1 /* 2 * fs/f2fs/super.c 3 * 4 * Copyright (c) 2012 Samsung Electronics Co., Ltd. 5 * http://www.samsung.com/ 6 * 7 * This program is free software; you can redistribute it and/or modify 8 * it under the terms of the GNU General Public License version 2 as 9 * published by the Free Software Foundation. 10 */ 11 #include <linux/module.h> 12 #include <linux/init.h> 13 #include <linux/fs.h> 14 #include <linux/statfs.h> 15 #include <linux/buffer_head.h> 16 #include <linux/backing-dev.h> 17 #include <linux/kthread.h> 18 #include <linux/parser.h> 19 #include <linux/mount.h> 20 #include <linux/seq_file.h> 21 #include <linux/proc_fs.h> 22 #include <linux/random.h> 23 #include <linux/exportfs.h> 24 #include <linux/blkdev.h> 25 #include <linux/quotaops.h> 26 #include <linux/f2fs_fs.h> 27 #include <linux/sysfs.h> 28 #include <linux/quota.h> 29 30 #include "f2fs.h" 31 #include "node.h" 32 #include "segment.h" 33 #include "xattr.h" 34 #include "gc.h" 35 #include "trace.h" 36 37 #define CREATE_TRACE_POINTS 38 #include <trace/events/f2fs.h> 39 40 static struct kmem_cache *f2fs_inode_cachep; 41 42 #ifdef CONFIG_F2FS_FAULT_INJECTION 43 44 char *f2fs_fault_name[FAULT_MAX] = { 45 [FAULT_KMALLOC] = "kmalloc", 46 [FAULT_KVMALLOC] = "kvmalloc", 47 [FAULT_PAGE_ALLOC] = "page alloc", 48 [FAULT_PAGE_GET] = "page get", 49 [FAULT_ALLOC_BIO] = "alloc bio", 50 [FAULT_ALLOC_NID] = "alloc nid", 51 [FAULT_ORPHAN] = "orphan", 52 [FAULT_BLOCK] = "no more block", 53 [FAULT_DIR_DEPTH] = "too big dir depth", 54 [FAULT_EVICT_INODE] = "evict_inode fail", 55 [FAULT_TRUNCATE] = "truncate fail", 56 [FAULT_IO] = "IO error", 57 [FAULT_CHECKPOINT] = "checkpoint error", 58 [FAULT_DISCARD] = "discard error", 59 }; 60 61 void f2fs_build_fault_attr(struct f2fs_sb_info *sbi, unsigned int rate, 62 unsigned int type) 63 { 64 struct f2fs_fault_info *ffi = &F2FS_OPTION(sbi).fault_info; 65 66 if (rate) { 67 atomic_set(&ffi->inject_ops, 0); 68 ffi->inject_rate = rate; 69 } 70 71 if (type) 72 ffi->inject_type = type; 73 74 if (!rate && !type) 75 memset(ffi, 0, sizeof(struct f2fs_fault_info)); 76 } 77 #endif 78 79 /* f2fs-wide shrinker description */ 80 static struct shrinker f2fs_shrinker_info = { 81 .scan_objects = f2fs_shrink_scan, 82 .count_objects = f2fs_shrink_count, 83 .seeks = DEFAULT_SEEKS, 84 }; 85 86 enum { 87 Opt_gc_background, 88 Opt_disable_roll_forward, 89 Opt_norecovery, 90 Opt_discard, 91 Opt_nodiscard, 92 Opt_noheap, 93 Opt_heap, 94 Opt_user_xattr, 95 Opt_nouser_xattr, 96 Opt_acl, 97 Opt_noacl, 98 Opt_active_logs, 99 Opt_disable_ext_identify, 100 Opt_inline_xattr, 101 Opt_noinline_xattr, 102 Opt_inline_xattr_size, 103 Opt_inline_data, 104 Opt_inline_dentry, 105 Opt_noinline_dentry, 106 Opt_flush_merge, 107 Opt_noflush_merge, 108 Opt_nobarrier, 109 Opt_fastboot, 110 Opt_extent_cache, 111 Opt_noextent_cache, 112 Opt_noinline_data, 113 Opt_data_flush, 114 Opt_reserve_root, 115 Opt_resgid, 116 Opt_resuid, 117 Opt_mode, 118 Opt_io_size_bits, 119 Opt_fault_injection, 120 Opt_fault_type, 121 Opt_lazytime, 122 Opt_nolazytime, 123 Opt_quota, 124 Opt_noquota, 125 Opt_usrquota, 126 Opt_grpquota, 127 Opt_prjquota, 128 Opt_usrjquota, 129 Opt_grpjquota, 130 Opt_prjjquota, 131 Opt_offusrjquota, 132 Opt_offgrpjquota, 133 Opt_offprjjquota, 134 Opt_jqfmt_vfsold, 135 Opt_jqfmt_vfsv0, 136 Opt_jqfmt_vfsv1, 137 Opt_whint, 138 Opt_alloc, 139 Opt_fsync, 140 Opt_test_dummy_encryption, 141 Opt_err, 142 }; 143 144 static match_table_t f2fs_tokens = { 145 {Opt_gc_background, "background_gc=%s"}, 146 {Opt_disable_roll_forward, "disable_roll_forward"}, 147 {Opt_norecovery, "norecovery"}, 148 {Opt_discard, "discard"}, 149 {Opt_nodiscard, "nodiscard"}, 150 {Opt_noheap, "no_heap"}, 151 {Opt_heap, "heap"}, 152 {Opt_user_xattr, "user_xattr"}, 153 {Opt_nouser_xattr, "nouser_xattr"}, 154 {Opt_acl, "acl"}, 155 {Opt_noacl, "noacl"}, 156 {Opt_active_logs, "active_logs=%u"}, 157 {Opt_disable_ext_identify, "disable_ext_identify"}, 158 {Opt_inline_xattr, "inline_xattr"}, 159 {Opt_noinline_xattr, "noinline_xattr"}, 160 {Opt_inline_xattr_size, "inline_xattr_size=%u"}, 161 {Opt_inline_data, "inline_data"}, 162 {Opt_inline_dentry, "inline_dentry"}, 163 {Opt_noinline_dentry, "noinline_dentry"}, 164 {Opt_flush_merge, "flush_merge"}, 165 {Opt_noflush_merge, "noflush_merge"}, 166 {Opt_nobarrier, "nobarrier"}, 167 {Opt_fastboot, "fastboot"}, 168 {Opt_extent_cache, "extent_cache"}, 169 {Opt_noextent_cache, "noextent_cache"}, 170 {Opt_noinline_data, "noinline_data"}, 171 {Opt_data_flush, "data_flush"}, 172 {Opt_reserve_root, "reserve_root=%u"}, 173 {Opt_resgid, "resgid=%u"}, 174 {Opt_resuid, "resuid=%u"}, 175 {Opt_mode, "mode=%s"}, 176 {Opt_io_size_bits, "io_bits=%u"}, 177 {Opt_fault_injection, "fault_injection=%u"}, 178 {Opt_fault_type, "fault_type=%u"}, 179 {Opt_lazytime, "lazytime"}, 180 {Opt_nolazytime, "nolazytime"}, 181 {Opt_quota, "quota"}, 182 {Opt_noquota, "noquota"}, 183 {Opt_usrquota, "usrquota"}, 184 {Opt_grpquota, "grpquota"}, 185 {Opt_prjquota, "prjquota"}, 186 {Opt_usrjquota, "usrjquota=%s"}, 187 {Opt_grpjquota, "grpjquota=%s"}, 188 {Opt_prjjquota, "prjjquota=%s"}, 189 {Opt_offusrjquota, "usrjquota="}, 190 {Opt_offgrpjquota, "grpjquota="}, 191 {Opt_offprjjquota, "prjjquota="}, 192 {Opt_jqfmt_vfsold, "jqfmt=vfsold"}, 193 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"}, 194 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"}, 195 {Opt_whint, "whint_mode=%s"}, 196 {Opt_alloc, "alloc_mode=%s"}, 197 {Opt_fsync, "fsync_mode=%s"}, 198 {Opt_test_dummy_encryption, "test_dummy_encryption"}, 199 {Opt_err, NULL}, 200 }; 201 202 void f2fs_msg(struct super_block *sb, const char *level, const char *fmt, ...) 203 { 204 struct va_format vaf; 205 va_list args; 206 207 va_start(args, fmt); 208 vaf.fmt = fmt; 209 vaf.va = &args; 210 printk_ratelimited("%sF2FS-fs (%s): %pV\n", level, sb->s_id, &vaf); 211 va_end(args); 212 } 213 214 static inline void limit_reserve_root(struct f2fs_sb_info *sbi) 215 { 216 block_t limit = (sbi->user_block_count << 1) / 1000; 217 218 /* limit is 0.2% */ 219 if (test_opt(sbi, RESERVE_ROOT) && 220 F2FS_OPTION(sbi).root_reserved_blocks > limit) { 221 F2FS_OPTION(sbi).root_reserved_blocks = limit; 222 f2fs_msg(sbi->sb, KERN_INFO, 223 "Reduce reserved blocks for root = %u", 224 F2FS_OPTION(sbi).root_reserved_blocks); 225 } 226 if (!test_opt(sbi, RESERVE_ROOT) && 227 (!uid_eq(F2FS_OPTION(sbi).s_resuid, 228 make_kuid(&init_user_ns, F2FS_DEF_RESUID)) || 229 !gid_eq(F2FS_OPTION(sbi).s_resgid, 230 make_kgid(&init_user_ns, F2FS_DEF_RESGID)))) 231 f2fs_msg(sbi->sb, KERN_INFO, 232 "Ignore s_resuid=%u, s_resgid=%u w/o reserve_root", 233 from_kuid_munged(&init_user_ns, 234 F2FS_OPTION(sbi).s_resuid), 235 from_kgid_munged(&init_user_ns, 236 F2FS_OPTION(sbi).s_resgid)); 237 } 238 239 static void init_once(void *foo) 240 { 241 struct f2fs_inode_info *fi = (struct f2fs_inode_info *) foo; 242 243 inode_init_once(&fi->vfs_inode); 244 } 245 246 #ifdef CONFIG_QUOTA 247 static const char * const quotatypes[] = INITQFNAMES; 248 #define QTYPE2NAME(t) (quotatypes[t]) 249 static int f2fs_set_qf_name(struct super_block *sb, int qtype, 250 substring_t *args) 251 { 252 struct f2fs_sb_info *sbi = F2FS_SB(sb); 253 char *qname; 254 int ret = -EINVAL; 255 256 if (sb_any_quota_loaded(sb) && !F2FS_OPTION(sbi).s_qf_names[qtype]) { 257 f2fs_msg(sb, KERN_ERR, 258 "Cannot change journaled " 259 "quota options when quota turned on"); 260 return -EINVAL; 261 } 262 if (f2fs_sb_has_quota_ino(sb)) { 263 f2fs_msg(sb, KERN_INFO, 264 "QUOTA feature is enabled, so ignore qf_name"); 265 return 0; 266 } 267 268 qname = match_strdup(args); 269 if (!qname) { 270 f2fs_msg(sb, KERN_ERR, 271 "Not enough memory for storing quotafile name"); 272 return -EINVAL; 273 } 274 if (F2FS_OPTION(sbi).s_qf_names[qtype]) { 275 if (strcmp(F2FS_OPTION(sbi).s_qf_names[qtype], qname) == 0) 276 ret = 0; 277 else 278 f2fs_msg(sb, KERN_ERR, 279 "%s quota file already specified", 280 QTYPE2NAME(qtype)); 281 goto errout; 282 } 283 if (strchr(qname, '/')) { 284 f2fs_msg(sb, KERN_ERR, 285 "quotafile must be on filesystem root"); 286 goto errout; 287 } 288 F2FS_OPTION(sbi).s_qf_names[qtype] = qname; 289 set_opt(sbi, QUOTA); 290 return 0; 291 errout: 292 kfree(qname); 293 return ret; 294 } 295 296 static int f2fs_clear_qf_name(struct super_block *sb, int qtype) 297 { 298 struct f2fs_sb_info *sbi = F2FS_SB(sb); 299 300 if (sb_any_quota_loaded(sb) && F2FS_OPTION(sbi).s_qf_names[qtype]) { 301 f2fs_msg(sb, KERN_ERR, "Cannot change journaled quota options" 302 " when quota turned on"); 303 return -EINVAL; 304 } 305 kfree(F2FS_OPTION(sbi).s_qf_names[qtype]); 306 F2FS_OPTION(sbi).s_qf_names[qtype] = NULL; 307 return 0; 308 } 309 310 static int f2fs_check_quota_options(struct f2fs_sb_info *sbi) 311 { 312 /* 313 * We do the test below only for project quotas. 