEV(i).bdev) == BLK_ZONED_HA ? 4124 "Host-aware" : "Host-managed"); 4125 continue; 4126 } 4127 #endif 4128 f2fs_info(sbi, "Mount Device [%2d]: %20s, %8u, %8x - %8x", 4129 i, FDEV(i).path, 4130 FDEV(i).total_segments, 4131 FDEV(i).start_blk, FDEV(i).end_blk); 4132 } 4133 f2fs_info(sbi, 4134 "IO Block Size: %8ld KB", F2FS_IO_SIZE_KB(sbi)); 4135 return 0; 4136 } 4137 f2fs_setup_casefold(struct f2fs_sb_info * sbi)4138 static int f2fs_setup_casefold(struct f2fs_sb_info *sbi) 4139 { 4140 #ifdef CONFIG_UNICODE 4141 if (f2fs_sb_has_casefold(sbi) && !sbi->sb->s_encoding) { 4142 const struct f2fs_sb_encodings *encoding_info; 4143 struct unicode_map *encoding; 4144 __u16 encoding_flags; 4145 4146 if (f2fs_sb_read_encoding(sbi->raw_super, &encoding_info, 4147 &encoding_flags)) { 4148 f2fs_err(sbi, 4149 "Encoding requested by superblock is unknown"); 4150 return -EINVAL; 4151 } 4152 4153 encoding = utf8_load(encoding_info->version); 4154 if (IS_ERR(encoding)) { 4155 f2fs_err(sbi, 4156 "can't mount with superblock charset: %s-%s " 4157 "not supported by the kernel. flags: 0x%x.", 4158 encoding_info->name, encoding_info->version, 4159 encoding_flags); 4160 return PTR_ERR(encoding); 4161 } 4162 f2fs_info(sbi, "Using encoding defined by superblock: " 4163 "%s-%s with flags 0x%hx", encoding_info->name, 4164 encoding_info->version?:"\b", encoding_flags); 4165 4166 sbi->sb->s_encoding = encoding; 4167 sbi->sb->s_encoding_flags = encoding_flags; 4168 } 4169 #else 4170 if (f2fs_sb_has_casefold(sbi)) { 4171 f2fs_err(sbi, "Filesystem with casefold feature cannot be mounted without CONFIG_UNICODE"); 4172 return -EINVAL; 4173 } 4174 #endif 4175 return 0; 4176 } 4177 f2fs_tuning_parameters(struct f2fs_sb_info * sbi)4178 static void f2fs_tuning_parameters(struct f2fs_sb_info *sbi) 4179 { 4180 /* adjust parameters according to the volume size */ 4181 if (MAIN_SEGS(sbi) <= SMALL_VOLUME_SEGMENTS) { 4182 if (f2fs_block_unit_discard(sbi)) 4183 SM_I(sbi)->dcc_info->discard_granularity = 4184 MIN_DISCARD_GRANULARITY; 4185 if (!f2fs_lfs_mode(sbi)) 4186 SM_I(sbi)->ipu_policy = BIT(F2FS_IPU_FORCE) | 4187 BIT(F2FS_IPU_HONOR_OPU_WRITE); 4188 } 4189 4190 sbi->readdir_ra = true; 4191 } 4192 f2fs_fill_super(struct super_block * sb,void * data,int silent)4193 static int f2fs_fill_super(struct super_block *sb, void *data, int silent) 4194 { 4195 struct f2fs_sb_info *sbi; 4196 struct f2fs_super_block *raw_super; 4197 struct inode *root; 4198 int err; 4199 bool skip_recovery = false, need_fsck = false; 4200 char *options = NULL; 4201 int recovery, i, valid_super_block; 4202 struct curseg_info *seg_i; 4203 int retry_cnt = 1; 4204 #ifdef CONFIG_QUOTA 4205 bool quota_enabled = false; 4206 #endif 4207 4208 try_onemore: 4209 err = -EINVAL; 4210 raw_super = NULL; 4211 valid_super_block = -1; 4212 recovery = 0; 4213 4214 /* allocate memory for f2fs-specific super block info */ 4215 sbi = kzalloc(sizeof(struct