'usrquota' and 314 * 'grpquota' mount options are allowed even without quota feature 315 * to support legacy quotas in quota files. 316 */ 317 if (test_opt(sbi, PRJQUOTA) && !f2fs_sb_has_project_quota(sbi->sb)) { 318 f2fs_msg(sbi->sb, KERN_ERR, "Project quota feature not enabled. " 319 "Cannot enable project quota enforcement."); 320 return -1; 321 } 322 if (F2FS_OPTION(sbi).s_qf_names[USRQUOTA] || 323 F2FS_OPTION(sbi).s_qf_names[GRPQUOTA] || 324 F2FS_OPTION(sbi).s_qf_names[PRJQUOTA]) { 325 if (test_opt(sbi, USRQUOTA) && 326 F2FS_OPTION(sbi).s_qf_names[USRQUOTA]) 327 clear_opt(sbi, USRQUOTA); 328 329 if (test_opt(sbi, GRPQUOTA) && 330 F2FS_OPTION(sbi).s_qf_names[GRPQUOTA]) 331 clear_opt(sbi, GRPQUOTA); 332 333 if (test_opt(sbi, PRJQUOTA) && 334 F2FS_OPTION(sbi).s_qf_names[PRJQUOTA]) 335 clear_opt(sbi, PRJQUOTA); 336 337 if (test_opt(sbi, GRPQUOTA) || test_opt(sbi, USRQUOTA) || 338 test_opt(sbi, PRJQUOTA)) { 339 f2fs_msg(sbi->sb, KERN_ERR, "old and new quota " 340 "format mixing"); 341 return -1; 342 } 343 344 if (!F2FS_OPTION(sbi).s_jquota_fmt) { 345 f2fs_msg(sbi->sb, KERN_ERR, "journaled quota format " 346 "not specified"); 347 return -1; 348 } 349 } 350 351 if (f2fs_sb_has_quota_ino(sbi->sb) && F2FS_OPTION(sbi).s_jquota_fmt) { 352 f2fs_msg(sbi->sb, KERN_INFO, 353 "QUOTA feature is enabled, so ignore jquota_fmt"); 354 F2FS_OPTION(sbi).s_jquota_fmt = 0; 355 } 356 return 0; 357 } 358 #endif 359 360 static int parse_options(struct super_block *sb, char *options) 361 { 362 struct f2fs_sb_info *sbi = F2FS_SB(sb); 363 struct request_queue *q; 364 substring_t args[MAX_OPT_ARGS]; 365 char *p, *name; 366 int arg = 0; 367 kuid_t uid; 368 kgid_t gid; 369 #ifdef CONFIG_QUOTA 370 int ret; 371 #endif 372 373 if (!options) 374 return 0; 375 376 while ((p = strsep(&options, ",")) != NULL) { 377 int token; 378 if (!*p) 379 continue; 380 /* 381 * Initialize args struct so we know whether arg was 382 * found; some options take optional arguments. 383 */ 384 args[0].to = args[0].from = NULL; 385 token = match_token(p, f2fs_tokens, args); 386 387 switch (token) { 388 case Opt_gc_background: 389 name = match_strdup(&args[0]); 390 391 if (!name) 392 return -ENOMEM; 393 if (strlen(name) == 2 && !strncmp(name, "on", 2)) { 394 set_opt(sbi, BG_GC); 395 clear_opt(sbi, FORCE_FG_GC); 396 } else if (strlen(name) == 3 && !strncmp(name, "off", 3)) { 397 clear_opt(sbi, BG_GC); 398 clear_opt(sbi, FORCE_FG_GC); 399 } else if (strlen(name) == 4 && !strncmp(name, "sync", 4)) { 400 set_opt(sbi, BG_GC); 401 set_opt(sbi, FORCE_FG_GC); 402 } else { 403 kfree(name); 404 return -EINVAL; 405 } 406 kfree(name); 407 break; 408 case Opt_disable_roll_forward: 409 set_opt(sbi, DISABLE_ROLL_FORWARD); 410 break; 411 case Opt_norecovery: 412 /* this option mounts f2fs with ro */ 413 set_opt(sbi, DISABLE_ROLL_FORWARD); 414 if (!f2fs_readonly(sb)) 415 return -EINVAL; 416 break; 417 case Opt_discard: 418 q = bdev_get_queue(sb->s_bdev); 419 if (blk_queue_discard(q)) { 420 set_opt(sbi, DISCARD); 421 } else if (!f2fs_sb_has_blkzoned(sb)) { 422 f2fs_msg(sb, KERN_WARNING, 423 "mounting with \"discard\" option, but " 424 "the device does not support discard"); 425 } 426 break; 427 case Opt_nodiscard: 428 if (f2fs_sb_has_blkzoned(sb)) { 429 f2fs_msg(sb, KERN_WARNING, 430 "discard is required for zoned block devices"); 431 return -EINVAL; 432 } 433 clear_opt(sbi, DISCARD); 434 break; 435 case Opt_noheap: 436 set_opt(sbi, NOHEAP); 437 break; 438 case Opt_heap: 439 clear_opt(sbi, NOHEAP); 440 break; 441 #ifdef CONFIG_F2FS_FS_XATTR 442 case Opt_user_xattr: 443 set_opt(sbi, XATTR_USER); 444 break; 445 case Opt_nouser_xattr: 446 clear_opt(sbi, XATTR_USER); 447 break; 448 case Opt_inline_xattr: 449 set_opt(sbi, INLINE_XATTR); 450 break; 451 case Opt_noinline_xattr: 452 clear_opt(sbi, INLINE_XATTR); 453 break; 454 case Opt_inline_xattr_size: 455 if (args->from && match_int(args, &arg)) 456 return -EINVAL; 457 set_opt(sbi, INLINE_XATTR_SIZE); 458 F2FS_OPTION(sbi).inline_xattr_size = arg; 459 break; 460 #else 461 case Opt_user_xattr: 462 f2fs_msg(sb, KERN_INFO, 463 "user_xattr options not supported"); 464 break; 465 case Opt_nouser_xattr: 466 f2fs_msg(sb, KERN_INFO, 467 "nouser_xattr options not supported"); 468 break; 469 case Opt_inline_xattr: 470 f2fs_msg(sb, KERN_INFO, 471 "inline_xattr options not supported"); 472 break; 473 case Opt_noinline_xattr: 474 f2fs_msg(sb, KERN_INFO, 475 "noinline_xattr options not supported"); 476 break; 477 #endif 478 #ifdef CONFIG_F2FS_FS_POSIX_ACL 479 case Opt_acl: 480 set_opt(sbi, POSIX_ACL); 481 break; 482 case Opt_noacl: 483 clear_opt(sbi, POSIX_ACL); 484 break; 485 #else 486 case Opt_acl: 487 f2fs_msg(sb, KERN_INFO, "acl options not supported"); 488 break; 489 case Opt_noacl: 490 f2fs_msg(sb, KERN_INFO, "noacl options not supported"); 491 break; 492 #endif 493 case Opt_active_logs: 494 if (args->from && match_int(args, &arg)) 495 return -EINVAL; 496 if (arg != 2 && arg != 4 && arg != NR_CURSEG_TYPE) 497 return -EINVAL; 498 F2FS_OPTION(sbi).active_logs = arg; 499 break; 500 case Opt_disable_ext_identify: 501 set_opt(sbi, DISABLE_EXT_IDENTIFY); 502 break; 503 case Opt_inline_data: 504 set_opt(sbi, INLINE_DATA); 505 break; 506 case Opt_inline_dentry: 507 set_opt(sbi, INLINE_DENTRY); 508 break; 509 case Opt_noinline_dentry: 510 clear_opt(sbi, INLINE_DENTRY); 511 break; 512 case Opt_flush_merge: 513 set_opt(sbi, FLUSH_MERGE); 514 break; 515 case Opt_noflush_merge: 516 clear_opt(sbi, FLUSH_MERGE); 517 break; 518 case Opt_nobarrier: 519 set_opt(sbi, NOBARRIER); 520 break; 521 case Opt_fastboot: 522 set_opt(sbi, FASTBOOT); 523 break; 524 case Opt_extent_cache: 525 set_opt(sbi, EXTENT_CACHE); 526 break; 527 case Opt_noextent_cache: 528 clear_opt(sbi, EXTENT_CACHE); 529 break; 530 case Opt_noinline_data: 531 clear_opt(sbi, INLINE_DATA); 532 break; 533 case Opt_data_flush: 534 set_opt(sbi, DATA_FLUSH); 535 break; 536 case Opt_reserve_root: 537 if (args->from && match_int(args, &arg)) 538 return -EINVAL; 539 if (test_opt(sbi, RESERVE_ROOT)) { 540 f2fs_msg(sb, KERN_INFO, 541 "Preserve previous reserve_root=%u", 542 F2FS_OPTION(sbi).root_reserved_blocks); 543 } else { 544 F2FS_OPTION(sbi).root_reserved_blocks = arg; 545 set_opt(sbi, RESERVE_ROOT); 546 } 547 break; 548 case Opt_resuid: 549 if (args->from && match_int(args, &arg)) 550 return -EINVAL; 551 uid = make_kuid(current_user_ns(), arg); 552 if (!uid_valid(uid)) { 553 f2fs_msg(sb, KERN_ERR, 554 "Invalid uid value %d", arg); 555 return -EINVAL; 556 } 557 F2FS_OPTION(sbi).s_resuid = uid; 558 break; 559 case Opt_resgid: 560 if (args->from && match_int(args, &arg)) 561 return -EINVAL; 562 gid = make_kgid(current_user_ns(), arg); 563 if (!gid_valid(gid)) { 564 f2fs_msg(sb, KERN_ERR, 565 "Invalid gid value %d", arg); 566 return -EINVAL; 567 } 568 F2FS_OPTION(sbi).s_resgid = gid; 569 break; 570 case Opt_mode: 571 name = match_strdup(&args[0]); 572 573 if (!name) 574 return -ENOMEM; 575 if (strlen(name) == 8 && 576 !strncmp(name, "adaptive", 8)) { 577 if (f2fs_sb_has_blkzoned(sb)) { 578 f2fs_msg(sb, KERN_WARNING, 579 "adaptive mode is not allowed with " 580 "zoned block device feature"); 581 kfree(name); 582 return -EINVAL; 583 } 584 set_opt_mode(sbi, F2FS_MOUNT_ADAPTIVE); 585 } else if (strlen(name) == 3 && 586 !strncmp(name, "lfs", 3)) { 587 set_opt_mode(sbi, F2FS_MOUNT_LFS); 588 } else { 589 kfree(name); 590 return -EINVAL; 591 } 592 kfree(name); 593 break; 594 case Opt_io_size_bits: 595 if (args->from && match_int(args, &arg)) 596 return -EINVAL; 597 if (arg > __ilog2_u32(BIO_MAX_PAGES)) { 598 f2fs_msg(sb, KERN_WARNING, 599 "Not support %d, larger than %d", 600 1 << arg, BIO_MAX_PAGES); 601 return -EINVAL; 602 } 603 F2FS_OPTION(sbi).write_io_size_bits = arg; 604 break; 605 case Opt_fault_injection: 606 if (args->from && match_int(args, &arg)) 607 return -EINVAL; 608 #ifdef CONFIG_F2FS_FAULT_INJECTION 609 f2fs_build_fault_attr(sbi, arg, F2FS_ALL_FAULT_TYPE); 610 set_opt(sbi, FAULT_INJECTION); 611 #else 612 f2fs_msg(sb, KERN_INFO, 613 "FAULT_INJECTION was not selected"); 614 #endif 615 break; 616 case Opt_fault_type: 617 if (args->from && match_int(args, &arg)) 618 return -EINVAL; 619 #ifdef CONFIG_F2FS_FAULT_INJECTION 620 f2fs_build_fault_attr(sbi, 0, arg); 621 set_opt(sbi, FAULT_INJECTION); 622 #else 623 f2fs_msg(sb, KERN_INFO, 624 "FAULT_INJECTION was not selected"); 625 #endif 626 break; 627 case Opt_lazytime: 628 sb->s_flags |= SB_LAZYTIME; 629 break; 630 case Opt_nolazytime: 631 sb->s_flags &= ~SB_LAZYTIME; 632 break; 633 #ifdef CONFIG_QUOTA 634 case Opt_quota: 635 case Opt_usrquota: 636 set_opt(sbi, USRQUOTA); 637 break; 638 case Opt_grpquota: 639 set_opt(sbi, GRPQUOTA); 640 break; 641 case Opt_prjquota: 642 set_opt(sbi, PRJQUOTA); 643 break; 644 case Opt_usrjquota: 645 ret = f2fs_set_qf_name(sb, USRQUOTA, &args[0]); 646 if (ret) 647 return ret; 648 break; 649 case Opt_grpjquota: 650 ret = f2fs_set_qf_name(sb, GRPQUOTA, &args[0]); 651 if (ret) 652 return ret; 653 break; 654 case Opt_prjjquota: 655 ret = f2fs_set_qf_name(sb, PRJQUOTA, &args[0]); 656 if (ret) 657 return ret; 658 break; 659 case Opt_offusrjquota: 660 ret = f2fs_clear_qf_name(sb, USRQUOTA); 661 if (ret) 662 return ret; 663 break; 664 case Opt_offgrpjquota: 665 ret = f2fs_clear_qf_name(sb, GRPQUOTA); 666 if (ret) 667 return ret; 668 break; 669 case Opt_offprjjquota: 670 ret = f2fs_clear_qf_name(sb, PRJQUOTA); 671 if (ret) 672 return ret; 673 break; 674 case Opt_jqfmt_vfsold: 675 F2FS_OPTION(sbi).s_jquota_fmt = QFMT_VFS_OLD; 676 break; 677 case Opt_jqfmt_vfsv0: 678 F2FS_OPTION(sbi).s_jquota_fmt = QFMT_VFS_V0; 679 break; 680 case Opt_jqfmt_vfsv1: 681 F2FS_OPTION(sbi).s_jquota_fmt = QFMT_VFS_V1; 682 break; 683 case Opt_noquota: 684 clear_opt(sbi, QUOTA); 685 clear_opt(sbi, USRQUOTA); 686 clear_opt(sbi, GRPQUOTA); 687 clear_opt(sbi, PRJQUOTA); 688 break; 689 #else 690 case Opt_quota: 691 case Opt_usrquota: 692 case Opt_grpquota: 693 case Opt_prjquota: 694 case Opt_usrjquota: 695 case Opt_grpjquota: 696 case Opt_prjjquota: 697 case Opt_offusrjquota: 698 case Opt_offgrpjquota: 699 case Opt_offprjjquota: 700 case Opt_jqfmt_vfsold: 701 case Opt_jqfmt_vfsv0: 702 case Opt_jqfmt_vfsv1: 703 case Opt_noquota: 704 f2fs_msg(sb, KERN_INFO, 705 "quota operations not supported"); 706 break; 707 #endif 708 case Opt_whint: 709 name = match_strdup(&args[0]); 710 if (!name) 711 return -ENOMEM; 712 if (strlen(name) == 10 && 713 !strncmp(name, "user-based", 10)) { 714 F2FS_OPTION(sbi).whint_mode = WHINT_MODE_USER; 715 } else if (strlen(name) == 3 && 716 !strncmp(name, "off", 3)) { 717 F2FS_OPTION(sbi).whint_mode = WHINT_MODE_OFF; 718 } else if (strlen(name) == 8 && 719 !strncmp(name, "fs-based", 8)) { 720 F2FS_OPTION(sbi).whint_mode = WHINT_MODE_FS; 721 } else { 722 kfree(name); 723 return -EINVAL; 724 } 725 kfree(name); 726 break; 727 case Opt_alloc: 728 name = match_strdup(&args[0]); 729 if (!name) 730 return -ENOMEM; 731 732 if (strlen(name) == 7 && 733 !strncmp(name, "default", 7)) { 734 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_DEFAULT; 735 } else if (strlen(name) == 5 && 736 !strncmp(name, "reuse", 5)) { 737 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_REUSE; 738 } else { 739 kfree(name); 740 return -EINVAL; 741 } 742 kfree(name); 743 break; 744 case Opt_fsync: 745 name = match_strdup(&args[0]); 746 if (!name) 747 return -ENOMEM; 748 if (strlen(name) == 5 && 749 !strncmp(name, "posix", 5)) { 750 F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_POSIX; 751 } else if (strlen(name) == 6 && 752 !strncmp(name, "strict", 6)) { 753 F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_STRICT; 754 } else if (strlen(name) == 9 && 755 !strncmp(name, "nobarrier", 9)) { 756 F2FS_OPTION(sbi).fsync_mode = 757 FSYNC_MODE_NOBARRIER; 758 } else { 759 kfree(name); 760 return -EINVAL; 761 } 762 kfree(name); 763 break; 764 case Opt_test_dummy_encryption: 765 #ifdef CONFIG_F2FS_FS_ENCRYPTION 766 if (!f2fs_sb_has_encrypt(sb)) { 767 f2fs_msg(sb, KERN_ERR, "Encrypt feature is off"); 768 return -EINVAL; 769 } 770 771 F2FS_OPTION(sbi).test_dummy_encryption = true; 772 f2fs_msg(sb, KERN_INFO, 773 "Test dummy encryption mode enabled"); 774 #else 775 f2fs_msg(sb, KERN_INFO, 776 "Test dummy encryption mount option ignored"); 777 #endif 778 break; 779 default: 780 f2fs_msg(sb, KERN_ERR, 781 "Unrecognized mount option \"%s\" or missing value", 782 p); 783 return -EINVAL; 784 } 785 } 786 #ifdef CONFIG_QUOTA 787 if (f2fs_check_quota_options(sbi)) 788 return -EINVAL; 789 #else 790 if (f2fs_sb_has_quota_ino(sbi->sb) && !f2fs_readonly(sbi->sb)) { 791 f2fs_msg(sbi->sb, KERN_INFO, 792 "Filesystem with quota feature cannot be mounted RDWR " 793 "without CONFIG_QUOTA"); 794 return -EINVAL; 795 } 796 if (f2fs_sb_has_project_quota(sbi->sb) && !f2fs_readonly(sbi->sb)) { 797 f2fs_msg(sb, KERN_ERR, 798 "Filesystem with project quota feature cannot be " 799 "mounted RDWR without CONFIG_QUOTA"); 800 return -EINVAL; 801 } 802 #endif 803 804 if (F2FS_IO_SIZE_BITS(sbi) && !test_opt(sbi, LFS)) { 805 f2fs_msg(sb, KERN_ERR, 806 "Should set mode=lfs with %uKB-sized IO", 807 F2FS_IO_SIZE_KB(sbi)); 808 return -EINVAL; 809 } 810 811 if (test_opt(sbi, INLINE_XATTR_SIZE)) { 812 if (!f2fs_sb_has_extra_attr(sb) || 813 !f2fs_sb_has_flexible_inline_xattr(sb)) { 814 f2fs_msg(sb, KERN_ERR, 815 "extra_attr or flexible_inline_xattr " 816 "feature is off"); 817 return -EINVAL; 818 } 819 if (!test_opt(sbi, INLINE_XATTR)) { 820 f2fs_msg(sb, KERN_ERR, 821 "inline_xattr_size option should be " 822 "set with inline_xattr option"); 823 return -EINVAL; 824 } 825 if (!F2FS_OPTION(sbi).inline_xattr_size || 826 F2FS_OPTION(sbi).inline_xattr_size >= 827 DEF_ADDRS_PER_INODE - 828 F2FS_TOTAL_EXTRA_ATTR_SIZE - 829 DEF_INLINE_RESERVED_SIZE - 830 DEF_MIN_INLINE_SIZE) { 831 f2fs_msg(sb, KERN_ERR, 832 "inline xattr size is out of range"); 833 return -EINVAL; 834 } 835 } 836 837 /* Not pass down write hints if the number of active logs is lesser 838 * than NR_CURSEG_TYPE. 839 */ 840 if (F2FS_OPTION(sbi).active_logs != NR_CURSEG_TYPE) 841 F2FS_OPTION(sbi).whint_mode = WHINT_MODE_OFF; 842 return 0; 843 } 844 845 static struct inode *f2fs_alloc_inode(struct super_block *sb) 846 { 847 struct f2fs_inode_info *fi; 848 849 fi = kmem_cache_alloc(f2fs_inode_cachep, GFP_F2FS_ZERO); 850 if (!fi) 851 return NULL; 852 853 init_once((void *) fi); 854 855 /* Initialize f2fs-specific inode info */ 856 atomic_set(&fi->dirty_pages, 0); 857 init_rwsem(&fi->i_sem); 858 INIT_LIST_HEAD(&fi->dirty_list); 859 INIT_LIST_HEAD(&fi->gdirty_list); 860 INIT_LIST_HEAD(&fi->inmem_ilist); 861 INIT_LIST_HEAD(&fi->inmem_pages); 862 mutex_init(&fi->inmem_lock); 863 init_rwsem(&fi->i_gc_rwsem[READ]); 864 init_rwsem(&fi->i_gc_rwsem[WRITE]); 865 init_rwsem(&fi->i_mmap_sem); 866 init_rwsem(&fi->i_xattr_sem); 867 868 /* Will be used by directory only */ 869 fi->i_dir_level = F2FS_SB(sb)->dir_level; 870 871 return &fi->vfs_inode; 872 } 873 874 static int f2fs_drop_inode(struct inode *inode) 875 { 876 int ret; 877 /* 878 * This is to avoid a deadlock condition like below. 