f2fs_sb_info), GFP_KERNEL); 4216 if (!sbi) 4217 return -ENOMEM; 4218 4219 sbi->sb = sb; 4220 4221 /* initialize locks within allocated memory */ 4222 init_f2fs_rwsem(&sbi->gc_lock); 4223 mutex_init(&sbi->writepages); 4224 init_f2fs_rwsem(&sbi->cp_global_sem); 4225 init_f2fs_rwsem(&sbi->node_write); 4226 init_f2fs_rwsem(&sbi->node_change); 4227 spin_lock_init(&sbi->stat_lock); 4228 init_f2fs_rwsem(&sbi->cp_rwsem); 4229 init_f2fs_rwsem(&sbi->quota_sem); 4230 init_waitqueue_head(&sbi->cp_wait); 4231 spin_lock_init(&sbi->error_lock); 4232 4233 for (i = 0; i < NR_INODE_TYPE; i++) { 4234 INIT_LIST_HEAD(&sbi->inode_list[i]); 4235 spin_lock_init(&sbi->inode_lock[i]); 4236 } 4237 mutex_init(&sbi->flush_lock); 4238 4239 /* Load the checksum driver */ 4240 sbi->s_chksum_driver = crypto_alloc_shash("crc32", 0, 0); 4241 if (IS_ERR(sbi->s_chksum_driver)) { 4242 f2fs_err(sbi, "Cannot load crc32 driver."); 4243 err = PTR_ERR(sbi->s_chksum_driver); 4244 sbi->s_chksum_driver = NULL; 4245 goto free_sbi; 4246 } 4247 4248 /* set a block size */ 4249 if (unlikely(!sb_set_blocksize(sb, F2FS_BLKSIZE))) { 4250 f2fs_err(sbi, "unable to set blocksize"); 4251 goto free_sbi; 4252 } 4253 4254 err = read_raw_super_block(sbi, &raw_super, &valid_super_block, 4255 &recovery); 4256 if (err) 4257 goto free_sbi; 4258 4259 sb->s_fs_info = sbi; 4260 sbi->raw_super = raw_super; 4261 4262 memcpy(sbi->errors, raw_super->s_errors, MAX_F2FS_ERRORS); 4263 4264 /* precompute checksum seed for metadata */ 4265 if (f2fs_sb_has_inode_chksum(sbi)) 4266 sbi->s_chksum_seed = f2fs_chksum(sbi, ~0, raw_super->uuid, 4267 sizeof(raw_super->uuid)); 4268 4269 default_options(sbi); 4270 /* parse mount options */ 4271 options = kstrdup((const char *)data, GFP_KERNEL); 4272 if (data && !options) { 4273 err = -ENOMEM; 4274 goto free_sb_buf; 4275 } 4276 4277 err = parse_options(sb, options, false); 4278 if (err) 4279 goto free_options; 4280 4281 sb->s_maxbytes = max_file_blocks(NULL) << 4282 le32_to_cpu(raw_super->log_blocksize); 4283 sb->s_max_links = F2FS_LINK_MAX; 4284 4285 err = f2fs_setup_casefold(sbi); 4286 if (err) 4287 goto free_options; 4288 4289 #ifdef CONFIG_QUOTA 4290 sb->dq_op = &f2fs_quota_operations; 4291 sb->s_qcop = &f2fs_quotactl_ops; 4292 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ; 4293 4294 if (f2fs_sb_has_quota_ino(sbi)) { 4295 for (i = 0; i < MAXQUOTAS; i++) { 4296 if (f2fs_qf_ino(sbi->sb, i)) 4297 sbi->nquota_files++; 4298 } 4299 } 4300 #endif 4301 4302 sb->s_op = &f2fs_sops; 4303 #ifdef CONFIG_FS_ENCRYPTION 4304 sb->s_cop = &f2fs_cryptops; 4305 #endif 4306 #ifdef CONFIG_FS_VERITY 4307 sb->s_vop = &f2fs_verityops; 4308 #endif 4309 sb->s_xattr = f2fs_xattr_handlers; 4310 sb->s_export_op = &f2fs_export_ops; 4311 sb->s_magic = F2FS_SUPER_MAGIC; 4312 sb->s_time_gran = 1; 4313 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) | 4314 (test_opt(sbi, POSIX_ACL) ? SB_POSIXACL : 0); 4315 memcpy(&sb->s_uuid, raw_super->uuid, sizeof(raw_super->uuid)); 4316 sb->s_iflags |= SB_I_CGROUPWB; 4317 4318 /* init f2fs-specific super block info */ 4319 sbi->valid_super_block = valid_super_block; 4320 4321 /* disallow all the data/node/meta page writes */ 4322 set_sbi_flag(sbi, SBI_POR_DOING); 4323 4324 err = f2fs_init_write_merge_io(sbi); 4325 if (err) 4326 goto free_bio_info; 4327 4328 init_sb_info(sbi); 4329 4330 err = f2fs_init_iostat(sbi); 4331 if (err) 4332 goto free_bio_info; 4333 4334 err = init_percpu_info(sbi); 4335 if (err) 4336 goto free_iostat; 4337 4338 if (F2FS_IO_ALIGNED(sbi)) { 4339 sbi->write_io_dummy = 4340 mempool_create_page_pool(2 * (F2FS_IO_SIZE(sbi) - 1), 0); 4341 if (!sbi->write_io_dummy) { 4342 err = -ENOMEM; 4343 goto free_percpu; 4344 } 4345 } 4346 4347 /* init per sbi slab cache */ 4348 err = f2fs_init_xattr_caches(sbi); 4349 if (err) 4350 goto free_io_dummy; 4351 err = f2fs_init_page_array_cache(sbi); 4352 if (err) 4353 goto free_xattr_cache; 4354 4355 /* get an inode for meta space */ 4356 sbi->meta_inode = f2fs_iget(sb, F2FS_META_INO(sbi)); 4357 if (IS_ERR(sbi->meta_inode)) { 4358 f2fs_err(sbi, "Failed to read F2FS meta data inode"); 4359 err = PTR_ERR(sbi->meta_inode); 4360 goto free_page_array_cache; 4361 } 4362 4363 err = f2fs_get_valid_checkpoint(sbi); 4364 if (err) { 4365 f2fs_err(sbi, "Failed to get valid F2FS checkpoint"); 4366 goto free_meta_inode; 4367 } 4368 4369 if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_QUOTA_NEED_FSCK_FLAG)) 4370 set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR); 4371 if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_DISABLED_QUICK_FLAG)) { 4372 set_sbi_flag(sbi, SBI_CP_DISABLED_QUICK); 4373 sbi->interval_time[DISABLE_TIME] = DEF_DISABLE_QUICK_INTERVAL; 4374 } 4375 4376 if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_FSCK_FLAG)) 4377 set_sbi_flag(sbi, SBI_NEED_FSCK); 4378 4379 /* Initialize device list */ 4380 err = f2fs_scan_devices(sbi); 4381 if (err) { 4382 f2fs_err(sbi, "Failed to find devices"); 4383 goto free_devices; 4384 } 4385 4386 err = f2fs_init_post_read_wq(sbi); 4387 if (err) { 4388 f2fs_err(sbi, "Failed to initialize post read workqueue"); 4389 goto free_devices; 4390 } 4391 4392 sbi->total_valid_node_count = 4393 le32_to_cpu(sbi->ckpt->valid_node_count); 4394 percpu_counter_set(&sbi->total_valid_inode_count, 4395 le32_to_cpu(sbi->ckpt->valid_inode_count)); 4396 sbi->user_block_count = le64_to_cpu(sbi->ckpt->user_block_count); 4397 sbi->total_valid_block_count = 4398 le64_to_cpu(sbi->ckpt->valid_block_count); 4399 sbi->last_valid_block_count = sbi->total_valid_block_count; 4400 sbi->reserved_blocks = 0; 4401 sbi->current_reserved_blocks = 0; 4402 limit_reserve_root(sbi); 4403 adjust_unusable_cap_perc(sbi); 4404 4405 f2fs_init_extent_cache_info(sbi); 4406 4407 f2fs_init_ino_entry_info(sbi); 4408 4409 