879 * writeback_single_inode(inode) 880 * - f2fs_write_data_page 881 * - f2fs_gc -> iput -> evict 882 * - inode_wait_for_writeback(inode) 883 */ 884 if ((!inode_unhashed(inode) && inode->i_state & I_SYNC)) { 885 if (!inode->i_nlink && !is_bad_inode(inode)) { 886 /* to avoid evict_inode call simultaneously */ 887 atomic_inc(&inode->i_count); 888 spin_unlock(&inode->i_lock); 889 890 /* some remained atomic pages should discarded */ 891 if (f2fs_is_atomic_file(inode)) 892 f2fs_drop_inmem_pages(inode); 893 894 /* should remain fi->extent_tree for writepage */ 895 f2fs_destroy_extent_node(inode); 896 897 sb_start_intwrite(inode->i_sb); 898 f2fs_i_size_write(inode, 0); 899 900 if (F2FS_HAS_BLOCKS(inode)) 901 f2fs_truncate(inode); 902 903 sb_end_intwrite(inode->i_sb); 904 905 spin_lock(&inode->i_lock); 906 atomic_dec(&inode->i_count); 907 } 908 trace_f2fs_drop_inode(inode, 0); 909 return 0; 910 } 911 ret = generic_drop_inode(inode); 912 trace_f2fs_drop_inode(inode, ret); 913 return ret; 914 } 915 916 int f2fs_inode_dirtied(struct inode *inode, bool sync) 917 { 918 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 919 int ret = 0; 920 921 spin_lock(&sbi->inode_lock[DIRTY_META]); 922 if (is_inode_flag_set(inode, FI_DIRTY_INODE)) { 923 ret = 1; 924 } else { 925 set_inode_flag(inode, FI_DIRTY_INODE); 926 stat_inc_dirty_inode(sbi, DIRTY_META); 927 } 928 if (sync && list_empty(&F2FS_I(inode)->gdirty_list)) { 929 list_add_tail(&F2FS_I(inode)->gdirty_list, 930 &sbi->inode_list[DIRTY_META]); 931 inc_page_count(sbi, F2FS_DIRTY_IMETA); 932 } 933 spin_unlock(&sbi->inode_lock[DIRTY_META]); 934 return ret; 935 } 936 937 void f2fs_inode_synced(struct inode *inode) 938 { 939 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 940 941 spin_lock(&sbi->inode_lock[DIRTY_META]); 942 if (!is_inode_flag_set(inode, FI_DIRTY_INODE)) { 943 spin_unlock(&sbi->inode_lock[DIRTY_META]); 944 return; 945 } 946 if (!list_empty(&F2FS_I(inode)->gdirty_list)) { 947 list_del_init(&F2FS_I(inode)->gdirty_list); 948 dec_page_count(sbi, F2FS_DIRTY_IMETA); 949 } 950 clear_inode_flag(inode, FI_DIRTY_INODE); 951 clear_inode_flag(inode, FI_AUTO_RECOVER); 952 stat_dec_dirty_inode(F2FS_I_SB(inode), DIRTY_META); 953 spin_unlock(&sbi->inode_lock[DIRTY_META]); 954 } 955 956 /* 957 * f2fs_dirty_inode() is called from __mark_inode_dirty() 958 * 959 * We should call set_dirty_inode to write the dirty inode through write_inode. 960 */ 961 static void f2fs_dirty_inode(struct inode *inode, int flags) 962 { 963 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 964 965 if (inode->i_ino == F2FS_NODE_INO(sbi) || 966 inode->i_ino == F2FS_META_INO(sbi)) 967 return; 968 969 if (flags == I_DIRTY_TIME) 970 return; 971 972 if (is_inode_flag_set(inode, FI_AUTO_RECOVER)) 973 clear_inode_flag(inode, FI_AUTO_RECOVER); 974 975 f2fs_inode_dirtied(inode, false); 976 } 977 978 static void f2fs_i_callback(struct rcu_head *head) 979 { 980 struct inode *inode = container_of(head, struct inode, i_rcu); 981 kmem_cache_free(f2fs_inode_cachep, F2FS_I(inode)); 982 } 983 984 static void f2fs_destroy_inode(struct inode *inode) 985 { 986 call_rcu(&inode->i_rcu, f2fs_i_callback); 987 } 988 989 static void destroy_percpu_info(struct f2fs_sb_info *sbi) 990 { 991 percpu_counter_destroy(&sbi->alloc_valid_block_count); 992 percpu_counter_destroy(&sbi->total_valid_inode_count); 993 } 994 995 static void destroy_device_list(struct f2fs_sb_info *sbi) 996 { 997 int i; 998 999 for (i = 0; i < sbi->s_ndevs; i++) { 1000 blkdev_put(FDEV(i).bdev, FMODE_EXCL); 1001 #ifdef CONFIG_BLK_DEV_ZONED 1002 kfree(FDEV(i).blkz_type); 1003 #endif 1004 } 1005 kfree(sbi->devs); 1006 } 1007 1008 static void f2fs_put_super(struct super_block *sb) 1009 { 1010 struct f2fs_sb_info *sbi = F2FS_SB(sb); 1011 int i; 1012 bool dropped; 1013 1014 f2fs_quota_off_umount(sb); 1015 1016 /* prevent remaining shrinker jobs */ 1017 mutex_lock(&sbi->umount_mutex); 1018 1019 /* 1020 * We don't need to do checkpoint when superblock is clean. 1021 * But, the previous checkpoint was not done by umount, it needs to do 1022 * clean checkpoint again. 1023 */ 1024 if (is_sbi_flag_set(sbi, SBI_IS_DIRTY) || 1025 !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG)) { 1026 struct cp_control cpc = { 1027 .reason = CP_UMOUNT, 1028 }; 1029 f2fs_write_checkpoint(sbi, &cpc); 1030 } 1031 1032 /* be sure to wait for any on-going discard commands */ 1033 dropped = f2fs_wait_discard_bios(sbi); 1034 1035 if (f2fs_discard_en(sbi) && !sbi->discard_blks && !dropped) { 1036 struct cp_control cpc = { 1037 .reason = CP_UMOUNT | CP_TRIMMED, 1038 }; 1039 f2fs_write_checkpoint(sbi, &cpc); 1040 } 1041 1042 /* f2fs_write_checkpoint can update stat informaion */ 1043 f2fs_destroy_stats(sbi); 1044 1045 /* 1046 * normally superblock is clean, so we need to release this. 1047 * In addition, EIO will skip do checkpoint, we need this as well. 1048 */ 1049 f2fs_release_ino_entry(sbi, true); 1050 1051 f2fs_leave_shrinker(sbi); 1052 mutex_unlock(&sbi->umount_mutex); 1053 1054 /* our cp_error case, we can wait for any writeback page */ 1055 f2fs_flush_merged_writes(sbi); 1056 1057 f2fs_wait_on_all_pages_writeback(sbi); 1058 1059 f2fs_bug_on(sbi, sbi->fsync_node_num); 1060 1061 iput(sbi->node_inode); 1062 iput(sbi->meta_inode); 1063 1064 /* destroy f2fs internal modules */ 1065 f2fs_destroy_node_manager(sbi); 1066 f2fs_destroy_segment_manager(sbi); 1067 1068 kfree(sbi->ckpt); 1069 1070 f2fs_unregister_sysfs(sbi); 1071 1072 sb->s_fs_info = NULL; 1073 if (sbi->s_chksum_driver) 1074 crypto_free_shash(sbi->s_chksum_driver); 1075 kfree(sbi->raw_super); 1076 1077 destroy_device_list(sbi); 1078 mempool_destroy(sbi->write_io_dummy); 1079 #ifdef CONFIG_QUOTA 1080 for (i = 0; i < MAXQUOTAS; i++) 1081 kfree(F2FS_OPTION(sbi).s_qf_names[i]); 1082 #endif 1083 destroy_percpu_info(sbi); 1084 for (i = 0; i < NR_PAGE_TYPE; i++) 1085 kfree(sbi->write_io[i]); 1086 kfree(sbi); 1087 } 1088 1089 int f2fs_sync_fs(struct super_block *sb, int sync) 1090 { 1091 struct f2fs_sb_info *sbi = F2FS_SB(sb); 1092 int err = 0; 1093 1094 if (unlikely(f2fs_cp_error(sbi))) 1095 return 0; 1096 1097 trace_f2fs_sync_fs(sb, sync); 1098 1099 if (unlikely(is_sbi_flag_set(sbi, SBI_POR_DOING))) 1100 return -EAGAIN; 1101 1102 if (sync) { 1103 struct cp_control cpc; 1104 1105 cpc.reason = __get_cp_reason(sbi); 1106 1107 mutex_lock(&sbi->gc_mutex); 1108 err = f2fs_write_checkpoint(sbi, &cpc); 1109 mutex_unlock(&sbi->gc_mutex); 1110 } 1111 f2fs_trace_ios(NULL, 1); 1112 1113 return err; 1114 } 1115 1116 static int f2fs_freeze(struct super_block *sb) 1117 { 1118 if (f2fs_readonly(sb)) 1119 return 0; 1120 1121 /* IO error happened before */ 1122 if (unlikely(f2fs_cp_error(F2FS_SB(sb)))) 1123 return -EIO; 1124 1125 /* must be clean, since sync_filesystem() was already called */ 1126 if (is_sbi_flag_set(F2FS_SB(sb), SBI_IS_DIRTY)) 1127 return -EINVAL; 1128 return 0; 1129 } 1130 1131 static int f2fs_unfreeze(struct super_block *sb) 1132 { 1133 return 0; 1134 } 1135 1136 #ifdef CONFIG_QUOTA 1137 static int f2fs_statfs_project(struct super_block *sb, 1138 kprojid_t projid, struct kstatfs *buf) 1139 { 1140 struct kqid qid; 1141 struct dquot *dquot; 1142 u64 limit; 1143 u64 curblock; 1144 1145 qid = make_kqid_projid(projid); 1146 dquot = dqget(sb, qid); 1147 if (IS_ERR(dquot)) 1148 return PTR_ERR(dquot); 1149 spin_lock(&dquot->dq_dqb_lock); 1150 1151 limit = (dquot->dq_dqb.dqb_bsoftlimit ? 1152 dquot->dq_dqb.dqb_bsoftlimit : 1153 dquot->dq_dqb.dqb_bhardlimit) >> sb->s_blocksize_bits; 1154 if (limit && buf->f_blocks > limit) { 1155 curblock = dquot->dq_dqb.dqb_curspace >> sb->s_blocksize_bits; 1156 buf->f_blocks = limit; 1157 buf->f_bfree = buf->f_bavail = 1158 (buf->f_blocks > curblock) ? 1159 (buf->f_blocks - curblock) : 0; 1160 } 1161 1162 limit = dquot->dq_dqb.dqb_isoftlimit ? 1163 dquot->dq_dqb.dqb_isoftlimit : 1164 dquot->dq_dqb.dqb_ihardlimit; 1165 if (limit && buf->f_files > limit) { 1166 buf->f_files = limit; 1167 buf->f_ffree = 1168 (buf->f_files > dquot->dq_dqb.dqb_curinodes) ? 1169 (buf->f_files - dquot->dq_dqb.dqb_curinodes) : 0; 1170 } 1171 1172 spin_unlock(&dquot->dq_dqb_lock); 1173 dqput(dquot); 1174 return 0; 1175 } 1176 #endif 1177 1178 static int f2fs_statfs(struct dentry *dentry, struct kstatfs *buf) 1179 { 1180 struct super_block *sb = dentry->d_sb; 1181 struct f2fs_sb_info *sbi = F2FS_SB(sb); 1182 u64 id = huge_encode_dev(sb->s_bdev->bd_dev); 1183 block_t total_count, user_block_count, start_count; 1184 u64 avail_node_count; 1185 1186 total_count = le64_to_cpu(sbi->raw_super->block_count); 1187 user_block_count = sbi->user_block_count; 1188 start_count = le32_to_cpu(sbi->raw_super->segment0_blkaddr); 1189 buf->f_type = F2FS_SUPER_MAGIC; 1190 buf->f_bsize = sbi->blocksize; 1191 1192 buf->f_blocks = total_count - start_count; 1193 buf->f_bfree = user_block_count - valid_user_blocks(sbi) - 1194 sbi->current_reserved_blocks; 1195 if (buf->f_bfree > F2FS_OPTION(sbi).root_reserved_blocks) 1196 buf->f_bavail = buf->f_bfree - 1197 F2FS_OPTION(sbi).root_reserved_blocks; 1198 else 1199 buf->f_bavail = 0; 1200 1201 avail_node_count = sbi->total_node_count - sbi->nquota_files - 1202 F2FS_RESERVED_NODE_NUM; 1203 1204 if (avail_node_count > user_block_count) { 1205 buf->f_files = user_block_count; 1206 buf->f_ffree = buf->f_bavail; 1207 } else { 1208 buf->f_files = avail_node_count; 1209 buf->f_ffree = min(avail_node_count - valid_node_count(sbi), 1210 buf->f_bavail); 1211 } 1212 1213 buf->f_namelen = F2FS_NAME_LEN; 1214 buf->f_fsid.val[0] = (u32)id; 1215 buf->f_fsid.val[1] = (u32)(id >> 32); 1216 1217 #ifdef CONFIG_QUOTA 1218 if (is_inode_flag_set(dentry->d_inode, FI_PROJ_INHERIT) && 1219 sb_has_quota_limits_enabled(sb, PRJQUOTA)) { 1220 f2fs_statfs_project(sb, F2FS_I(dentry->d_inode)->i_projid, buf); 1221 } 1222 #endif 1223 return 0; 1224 } 1225 1226 static inline void f2fs_show_quota_options(struct seq_file *seq, 1227 struct super_block *sb) 1228 { 1229 #ifdef CONFIG_QUOTA 1230 struct f2fs_sb_info *sbi = F2FS_SB(sb); 1231 1232 if (F2FS_OPTION(sbi).s_jquota_fmt) { 1233 char *fmtname = ""; 1234 1235 switch (F2FS_OPTION(sbi).s_jquota_fmt) { 1236 case QFMT_VFS_OLD: 1237 fmtname = "vfsold"; 1238 break; 1239 case QFMT_VFS_V0: 1240 fmtname = "vfsv0"; 1241 break; 1242 case QFMT_VFS_V1: 1243 fmtname = "vfsv1"; 1244 break; 1245 } 1246 seq_printf(seq, ",jqfmt=%s", fmtname); 1247 } 1248 1249 if (F2FS_OPTION(sbi).s_qf_names[USRQUOTA]) 1250 seq_show_option(seq, "usrjquota", 1251 F2FS_OPTION(sbi).s_qf_names[USRQUOTA]); 1252 1253 if (F2FS_OPTION(sbi).s_qf_names[GRPQUOTA]) 1254 seq_show_option(seq, "grpjquota", 1255 F2FS_OPTION(sbi).s_qf_names[GRPQUOTA]); 1256 1257 if (F2FS_OPTION(sbi).s_qf_names[PRJQUOTA]) 1258 seq_show_option(seq, "prjjquota", 1259 F2FS_OPTION(sbi).s_qf_names[PRJQUOTA]); 1260 #endif 1261 } 1262 1263 static int f2fs_show_options(struct seq_file *seq, struct dentry *root) 1264 { 1265 struct f2fs_sb_info *sbi = F2FS_SB(root->d_sb); 1266 1267 if (!f2fs_readonly(sbi->sb) && test_opt(sbi, BG_GC)) { 1268 if (test_opt(sbi, FORCE_FG_GC)) 1269 seq_printf(seq, ",background_gc=%s", "sync"); 1270 else 1271 seq_printf(seq, ",background_gc=%s", "on"); 1272 } else { 1273 seq_printf(seq, ",background_gc=%s", "off"); 1274 } 1275 if (test_opt(sbi, DISABLE_ROLL_FORWARD)) 1276 seq_puts(seq, ",disable_roll_forward"); 1277 if (test_opt(sbi, DISCARD)) 1278 seq_puts(seq, ",discard"); 1279 if (test_opt(sbi, NOHEAP)) 1280 seq_puts(seq, ",no_heap"); 1281 else 1282 seq_puts(seq, ",heap"); 1283 #ifdef CONFIG_F2FS_FS_XATTR 1284 if (test_opt(sbi, XATTR_USER)) 1285 seq_puts(seq, ",user_xattr"); 1286 else 1287 seq_puts(seq, ",nouser_xattr"); 1288 if (test_opt(sbi, INLINE_XATTR)) 1289 seq_puts(seq, ",inline_xattr"); 1290 else 1291 seq_puts(seq, ",noinline_xattr"); 1292 if (test_opt(sbi, INLINE_XATTR_SIZE)) 1293 seq_printf(seq, ",inline_xattr_size=%u", 1294 F2FS_OPTION(sbi).inline_xattr_size); 1295 #endif 1296 #ifdef CONFIG_F2FS_FS_POSIX_ACL 1297 if (test_opt(sbi, POSIX_ACL)) 1298 seq_puts(seq, ",acl"); 1299 else 1300 seq_puts(seq, ",noacl"); 1301 #endif 1302 if (test_opt(sbi, DISABLE_EXT_IDENTIFY)) 1303 seq_puts(seq, ",disable_ext_identify"); 1304 if (test_opt(sbi, INLINE_DATA)) 1305 seq_puts(seq, ",inline_data"); 1306 else 1307 seq_puts(seq, ",noinline_data"); 1308 if (test_opt(sbi, INLINE_DENTRY)) 1309 seq_puts(seq, ",inline_dentry"); 1310 else 1311 seq_puts(seq, ",noinline_dentry"); 1312 if (!f2fs_readonly(sbi->sb) && test_opt(sbi, FLUSH_MERGE)) 1313 seq_puts(seq, ",flush_merge"); 1314 if (test_opt(sbi, NOBARRIER)) 1315 seq_puts(seq, ",nobarrier"); 1316 if (test_opt(sbi, FASTBOOT)) 1317 seq_puts(seq, ",fastboot"); 1318 if (test_opt(sbi, EXTENT_CACHE)) 1319 seq_puts(seq, ",extent_cache"); 1320 else 1321 seq_puts(seq, ",noextent_cache"); 1322 if (test_opt(sbi, DATA_FLUSH)) 1323 seq_puts(seq, ",data_flush"); 1324 1325 seq_puts(seq, ",mode="); 1326 if (test_opt(sbi, ADAPTIVE)) 1327 seq_puts(seq, "adaptive"); 1328 else if (test_opt(sbi, LFS)) 1329 seq_puts(seq, "lfs"); 1330 seq_printf(seq, ",active_logs=%u", F2FS_OPTION(sbi).active_logs); 1331 if (test_opt(sbi, RESERVE_ROOT)) 1332 seq_printf(seq, ",reserve_root=%u,resuid=%u,resgid=%u", 1333 F2FS_OPTION(sbi).root_reserved_blocks, 1334 from_kuid_munged(&init_user_ns, 1335 F2FS_OPTION(sbi).s_resuid), 1336 from_kgid_munged(&init_user_ns, 1337 F2FS_OPTION(sbi).s_resgid)); 1338 if (F2FS_IO_SIZE_BITS(sbi)) 1339 seq_printf(seq, ",io_size=%uKB", F2FS_IO_SIZE_KB(sbi)); 1340 #ifdef CONFIG_F2FS_FAULT_INJECTION 1341 if (test_opt(sbi, FAULT_INJECTION)) { 1342 seq_printf(seq, ",fault_injection=%u", 1343 F2FS_OPTION(sbi).fault_info.inject_rate); 1344 seq_printf(seq, ",fault_type=%u", 1345 F2FS_OPTION(sbi).fault_info.inject_type); 1346 } 1347 #endif 1348 #ifdef CONFIG_QUOTA 1349 if (test_opt(sbi, QUOTA)) 1350 seq_puts(seq, ",quota"); 1351 if (test_opt(sbi, USRQUOTA)) 1352 seq_puts(seq, ",usrquota"); 1353 if (test_opt(sbi, GRPQUOTA)) 1354 seq_puts(seq, ",grpquota"); 1355 if (test_opt(sbi, PRJQUOTA)) 1356 seq_puts(seq, ",prjquota"); 1357 #endif 1358 f2fs_show_quota_options(seq, sbi->sb); 1359 if (F2FS_OPTION(sbi).whint_mode == WHINT_MODE_USER) 1360 seq_printf(seq, ",whint_mode=%s", "user-based"); 1361 else if (F2FS_OPTION(sbi).whint_mode == WHINT_MODE_FS) 1362 seq_printf(seq, ",whint_mode=%s", "fs-based"); 1363 #ifdef CONFIG_F2FS_FS_ENCRYPTION 1364 if (F2FS_OPTION(sbi).test_dummy_encryption) 1365 seq_puts(seq, ",test_dummy_encryption"); 1366 #endif 1367 1368 if (F2FS_OPTION(sbi).alloc_mode == ALLOC_MODE_DEFAULT) 1369 seq_printf(seq, ",alloc_mode=%s", "default"); 1370 else if (F2FS_OPTION(sbi).alloc_mode == ALLOC_MODE_REUSE) 1371 seq_printf(seq, ",alloc_mode=%s", "reuse"); 1372 1373 if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_POSIX) 1374 seq_printf(seq, ",fsync_mode=%s", "posix"); 1375 else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_STRICT) 1376 seq_printf(seq, ",fsync_mode=%s", "strict"); 1377 else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_NOBARRIER) 1378 seq_printf(seq, ",fsync_mode=%s", "nobarrier"); 1379 return 0; 1380 } 1381 1382 static void default_options(struct f2fs_sb_info *sbi) 1383 { 1384 /* init some FS parameters */ 1385 F2FS_OPTION(sbi).active_logs = NR_CURSEG_TYPE; 1386 F2FS_OPTION(sbi).inline_xattr_size = DEFAULT_INLINE_XATTR_ADDRS; 1387 F2FS_OPTION(sbi).whint_mode = WHINT_MODE_OFF; 1388 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_DEFAULT; 1389 F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_POSIX; 1390 F2FS_OPTION(sbi).test_dummy_encryption = false; 1391 F2FS_OPTION(sbi).s_resuid = make_kuid(&init_user_ns, F2FS_DEF_RESUID); 1392 F2FS_OPTION(sbi).s_resgid = make_kgid(&init_user_ns, F2FS_DEF_RESGID); 1393 1394 set_opt(sbi, BG_GC); 1395 set_opt(sbi, INLINE_XATTR); 1396 set_opt(sbi, INLINE_DATA); 1397 set_opt(sbi, INLINE_DENTRY); 1398 set_opt(sbi, EXTENT_CACHE); 1399 set_opt(sbi, NOHEAP); 1400 sbi->sb->s_flags |= SB_LAZYTIME; 1401 set_opt(sbi, FLUSH_MERGE); 1402 if (blk_queue_discard(bdev_get_queue(sbi->sb->s_bdev))) 1403 set_opt(sbi, DISCARD); 1404 if (f2fs_sb_has_blkzoned(sbi->sb)) 1405 set_opt_mode(sbi, F2FS_MOUNT_LFS); 1406 else 1407 set_opt_mode(sbi, F2FS_MOUNT_ADAPTIVE); 1408 1409 #ifdef CONFIG_F2FS_FS_XATTR 1410 set_opt(sbi, XATTR_USER); 1411 #endif 1412 #ifdef CONFIG_F2FS_FS_POSIX_ACL 1413 set_opt(sbi, POSIX_ACL); 1414 #endif 1415 1416 f2fs_build_fault_attr(sbi, 0, 0); 1417 } 1418 1419 #ifdef CONFIG_QUOTA 1420 static int f2fs_enable_quotas(struct super_block *sb); 1421 #endif 1422 static int f2fs_remount(struct super_block *sb, int *flags, char *data) 1423 { 1424 struct f2fs_sb_info *sbi = F2FS_SB(sb); 1425 struct f2fs_mount_info org_mount_opt; 1426 unsigned long old_sb_flags; 1427 int err; 1428 bool need_restart_gc = false; 1429 bool need_stop_gc = false; 1430 bool no_extent_cache = !test_opt(sbi, EXTENT_CACHE); 1431 #ifdef CONFIG_QUOTA 1432 int i, j; 1433 #endif 1434 1435 /* 1436 * Save the old mount options in case we 1437 * need to restore them. 1438 */ 1439 org_mount_opt = sbi->mount_opt; 1440 old_sb_flags = sb->s_flags; 1441 1442 #ifdef CONFIG_QUOTA 1443 org_mount_opt.s_jquota_fmt = F2FS_OPTION(sbi).s_jquota_fmt; 1444 for (i = 0; i < MAXQUOTAS; i++) { 1445 if (F2FS_OPTION(sbi).s_qf_names[i]) { 1446 org_mount_opt.s_qf_names[i] = 1447 kstrdup(F2FS_OPTION(sbi).s_qf_names[i], 1448 GFP_KERNEL); 1449 if (!org_mount_opt.s_qf_names[i]) { 1450 for (j = 0; j < i; j++) 1451 kfree(org_mount_opt.s_qf_names[j]); 1452 return -ENOMEM; 1453 } 1454 } else { 1455 org_mount_opt.s_qf_names[i] = NULL; 1456 } 1457 } 1458 #endif 1459 1460 /* recover superblocks we couldn't write due to previous RO mount */ 1461 if (!