f2fs_init_fsync_node_info(sbi); 4410 4411 /* setup checkpoint request control and start checkpoint issue thread */ 4412 f2fs_init_ckpt_req_control(sbi); 4413 if (!f2fs_readonly(sb) && !test_opt(sbi, DISABLE_CHECKPOINT) && 4414 test_opt(sbi, MERGE_CHECKPOINT)) { 4415 err = f2fs_start_ckpt_thread(sbi); 4416 if (err) { 4417 f2fs_err(sbi, 4418 "Failed to start F2FS issue_checkpoint_thread (%d)", 4419 err); 4420 goto stop_ckpt_thread; 4421 } 4422 } 4423 4424 /* setup f2fs internal modules */ 4425 err = f2fs_build_segment_manager(sbi); 4426 if (err) { 4427 f2fs_err(sbi, "Failed to initialize F2FS segment manager (%d)", 4428 err); 4429 goto free_sm; 4430 } 4431 err = f2fs_build_node_manager(sbi); 4432 if (err) { 4433 f2fs_err(sbi, "Failed to initialize F2FS node manager (%d)", 4434 err); 4435 goto free_nm; 4436 } 4437 4438 err = adjust_reserved_segment(sbi); 4439 if (err) 4440 goto free_nm; 4441 4442 /* For write statistics */ 4443 sbi->sectors_written_start = f2fs_get_sectors_written(sbi); 4444 4445 /* Read accumulated write IO statistics if exists */ 4446 seg_i = CURSEG_I(sbi, CURSEG_HOT_NODE); 4447 if (__exist_node_summaries(sbi)) 4448 sbi->kbytes_written = 4449 le64_to_cpu(seg_i->journal->info.kbytes_written); 4450 4451 f2fs_build_gc_manager(sbi); 4452 4453 err = f2fs_build_stats(sbi); 4454 if (err) 4455 goto free_nm; 4456 4457 /* get an inode for node space */ 4458 sbi->node_inode = f2fs_iget(sb, F2FS_NODE_INO(sbi)); 4459 if (IS_ERR(sbi->node_inode)) { 4460 f2fs_err(sbi, "Failed to read node inode"); 4461 err = PTR_ERR(sbi->node_inode); 4462 goto free_stats; 4463 } 4464 4465 /* read root inode and dentry */ 4466 root = f2fs_iget(sb, F2FS_ROOT_INO(sbi)); 4467 if (IS_ERR(root)) { 4468 f2fs_err(sbi, "Failed to read root inode"); 4469 err = PTR_ERR(root); 4470 goto free_node_inode; 4471 } 4472 if (!S_ISDIR(root->i_mode) || !root->i_blocks || 4473 !root->i_size || !root->i_nlink) { 4474 iput(root); 4475 err = -EINVAL; 4476 goto free_node_inode; 4477 } 4478 4479 sb->s_root = d_make_root(root); /* allocate root dentry */ 4480 if (!sb->s_root) { 4481 err = -ENOMEM; 4482 goto free_node_inode; 4483 } 4484 4485 err = f2fs_init_compress_inode(sbi); 4486 if (err) 4487 goto free_root_inode; 4488 4489 err = f2fs_register_sysfs(sbi); 4490 if (err) 4491 goto free_compress_inode; 4492 4493 #ifdef CONFIG_QUOTA 4494 /* Enable quota usage during mount */ 4495 if (f2fs_sb_has_quota_ino(sbi) && !f2fs_readonly(sb)) { 4496 err = f2fs_enable_quotas(sb); 4497 if (err) 4498 f2fs_err(sbi, "Cannot turn on quotas: error %d", err); 4499 } 4500 4501 quota_enabled = f2fs_recover_quota_begin(sbi); 4502 #endif 4503 /* if there are any orphan inodes, free them */ 4504 err = f2fs_recover_orphan_inodes(sbi); 4505 if (err) 4506 goto free_meta; 4507 4508 if (unlikely(is_set_ckpt_flags(sbi, CP_DISABLED_FLAG))) 4509 goto reset_checkpoint; 4510 4511 /* recover fsynced data */ 4512 if (!test_opt(sbi, DISABLE_ROLL_FORWARD) && 4513 !test_opt(sbi, NORECOVERY)) { 4514 /* 4515 * mount should be failed, when device has readonly mode, and 4516 * previous checkpoint was not done by clean system shutdown. 