(*flags & SB_RDONLY) && is_sbi_flag_set(sbi, SBI_NEED_SB_WRITE)) { 1462 err = f2fs_commit_super(sbi, false); 1463 f2fs_msg(sb, KERN_INFO, 1464 "Try to recover all the superblocks, ret: %d", err); 1465 if (!err) 1466 clear_sbi_flag(sbi, SBI_NEED_SB_WRITE); 1467 } 1468 1469 default_options(sbi); 1470 1471 /* parse mount options */ 1472 err = parse_options(sb, data); 1473 if (err) 1474 goto restore_opts; 1475 1476 /* 1477 * Previous and new state of filesystem is RO, 1478 * so skip checking GC and FLUSH_MERGE conditions. 1479 */ 1480 if (f2fs_readonly(sb) && (*flags & SB_RDONLY)) 1481 goto skip; 1482 1483 #ifdef CONFIG_QUOTA 1484 if (!f2fs_readonly(sb) && (*flags & SB_RDONLY)) { 1485 err = dquot_suspend(sb, -1); 1486 if (err < 0) 1487 goto restore_opts; 1488 } else if (f2fs_readonly(sb) && !(*flags & MS_RDONLY)) { 1489 /* dquot_resume needs RW */ 1490 sb->s_flags &= ~SB_RDONLY; 1491 if (sb_any_quota_suspended(sb)) { 1492 dquot_resume(sb, -1); 1493 } else if (f2fs_sb_has_quota_ino(sb)) { 1494 err = f2fs_enable_quotas(sb); 1495 if (err) 1496 goto restore_opts; 1497 } 1498 } 1499 #endif 1500 /* disallow enable/disable extent_cache dynamically */ 1501 if (no_extent_cache == !!test_opt(sbi, EXTENT_CACHE)) { 1502 err = -EINVAL; 1503 f2fs_msg(sbi->sb, KERN_WARNING, 1504 "switch extent_cache option is not allowed"); 1505 goto restore_opts; 1506 } 1507 1508 /* 1509 * We stop the GC thread if FS is mounted as RO 1510 * or if background_gc = off is passed in mount 1511 * option. Also sync the filesystem. 1512 */ 1513 if ((*flags & SB_RDONLY) || !test_opt(sbi, BG_GC)) { 1514 if (sbi->gc_thread) { 1515 f2fs_stop_gc_thread(sbi); 1516 need_restart_gc = true; 1517 } 1518 } else if (!sbi->gc_thread) { 1519 err = f2fs_start_gc_thread(sbi); 1520 if (err) 1521 goto restore_opts; 1522 need_stop_gc = true; 1523 } 1524 1525 if (*flags & SB_RDONLY || 1526 F2FS_OPTION(sbi).whint_mode != org_mount_opt.whint_mode) { 1527 writeback_inodes_sb(sb, WB_REASON_SYNC); 1528 sync_inodes_sb(sb); 1529 1530 set_sbi_flag(sbi, SBI_IS_DIRTY); 1531 set_sbi_flag(sbi, SBI_IS_CLOSE); 1532 f2fs_sync_fs(sb, 1); 1533 clear_sbi_flag(sbi, SBI_IS_CLOSE); 1534 } 1535 1536 /* 1537 * We stop issue flush thread if FS is mounted as RO 1538 * or if flush_merge is not passed in mount option. 1539 */ 1540 if ((*flags & SB_RDONLY) || !test_opt(sbi, FLUSH_MERGE)) { 1541 clear_opt(sbi, FLUSH_MERGE); 1542 f2fs_destroy_flush_cmd_control(sbi, false); 1543 } else { 1544 err = f2fs_create_flush_cmd_control(sbi); 1545 if (err) 1546 goto restore_gc; 1547 } 1548 skip: 1549 #ifdef CONFIG_QUOTA 1550 /* Release old quota file names */ 1551 for (i = 0; i < MAXQUOTAS; i++) 1552 kfree(org_mount_opt.s_qf_names[i]); 1553 #endif 1554 /* Update the POSIXACL Flag */ 1555 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) | 1556 (test_opt(sbi, POSIX_ACL) ? SB_POSIXACL : 0); 1557 1558 limit_reserve_root(sbi); 1559 return 0; 1560 restore_gc: 1561 if (need_restart_gc) { 1562 if (f2fs_start_gc_thread(sbi)) 1563 f2fs_msg(sbi->sb, KERN_WARNING, 1564 "background gc thread has stopped"); 1565 } else if (need_stop_gc) { 1566 f2fs_stop_gc_thread(sbi); 1567 } 1568 restore_opts: 1569 #ifdef CONFIG_QUOTA 1570 F2FS_OPTION(sbi).s_jquota_fmt = org_mount_opt.s_jquota_fmt; 1571 for (i = 0; i < MAXQUOTAS; i++) { 1572 kfree(F2FS_OPTION(sbi).s_qf_names[i]); 1573 F2FS_OPTION(sbi).s_qf_names[i] = org_mount_opt.s_qf_names[i]; 1574 } 1575 #endif 1576 sbi->mount_opt = org_mount_opt; 1577 sb->s_flags = old_sb_flags; 1578 return err; 1579 } 1580 1581 #ifdef CONFIG_QUOTA 1582 /* Read data from quotafile */ 1583 static ssize_t f2fs_quota_read(struct super_block *sb, int type, char *data, 1584 size_t len, loff_t off) 1585 { 1586 struct inode *inode = sb_dqopt(sb)->files[type]; 1587 struct address_space *mapping = inode->i_mapping; 1588 block_t blkidx = F2FS_BYTES_TO_BLK(off); 1589 int offset = off & (sb->s_blocksize - 1); 1590 int tocopy; 1591 size_t toread; 1592 loff_t i_size = i_size_read(inode); 1593 struct page *page; 1594 char *kaddr; 1595 1596 if (off > i_size) 1597 return 0; 1598 1599 if (off + len > i_size) 1600 len = i_size - off; 1601 toread = len; 1602 while (toread > 0) { 1603 tocopy = min_t(unsigned long, sb->s_blocksize - offset, toread); 1604 repeat: 1605 page = read_cache_page_gfp(mapping, blkidx, GFP_NOFS); 1606 if (IS_ERR(page)) { 1607 if (PTR_ERR(page) == -ENOMEM) { 1608 congestion_wait(BLK_RW_ASYNC, HZ/50); 1609 goto repeat; 1610 } 1611 return PTR_ERR(page); 1612 } 1613 1614 lock_page(page); 1615 1616 if (unlikely(page->mapping != mapping)) { 1617 f2fs_put_page(page, 1); 1618 goto repeat; 1619 } 1620 if (unlikely(!PageUptodate(page))) { 1621 f2fs_put_page(page, 1); 1622 return -EIO; 1623 } 1624 1625 kaddr = kmap_atomic(page); 1626 memcpy(data, kaddr + offset, tocopy); 1627 kunmap_atomic(kaddr); 1628 f2fs_put_page(page, 1); 1629 1630 offset = 0; 1631 toread -= tocopy; 1632 data += tocopy; 1633 blkidx++; 1634 } 1635 return len; 1636 } 1637 1638 /* Write to quotafile */ 1639 static ssize_t f2fs_quota_write(struct super_block *sb, int type, 1640 const char *data, size_t len, loff_t off) 1641 { 1642 struct inode *inode = sb_dqopt(sb)->files[type]; 1643 struct address_space *mapping = inode->i_mapping; 1644 const struct address_space_operations *a_ops = mapping->a_ops; 1645 int offset = off & (sb->s_blocksize - 1); 1646 size_t towrite = len; 1647 struct page *page; 1648 char *kaddr; 1649 int err = 0; 1650 int tocopy; 1651 1652 while (towrite > 0) { 1653 tocopy = min_t(unsigned long, sb->s_blocksize - offset, 1654 towrite); 1655 retry: 1656 err = a_ops->write_begin(NULL, mapping, off, tocopy, 0, 1657 &page, NULL); 1658 if (unlikely(err)) { 1659 if (err == -ENOMEM) { 1660 congestion_wait(BLK_RW_ASYNC, HZ/50); 1661 goto retry; 1662 } 1663 break; 1664 } 1665 1666 kaddr = kmap_atomic(page); 1667 memcpy(kaddr + offset, data, tocopy); 1668 kunmap_atomic(kaddr); 1669 flush_dcache_page(page); 1670 1671 a_ops->write_end(NULL, mapping, off, tocopy, tocopy, 1672 page, NULL); 1673 offset = 0; 1674 towrite -= tocopy; 1675 off += tocopy; 1676 data += tocopy; 1677 cond_resched(); 1678 } 1679 1680 if (len == towrite) 1681 return err; 1682 inode->i_mtime = inode->i_ctime = current_time(inode); 1683 f2fs_mark_inode_dirty_sync(inode, false); 1684 return len - towrite; 1685 } 1686 1687 static struct dquot **f2fs_get_dquots(struct inode *inode) 1688 { 1689 return F2FS_I(inode)->i_dquot; 1690 } 1691 1692 static qsize_t *f2fs_get_reserved_space(struct inode *inode) 1693 { 1694 return &F2FS_I(inode)->i_reserved_quota; 1695 } 1696 1697 static int f2fs_quota_on_mount(struct f2fs_sb_info *sbi, int type) 1698 { 1699 return dquot_quota_on_mount(sbi->sb, F2FS_OPTION(sbi).s_qf_names[type], 1700 F2FS_OPTION(sbi).s_jquota_fmt, type); 1701 } 1702 1703 int f2fs_enable_quota_files(struct f2fs_sb_info *sbi, bool rdonly) 1704 { 1705 int enabled = 0; 1706 int i, err; 1707 1708 if (f2fs_sb_has_quota_ino(sbi->sb) && rdonly) { 1709 err = f2fs_enable_quotas(sbi->sb); 1710 if (err) { 1711 f2fs_msg(sbi->sb, KERN_ERR, 1712 "Cannot turn on quota_ino: %d", err); 1713 return 0; 1714 } 1715 return 1; 1716 } 1717 1718 for (i = 0; i < MAXQUOTAS; i++) { 1719 if (F2FS_OPTION(sbi).s_qf_names[i]) { 1720 err = f2fs_quota_on_mount(sbi, i); 1721 if (!err) { 1722 enabled = 1; 1723 continue; 1724 } 1725 f2fs_msg(sbi->sb, KERN_ERR, 1726 "Cannot turn on quotas: %d on %d", err, i); 1727 } 1728 } 1729 return enabled; 1730 } 1731 1732 static int f2fs_quota_enable(struct super_block *sb, int type, int format_id, 1733 unsigned int flags) 1734 { 1735 struct inode *qf_inode; 1736 unsigned long qf_inum; 1737 int err; 1738 1739 BUG_ON(!f2fs_sb_has_quota_ino(sb)); 1740 1741 qf_inum = f2fs_qf_ino(sb, type); 1742 if (!qf_inum) 1743 return -EPERM; 1744 1745 qf_inode = f2fs_iget(sb, qf_inum); 1746 if (IS_ERR(qf_inode)) { 1747 f2fs_msg(sb, KERN_ERR, 1748 "Bad quota inode %u:%lu", type, qf_inum); 1749 return PTR_ERR(qf_inode); 1750 } 1751 1752 /* Don't account quota for quota files to avoid