4517 */ 4518 if (f2fs_hw_is_readonly(sbi)) { 4519 if (!is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG)) { 4520 err = f2fs_recover_fsync_data(sbi, true); 4521 if (err > 0) { 4522 err = -EROFS; 4523 f2fs_err(sbi, "Need to recover fsync data, but " 4524 "write access unavailable, please try " 4525 "mount w/ disable_roll_forward or norecovery"); 4526 } 4527 if (err < 0) 4528 goto free_meta; 4529 } 4530 f2fs_info(sbi, "write access unavailable, skipping recovery"); 4531 goto reset_checkpoint; 4532 } 4533 4534 if (need_fsck) 4535 set_sbi_flag(sbi, SBI_NEED_FSCK); 4536 4537 if (skip_recovery) 4538 goto reset_checkpoint; 4539 4540 err = f2fs_recover_fsync_data(sbi, false); 4541 if (err < 0) { 4542 if (err != -ENOMEM) 4543 skip_recovery = true; 4544 need_fsck = true; 4545 f2fs_err(sbi, "Cannot recover all fsync data errno=%d", 4546 err); 4547 goto free_meta; 4548 } 4549 } else { 4550 err = f2fs_recover_fsync_data(sbi, true); 4551 4552 if (!f2fs_readonly(sb) && err > 0) { 4553 err = -EINVAL; 4554 f2fs_err(sbi, "Need to recover fsync data"); 4555 goto free_meta; 4556 } 4557 } 4558 4559 #ifdef CONFIG_QUOTA 4560 f2fs_recover_quota_end(sbi, quota_enabled); 4561 #endif 4562 4563 /* 4564 * If the f2fs is not readonly and fsync data recovery succeeds, 4565 * check zoned block devices' write pointer consistency. 4566 */ 4567 if (!err && !f2fs_readonly(sb) && f2fs_sb_has_blkzoned(sbi)) { 4568 err = f2fs_check_write_pointer(sbi); 4569 if (err) 4570 goto free_meta; 4571 } 4572 4573 reset_checkpoint: 4574 f2fs_init_inmem_curseg(sbi); 4575 4576 /* f2fs_recover_fsync_data() cleared this already */ 4577 clear_sbi_flag(sbi, SBI_POR_DOING); 4578 4579 if (test_opt(sbi, DISABLE_CHECKPOINT)) { 4580 err = f2fs_disable_checkpoint(sbi); 4581 if (err) 4582 goto sync_free_meta; 4583 } else if (is_set_ckpt_flags(sbi, CP_DISABLED_FLAG)) { 4584 f2fs_enable_checkpoint(sbi); 4585 } 4586 4587 /* 4588 * If filesystem is not mounted as read-only then 4589 * do start the gc_thread. 4590 */ 4591 if ((F2FS_OPTION(sbi).bggc_mode != BGGC_MODE_OFF || 4592 test_opt(sbi, GC_MERGE)) && !f2fs_readonly(sb)) { 4593 /* After POR, we can run background GC thread.*/ 4594 err = f2fs_start_gc_thread(sbi); 4595 if (err) 4596 goto sync_free_meta; 4597 } 4598 kvfree(options); 4599 4600 /* recover broken superblock */ 4601 if (recovery) { 4602 err = f2fs_commit_super(sbi, true); 4603 f2fs_info(sbi, "Try to recover %dth superblock, ret: %d", 4604 sbi->valid_super_block ? 1 : 2, err); 4605 } 4606 4607 f2fs_join_shrinker(sbi); 4608 4609 f2fs_tuning_parameters(sbi); 4610 4611 f2fs_notice(sbi, "Mounted with checkpoint version = %llx", 4612 cur_cp_version(F2FS_CKPT(sbi))); 4613 f2fs_update_time(sbi, CP_TIME); 4614 f2fs_update_time(sbi, REQ_TIME); 4615 clear_sbi_flag(sbi, SBI_CP_DISABLED_QUICK); 4616 4617 cleancache_init_fs(sb); 4618 return 0; 4619 4620 sync_free_meta: 4621 /* safe to flush all the data */ 4622 sync_filesystem(sbi->sb); 4623 retry_cnt = 0; 4624 4625 free_meta: 4626 #ifdef CONFIG_QUOTA 4627 f2fs_truncate_quota_inode_pages(sb); 4628 if (f2fs_sb_has_quota_ino(sbi) && !f2fs_readonly(sb)) 4629 f2fs_quota_off_umount(sbi->sb); 4630 #endif 4631 /* 4632 * Some dirty meta pages can be produced by f2fs_recover_orphan_inodes() 4633 * failed by EIO. Then, iput(node_inode) can trigger balance_fs_bg() 4634 * followed by f2fs_write_checkpoint() through f2fs_write_node_pages(), which 4635 * falls into an infinite loop in f2fs_sync_meta_pages(). 4636 */ 4637 truncate_inode_pages_final(META_MAPPING(sbi)); 4638 /* evict some inodes being cached by GC */ 4639 evict_inodes(sb); 4640 f2fs_unregister_sysfs(sbi); 4641 free_compress_inode: 4642 f2fs_destroy_compress_inode(sbi); 4643 free_root_inode: 4644 dput(sb->s_root); 4645 sb->s_root = NULL; 4646 free_node_inode: 4647 f2fs_release_ino_entry(sbi, true); 4648 truncate_inode_pages_final(NODE_MAPPING(sbi)); 4649 iput(sbi->node_inode); 4650 sbi->node_inode = NULL; 4651 free_stats: 4652 f2fs_destroy_stats(sbi); 4653 free_nm: 4654 /* stop discard thread before destroying node manager */ 4655 f2fs_stop_discard_thread(sbi); 4656 f2fs_destroy_node_manager(sbi); 4657 free_sm: 4658 f2fs_destroy_segment_manager(sbi); 4659 stop_ckpt_thread: 4660 f2fs_stop_ckpt_thread(sbi); 4661 f2fs_destroy_post_read_wq(sbi); 4662 free_devices: 4663 destroy_device_list(sbi); 4664 kvfree(sbi->ckpt); 4665 free_meta_inode: 4666 make_bad_inode(sbi->meta_inode); 4667 iput(sbi->meta_inode); 4668 sbi->meta_inode = NULL; 4669 free_page_array_cache: 4670 f2fs_destroy_page_array_cache(sbi); 4671 free_xattr_cache: 4672 f2fs_destroy_xattr_caches(sbi); 4673 free_io_dummy: 4674 mempool_destroy(sbi->write_io_dummy); 4675 free_percpu: 4676 destroy_percpu_info(sbi); 4677 free_iostat: 4678 f2fs_destroy_iostat(sbi); 4679 free_bio_info: 4680 for (i = 0; i < NR_PAGE_TYPE; i++) 4681 kvfree(sbi->write_io[i]); 4682 4683 #ifdef CONFIG_UNICODE 4684 utf8_unload(sb->s_encoding); 4685 sb->s_encoding = NULL; 4686 #endif 4687 free_options: 4688 #ifdef CONFIG_QUOTA 4689 for (i = 0; i < MAXQUOTAS; i++) 4690 kfree(F2FS_OPTION(sbi).s_qf_names[i]); 4691 #endif 4692 fscrypt_free_dummy_policy(&F2FS_OPTION(sbi).dummy_enc_policy); 4693 kvfree(options); 4694 free_sb_buf: 4695 kfree(raw_super); 4696 free_sbi: 4697 if (sbi->s_chksum_driver) 4698 crypto_free_shash(sbi->s_chksum_driver); 4699 kfree(sbi); 4700 4701 /* give only one another chance */ 4702 if (retry_cnt > 0 && skip_recovery) { 4703 retry_cnt--; 4704 shrink_dcache_sb(sb); 4705 goto try_onemore; 4706 } 4707 return err; 4708 } 4709 f2fs_mount(struct file_system_type * fs_type,int flags,const char * dev_name,void * data)4710 static struct dentry *f2fs_mount(struct file_system_type *fs_type, int flags, 4711 const char *dev_name, void *data) 4712 { 4713 return mount_bdev(fs_type, flags, dev_name, data, f2fs_fill_super); 4714 } 4715 kill_f2fs_super(struct super_block * sb)4716 static void kill_f2fs_super(struct super_block *sb) 4717 { 4718 if (sb->s_root) { 4719 struct f2fs_sb_info *sbi = F2FS_SB(sb); 4720 4721 set_sbi_flag(sbi, SBI_IS_CLOSE); 4722 f2fs_stop_gc_thread(sbi); 4723 f2fs_stop_discard_thread(sbi); 4724 4725 #ifdef CONFIG_F2FS_FS_COMPRESSION 4726 /* 4727 * latter evict_inode() can bypass checking and invalidating 4728 * compress inode cache. 4729 */ 4730 if (test_opt(sbi, COMPRESS_CACHE)) 4731 truncate_inode_pages_final(COMPRESS_MAPPING(sbi)); 4732 #endif 4733 4734 if (is_sbi_flag_set(sbi, SBI_IS_DIRTY) || 4735 !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG)) { 4736 struct cp_control cpc = { 4737 .reason = CP_UMOUNT, 4738 }; 4739 f2fs_write_checkpoint(sbi, &cpc); 4740 } 4741 4742 if (is_sbi_flag_set(sbi, SBI_IS_RECOVERED) && f2fs_readonly(sb)) 4743 sb->s_flags &= ~SB_RDONLY; 4744 } 4745 kill_block_super(sb); 4746 } 4747 4748 static struct file_system_type f2fs_fs_type = { 4749 .owner = THIS_MODULE, 4750 .name = "f2fs", 4751 .mount = f2fs_mount, 4752 .kill_sb = kill_f2fs_super, 4753 .fs_flags = FS_REQUIRES_DEV | FS_ALLOW_IDMAP, 4754 }; 4755 MODULE_ALIAS_FS("f2fs"); 4756 init_inodecache(void)4757 static int __init init_inodecache(void) 4758 { 4759 f2fs_inode_cachep = kmem_cache_create("f2fs_inode_cache", 4760 sizeof(struct f2fs_inode_info), 0, 4761 SLAB_RECLAIM_ACCOUNT|SLAB_ACCOUNT, NULL); 4762 return f2fs_inode_cachep ? 0 : -ENOMEM; 4763 } 4764 destroy_inodecache(void)4765 static void destroy_inodecache(void) 4766 { 4767 /* 4768 * Make sure all delayed rcu free inodes are flushed before we 4769 * destroy cache. 4770 */ 4771 rcu_barrier(); 4772 kmem_cache_destroy(f2fs_inode_cachep); 4773 } 4774 init_f2fs_fs(void)4775 static int __init init_f2fs_fs(void) 4776 { 4777 int err; 4778 4779 if (PAGE_SIZE != F2FS_BLKSIZE) { 4780 printk("F2FS not supported on PAGE_SIZE(%lu) != BLOCK_SIZE(%lu)\n", 4781 PAGE_SIZE, F2FS_BLKSIZE); 4782 return -EINVAL; 4783 } 4784 4785 err = init_inodecache(); 4786 if (err) 4787 goto fail; 4788 err = f2fs_create_node_manager_caches(); 4789 if (err) 4790 goto free_inodecache; 4791 err = f2fs_create_segment_manager_caches(); 4792 if (err) 4793 goto free_node_manager_caches; 4794 err = f2fs_create_checkpoint_caches(); 4795 if (err) 4796 goto free_segment_manager_caches; 4797 err = f2fs_create_recovery_cache(); 4798 if (err) 4799 goto free_checkpoint_caches; 4800 err = f2fs_create_extent_cache(); 4801 if (err) 4802 goto free_recovery_cache; 4803 err = f2fs_create_garbage_collection_cache(); 4804 if (err) 4805 goto free_extent_cache; 4806 err = f2fs_init_sysfs(); 4807 if (err) 4808 goto free_garbage_collection_cache; 4809 err = register_shrinker(&f2fs_shrinker_info); 4810 if (err) 4811 goto free_sysfs; 4812 err = register_filesystem(&f2fs_fs_type); 4813 if (err) 4814 goto free_shrinker; 4815 f2fs_create_root_stats(); 4816 err = f2fs_init_post_read_processing(); 4817 if (err) 4818 goto free_root_stats; 4819 err = f2fs_init_iostat_processing(); 4820 if (err) 4821 goto free_post_read; 4822 err = f2fs_init_bio_entry_cache(); 4823 if (err) 4824 goto free_iostat; 4825 err = f2fs_init_bioset(); 4826 if (err) 4827 goto free_bio_entry_cache; 4828 err = f2fs_init_compress_mempool(); 4829 if (err) 4830 goto free_bioset; 4831 err = f2fs_init_compress_cache(); 4832 if (err) 4833 goto free_compress_mempool; 4834 err = f2fs_create_casefold_cache(); 4835 if (err) 4836 goto free_compress_cache; 4837 return 0; 4838 free_compress_cache: 4839 f2fs_destroy_compress_cache(); 4840 free_compress_mempool: 4841 f2fs_destroy_compress_mempool(); 4842 free_bioset: 4843 f2fs_destroy_bioset(); 4844 free_bio_entry_cache: 4845 f2fs_destroy_bio_entry_cache(); 4846 free_iostat: 4847 f2fs_destroy_iostat_processing(); 4848 free_post_read: 4849 f2fs_destroy_post_read_processing(); 4850 free_root_stats: 4851 f2fs_destroy_root_stats(); 4852 unregister_filesystem(&f2fs_fs_type); 4853 free_shrinker: 4854 unregister_shrinker(&f2fs_shrinker_info); 4855 free_sysfs: 4856 f2fs_exit_sysfs(); 4857 free_garbage_collection_cache: 4858 f2fs_destroy_garbage_collection_cache(); 4859 free_extent_cache: 4860 f2fs_destroy_extent_cache(); 4861 free_recovery_cache: 4862 f2fs_destroy_recovery_cache(); 4863 free_checkpoint_caches: 4864 f2fs_destroy_checkpoint_caches(); 4865 free_segment_manager_caches: 4866 f2fs_destroy_segment_manager_caches(); 4867 free_node_manager_caches: 4868 f2fs_destroy_node_manager_caches(); 4869 free_inodecache: 4870 destroy_inodecache(); 4871 fail: 4872 return err; 4873 } 4874 exit_f2fs_fs(void)4875 static void __exit exit_f2fs_fs(void) 4876 { 4877 f2fs_destroy_casefold_cache(); 4878 f2fs_destroy_compress_cache(); 4879 f2fs_destroy_compress_mempool(); 4880 f2fs_destroy_bioset(); 4881 f2fs_destroy_bio_entry_cache(); 4882 f2fs_destroy_iostat_processing(); 4883 f2fs_destroy_post_read_processing(); 4884 f2fs_destroy_root_stats(); 4885 unregister_filesystem(&f2fs_fs_type); 4886 unregister_shrinker(&f2fs_shrinker_info); 4887 f2fs_exit_sysfs(); 4888 f2fs_destroy_garbage_collection_cache(); 4889 f2fs_destroy_extent_cache(); 4890 f2fs_destroy_recovery_cache(); 4891 f2fs_destroy_checkpoint_caches(); 4892 f2fs_destroy_segment_manager_caches(); 4893 f2fs_destroy_node_manager_caches(); 4894 destroy_inodecache(); 4895 } 4896 4897 module_init(init_f2fs_fs) 4898 module_exit(exit_f2fs_fs) 4899 4900 MODULE_AUTHOR("Samsung Electronics's Praesto Team"); 4901 MODULE_DESCRIPTION("Flash Friendly File System"); 4902 MODULE_LICENSE("GPL"); 4903 MODULE_IMPORT_NS(ANDROID_GKI_VFS_EXPORT_ONLY); 4904 MODULE_SOFTDEP("pre: crc32"); 4905 4906