recursion */ 1753 qf_inode->i_flags |= S_NOQUOTA; 1754 err = dquot_enable(qf_inode, type, format_id, flags); 1755 iput(qf_inode); 1756 return err; 1757 } 1758 1759 static int f2fs_enable_quotas(struct super_block *sb) 1760 { 1761 int type, err = 0; 1762 unsigned long qf_inum; 1763 bool quota_mopt[MAXQUOTAS] = { 1764 test_opt(F2FS_SB(sb), USRQUOTA), 1765 test_opt(F2FS_SB(sb), GRPQUOTA), 1766 test_opt(F2FS_SB(sb), PRJQUOTA), 1767 }; 1768 1769 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE | DQUOT_NOLIST_DIRTY; 1770 for (type = 0; type < MAXQUOTAS; type++) { 1771 qf_inum = f2fs_qf_ino(sb, type); 1772 if (qf_inum) { 1773 err = f2fs_quota_enable(sb, type, QFMT_VFS_V1, 1774 DQUOT_USAGE_ENABLED | 1775 (quota_mopt[type] ? DQUOT_LIMITS_ENABLED : 0)); 1776 if (err) { 1777 f2fs_msg(sb, KERN_ERR, 1778 "Failed to enable quota tracking " 1779 "(type=%d, err=%d). Please run " 1780 "fsck to fix.", type, err); 1781 for (type--; type >= 0; type--) 1782 dquot_quota_off(sb, type); 1783 return err; 1784 } 1785 } 1786 } 1787 return 0; 1788 } 1789 1790 static int f2fs_quota_sync(struct super_block *sb, int type) 1791 { 1792 struct quota_info *dqopt = sb_dqopt(sb); 1793 int cnt; 1794 int ret; 1795 1796 ret = dquot_writeback_dquots(sb, type); 1797 if (ret) 1798 return ret; 1799 1800 /* 1801 * Now when everything is written we can discard the pagecache so 1802 * that userspace sees the changes. 1803 */ 1804 for (cnt = 0; cnt < MAXQUOTAS; cnt++) { 1805 if (type != -1 && cnt != type) 1806 continue; 1807 if (!sb_has_quota_active(sb, cnt)) 1808 continue; 1809 1810 ret = filemap_write_and_wait(dqopt->files[cnt]->i_mapping); 1811 if (ret) 1812 return ret; 1813 1814 inode_lock(dqopt->files[cnt]); 1815 truncate_inode_pages(&dqopt->files[cnt]->i_data, 0); 1816 inode_unlock(dqopt->files[cnt]); 1817 } 1818 return 0; 1819 } 1820 1821 static int f2fs_quota_on(struct super_block *sb, int type, int format_id, 1822 const struct path *path) 1823 { 1824 struct inode *inode; 1825 int err; 1826 1827 err = f2fs_quota_sync(sb, type); 1828 if (err) 1829 return err; 1830 1831 err = dquot_quota_on(sb, type, format_id, path); 1832 if (err) 1833 return err; 1834 1835 inode = d_inode(path->dentry); 1836 1837 inode_lock(inode); 1838 F2FS_I(inode)->i_flags |= F2FS_NOATIME_FL | F2FS_IMMUTABLE_FL; 1839 inode_set_flags(inode, S_NOATIME | S_IMMUTABLE, 1840 S_NOATIME | S_IMMUTABLE); 1841 inode_unlock(inode); 1842 f2fs_mark_inode_dirty_sync(inode, false); 1843 1844 return 0; 1845 } 1846 1847 static int f2fs_quota_off(struct super_block *sb, int type) 1848 { 1849 struct inode *inode = sb_dqopt(sb)->files[type]; 1850 int err; 1851 1852 if (!inode || !igrab(inode)) 1853 return dquot_quota_off(sb, type); 1854 1855 f2fs_quota_sync(sb, type); 1856 1857 err = dquot_quota_off(sb, type); 1858 if (err || f2fs_sb_has_quota_ino(sb)) 1859 goto out_put; 1860 1861 inode_lock(inode); 1862 F2FS_I(inode)->i_flags &= ~(F2FS_NOATIME_FL | F2FS_IMMUTABLE_FL); 1863 inode_set_flags(inode, 0, S_NOATIME | S_IMMUTABLE); 1864 inode_unlock(inode); 1865 f2fs_mark_inode_dirty_sync(inode, false); 1866 out_put: 1867 iput(inode); 1868 return err; 1869 } 1870 1871 void f2fs_quota_off_umount(struct super_block *sb) 1872 { 1873 int type; 1874 1875 for (type = 0; type < MAXQUOTAS; type++) 1876 f2fs_quota_off(sb, type); 1877 } 1878 1879 static int f2fs_get_projid(struct inode *inode, kprojid_t *projid) 1880 { 1881 *projid = F2FS_I(inode)->i_projid; 1882 return 0; 1883 } 1884 1885 static const struct dquot_operations f2fs_quota_operations = { 1886 .get_reserved_space = f2fs_get_reserved_space, 1887 .write_dquot = dquot_commit, 1888 .acquire_dquot = dquot_acquire, 1889 .release_dquot = dquot_release, 1890 .mark_dirty = dquot_mark_dquot_dirty, 1891 .write_info = dquot_commit_info, 1892 .alloc_dquot = dquot_alloc, 1893 .destroy_dquot = dquot_destroy, 1894 .get_projid = f2fs_get_projid, 1895 .get_next_id = dquot_get_next_id, 1896 }; 1897 1898 static const struct quotactl_ops f2fs_quotactl_ops = { 1899 .quota_on = f2fs_quota_on, 1900 .quota_off = f2fs_quota_off, 1901 .quota_sync = f2fs_quota_sync, 1902 .get_state = dquot_get_state, 1903 .set_info = dquot_set_dqinfo, 1904 .get_dqblk = dquot_get_dqblk, 1905 .set_dqblk = dquot_set_dqblk, 1906 .get_nextdqblk = dquot_get_next_dqblk, 1907 }; 1908 #else 1909 void f2fs_quota_off_umount(struct super_block *sb) 1910 { 1911 } 1912 #endif 1913 1914 static const struct super_operations f2fs_sops = { 1915 .alloc_inode = f2fs_alloc_inode, 1916 .drop_inode = f2fs_drop_inode, 1917 .destroy_inode = f2fs_destroy_inode, 1918 .write_inode = f2fs_write_inode, 1919 .dirty_inode = f2fs_dirty_inode, 1920 .show_options = f2fs_show_options, 1921 #ifdef CONFIG_QUOTA 1922 .quota_read = f2fs_quota_read, 1923 .quota_write = f2fs_quota_write, 1924 .get_dquots = f2fs_get_dquots, 1925 #endif 1926 .evict_inode = f2fs_evict_inode, 1927 .put_super = f2fs_put_super, 1928 .sync_fs = f2fs_sync_fs, 1929 .freeze_fs = f2fs_freeze, 1930 .unfreeze_fs = f2fs_unfreeze, 1931 .statfs = f2fs_statfs, 1932 .remount_fs = f2fs_remount, 1933 }; 1934 1935 #ifdef CONFIG_F2FS_FS_ENCRYPTION 1936 static int f2fs_get_context(struct inode *inode, void *ctx, size_t len) 1937 { 1938 return f2fs_getxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION, 1939 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT, 1940 ctx, len, NULL); 1941 } 1942 1943 static int f2fs_set_context(struct inode *inode, const void *ctx, size_t len, 1944 void *fs_data) 1945 { 1946 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 1947 1948 /* 1949 * Encrypting the root directory is not allowed because fsck 1950 * expects lost+found directory to exist and remain unencrypted 1951 * if LOST_FOUND feature is enabled. 1952 * 1953 */ 1954 if (f2fs_sb_has_lost_found(sbi->sb) && 1955 inode->i_ino == F2FS_ROOT_INO(sbi)) 1956 return -EPERM; 1957 1958 return f2fs_setxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION, 1959 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT, 1960 ctx, len, fs_data, XATTR_CREATE); 1961 } 1962 1963 static bool f2fs_dummy_context(struct inode *inode) 1964 { 1965 return DUMMY_ENCRYPTION_ENABLED(F2FS_I_SB(inode)); 1966 } 1967 1968 static const struct fscrypt_operations f2fs_cryptops = { 1969 .key_prefix = "f2fs:", 1970 .get_context = f2fs_get_context, 1971 .set_context = f2fs_set_context, 1972 .dummy_context = f2fs_dummy_context, 1973 .empty_dir = f2fs_empty_dir, 1974 .max_namelen = F2FS_NAME_LEN, 1975 }; 1976 #endif 1977 1978 static struct inode *f2fs_nfs_get_inode(struct super_block *sb, 1979 u64 ino, u32 generation) 1980 { 1981 struct f2fs_sb_info *sbi = F2FS_SB(sb); 1982 struct inode *inode; 1983 1984 if (f2fs_check_nid_range(sbi, ino)) 1985 return ERR_PTR(-ESTALE); 1986 1987 /* 1988 * f2fs_iget isn't quite right if the inode is currently unallocated! 1989 * However f2fs_iget currently does appropriate checks to handle stale 1990 * inodes so everything is OK. 1991 */ 1992 inode = f2fs_iget(sb, ino); 1993 if (IS_ERR(inode)) 1994 return ERR_CAST(inode); 1995 if (unlikely(generation && inode->i_generation != generation)) { 1996 /* we didn't find the right inode.. */ 1997 iput(inode); 1998 return ERR_PTR(-ESTALE); 1999 } 2000 return inode; 2001 } 2002 2003 static struct dentry *f2fs_fh_to_dentry(struct super_block *sb, struct fid *fid, 2004 int fh_len, int fh_type) 2005 { 2006 return generic_fh_to_dentry(sb, fid, fh_len, fh_type, 2007 f2fs_nfs_get_inode); 2008 } 2009 2010 static struct dentry *f2fs_fh_to_parent(struct super_block *sb, struct fid *fid, 2011 int fh_len, int fh_type) 2012 { 2013 return generic_fh_to_parent(sb, fid, fh_len, fh_type, 2014 f2fs_nfs_get_inode); 2015 } 2016 2017 static const struct export_operations f2fs_export_ops = { 2018 .fh_to_dentry = f2fs_fh_to_dentry, 2019 .fh_to_parent = f2fs_fh_to_parent, 2020 .get_parent = f2fs_get_parent, 2021 }; 2022 2023 static loff_t max_file_blocks(void) 2024 { 2025 loff_t result = 0; 2026 loff_t leaf_count = ADDRS_PER_BLOCK; 2027 2028 /* 2029 * note: previously, result is equal to (DEF_ADDRS_PER_INODE - 2030 * DEFAULT_INLINE_XATTR_ADDRS), but now f2fs try to reserve more 2031 * space in inode.i_addr, it will be more safe to reassign 2032 * result as zero. 2033 */ 2034 2035 /* two direct node blocks */ 2036 result += (leaf_count * 2); 2037 2038 /* two indirect node blocks */ 2039 leaf_count *= NIDS_PER_BLOCK; 2040 result += (leaf_count * 2); 2041 2042 /* one double indirect node block */ 2043 leaf_count *= NIDS_PER_BLOCK; 2044 result += leaf_count; 2045 2046 return result; 2047 } 2048 2049 static int __f2fs_commit_super(struct buffer_head *bh, 2050 struct f2fs_super_block *super) 2051 { 2052 lock_buffer(bh); 2053 if (super) 2054 memcpy(bh->b_data + F2FS_SUPER_OFFSET, super, sizeof(*super)); 2055 set_buffer_dirty(bh); 2056 unlock_buffer(bh); 2057 2058 /* it's rare case, we can do fua all the time */ 2059 return __sync_dirty_buffer(bh, REQ_SYNC | REQ_PREFLUSH | REQ_FUA); 2060 } 2061 2062 static inline bool sanity_check_area_boundary(struct f2fs_sb_info *sbi, 2063 struct buffer_head *bh) 2064 { 2065 struct f2fs_super_block *raw_super = (struct f2fs_super_block *) 2066 (bh->b_data + F2FS_SUPER_OFFSET); 2067 struct super_block *sb = sbi->sb; 2068 u32 segment0_blkaddr = le32_to_cpu(raw_super->segment0_blkaddr); 2069 u32 cp_blkaddr = le32_to_cpu(raw_super->cp_blkaddr); 2070 u32 sit_blkaddr = le32_to_cpu(raw_super->sit_blkaddr); 2071 u32 nat_blkaddr = le32_to_cpu(raw_super->nat_blkaddr); 2072 u32 ssa_blkaddr = le32_to_cpu(raw_super->ssa_blkaddr); 2073 u32 main_blkaddr = le32_to_cpu(raw_super->main_blkaddr); 2074 u32 segment_count_ckpt = le32_to_cpu(raw_super->segment_count_ckpt); 2075 u32 segment_count_sit = le32_to_cpu(raw_super->segment_count_sit); 2076 u32 segment_count_nat = le32_to_cpu(raw_super->segment_count_nat); 2077 u32 segment_count_ssa = le32_to_cpu(raw_super->segment_count_ssa); 2078 u32 segment_count_main = le32_to_cpu(raw_super->segment_count_main); 2079 u32 segment_count = le32_to_cpu(raw_super->segment_count); 2080 u32 log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg); 2081 u64 main_end_blkaddr = main_blkaddr + 2082 (segment_count_main << log_blocks_per_seg); 2083 u64 seg_end_blkaddr = segment0_blkaddr + 2084 (segment_count << log_blocks_per_seg); 2085 2086 if (segment0_blkaddr != cp_blkaddr) { 2087 f2fs_msg(sb, KERN_INFO, 2088 "Mismatch start address, segment0(%u) cp_blkaddr(%u)", 2089 segment0_blkaddr, cp_blkaddr); 2090 return true; 2091 } 2092 2093 if (cp_blkaddr + (segment_count_ckpt << log_blocks_per_seg) != 2094 sit_blkaddr) { 2095 f2fs_msg(sb, KERN_INFO, 2096 "Wrong CP boundary, start(%u) end(%u) blocks(%u)", 2097 cp_blkaddr, sit_blkaddr, 2098 segment_count_ckpt << log_blocks_per_seg); 2099 return true; 2100 } 2101 2102 if (sit_blkaddr + (segment_count_sit << log_blocks_per_seg) != 2103 nat_blkaddr) { 2104 f2fs_msg(sb, KERN_INFO, 2105 "Wrong SIT boundary, start(%u) end(%u) blocks(%u)", 2106 sit_blkaddr, nat_blkaddr, 2107 segment_count_sit << log_blocks_per_seg); 2108 return true; 2109 } 2110 2111 if (nat_blkaddr + (segment_count_nat << log_blocks_per_seg) != 2112 ssa_blkaddr) { 2113 f2fs_msg(sb, KERN_INFO, 2114 "Wrong NAT boundary, start(%u) end(%u) blocks(%u)", 2115 nat_blkaddr, ssa_blkaddr, 2116 segment_count_nat << log_blocks_per_seg); 2117 return true; 2118 } 2119 2120 if (ssa_blkaddr + (segment_count_ssa << log_blocks_per_seg) != 2121 main_blkaddr) { 2122 f2fs_msg(sb, KERN_INFO, 2123 "Wrong SSA boundary, start(%u) end(%u) blocks(%u)", 2124 ssa_blkaddr, main_blkaddr, 2125 segment_count_ssa << log_blocks_per_seg); 2126 return true; 2127 } 2128 2129 if (main_end_blkaddr > seg_end_blkaddr) { 2130 f2fs_msg(sb, KERN_INFO, 2131 "Wrong MAIN_AREA boundary, start(%u) end(%u) block(%u)", 2132 main_blkaddr, 2133 segment0_blkaddr + 2134 (segment_count << log_blocks_per_seg), 2135 segment_count_main << log_blocks_per_seg); 2136 return true; 2137 } else if (main_end_blkaddr < seg_end_blkaddr) { 2138 int err = 0; 2139 char *res; 2140 2141 /* fix in-memory information all the time */ 2142 raw_super->segment_count = cpu_to_le32((main_end_blkaddr - 2143 segment0_blkaddr) >> log_blocks_per_seg); 2144 2145 if (f2fs_readonly(sb) || bdev_read_only(sb->s_bdev)) { 2146 set_sbi_flag(sbi, SBI_NEED_SB_WRITE); 2147 res = "internally"; 2148 } else { 2149 err = __f2fs_commit_super(bh, NULL); 2150 res = err ? "failed" : "done"; 2151 } 2152 f2fs_msg(sb, KERN_INFO, 2153 "Fix alignment : %s, start(%u) end(%u) block(%u)", 2154 res, main_blkaddr, 2155 segment0_blkaddr + 2156 (segment_count << log_blocks_per_seg), 2157 segment_count_main << log_blocks_per_seg); 2158 if (err) 2159 return true; 2160 } 2161 return false; 2162 } 2163 2164 static int sanity_check_raw_super(struct f2fs_sb_info *sbi, 2165 struct buffer_head *bh) 2166 { 2167 block_t segment_count, segs_per_sec, secs_per_zone; 2168 block_t total_sections, blocks_per_seg; 2169 struct f2fs_super_block *raw_super = (struct f2fs_super_block *) 2170 (bh->b_data + F2FS_SUPER_OFFSET); 2171 struct super_block *sb = sbi->sb; 2172 unsigned int blocksize; 2173 2174 if (F2FS_SUPER_MAGIC != le32_to_cpu(raw_super->magic)) { 2175 f2fs_msg(sb, KERN_INFO, 2176 "Magic Mismatch, valid(0x%x) - read(0x%x)", 2177 F2FS_SUPER_MAGIC, le32_to_cpu(raw_super->magic)); 2178 return 1; 2179 } 2180 2181 /* Currently, support only 4KB page cache size */ 2182 if (F2FS_BLKSIZE != PAGE_SIZE) { 2183 f2fs_msg(sb, KERN_INFO, 2184 "Invalid page_cache_size (%lu), supports only 4KB\n", 2185 PAGE_SIZE); 2186 return 1; 2187 } 2188 2189 /* Currently, support only 4KB block size */ 2190 blocksize = 1 << le32_to_cpu(raw_super->log_blocksize); 2191 if (blocksize != F2FS_BLKSIZE) { 2192 f2fs_msg(sb, KERN_INFO, 2193 "Invalid blocksize (%u), supports only 4KB\n", 2194 blocksize); 2195 return 1; 2196 } 2197 2198 /* check log blocks per segment */ 2199 if (le32_to_cpu(raw_super->log_blocks_per_seg) != 9) { 2200 f2fs_msg(sb, KERN_INFO, 2201 "Invalid log blocks per segment (%u)\n", 2202 le32_to_cpu(raw_super->log_blocks_per_seg)); 2203 return 1; 2204 } 2205 2206 /* Currently, support 512/1024/2048/4096 bytes sector size */ 2207 if (le32_to_cpu(raw_super->log_sectorsize) > 2208 F2FS_MAX_LOG_SECTOR_SIZE || 2209 le32_to_cpu(raw_super->log_sectorsize) < 2210 F2FS_MIN_LOG_SECTOR_SIZE) { 2211 f2fs_msg(sb, KERN_INFO, "Invalid log sectorsize (%u)", 2212 le32_to_cpu(raw_super->log_sectorsize)); 2213 return 1; 2214 } 2215 if (le32_to_cpu(raw_super->log_sectors_per_block) + 2216 le32_to_cpu(raw_super->log_sectorsize) != 2217 F2FS_MAX_LOG_SECTOR_SIZE) { 2218 f2fs_msg(sb, KERN_INFO, 2219 "Invalid log sectors per block(%u) log sectorsize(%u)", 2220 le32_to_cpu(raw_super->log_sectors_per_block), 2221 le32_to_cpu(raw_super->log_sectorsize)); 2222 return 1; 2223 } 2224 2225 segment_count = le32_to_cpu(raw_super->segment_count); 2226 segs_per_sec = le32_to_cpu(raw_super->segs_per_sec); 2227 secs_per_zone = le32_to_cpu(raw_super->secs_per_zone); 2228 total_sections = le32_to_cpu(raw_super->section_count); 2229 2230 /* blocks_per_seg should be 512, given the above check */ 2231 blocks_per_seg = 1 << le32_to_cpu(raw_super->log_blocks_per_seg); 2232 2233 if (segment_count > F2FS_MAX_SEGMENT || 2234 segment_count < F2FS_MIN_SEGMENTS) { 2235 f2fs_msg(sb, KERN_INFO, 2236 "Invalid segment count (%u)", 2237 segment_count); 2238 return 1; 2239 } 2240 2241 if (total_sections > segment_count || 2242 total_sections < F2FS_MIN_SEGMENTS || 2243 segs_per_sec > segment_count || !segs_per_sec) { 2244 f2fs_msg(sb, KERN_INFO, 2245 "Invalid segment/section count (%u, %u x %u)", 2246 segment_count, total_sections, segs_per_sec); 2247 return 1; 2248 } 2249 2250 if ((segment_count / segs_per_sec) < total_sections) { 2251 f2fs_msg(sb, KERN_INFO, 2252 "Small segment_count (%u < %u * %u)", 2253 segment_count, segs_per_sec, total_sections); 2254 return 1; 2255 } 2256 2257 if (segment_count > (le32_to_cpu(raw_super->block_count) >> 9)) { 2258 f2fs_msg(sb, KERN_INFO, 2259 "Wrong segment_count / block_count (%u > %u)", 2260 segment_count, le32_to_cpu(raw_super->block_count)); 2261 return 1; 2262 } 2263 2264 if (secs_per_zone > total_sections || !secs_per_zone) { 2265 f2fs_msg(sb, KERN_INFO, 2266 "Wrong secs_per_zone / total_sections (%u, %u)", 2267 secs_per_zone, total_sections); 2268 return 1; 2269 } 2270 if (le32_to_cpu(raw_super->extension_count) > F2FS_MAX_EXTENSION || 2271 raw_super->hot_ext_count > F2FS_MAX_EXTENSION || 2272 (le32_to_cpu(raw_super->extension_count) + 2273 raw_super->hot_ext_count) > F2FS_MAX_EXTENSION) { 2274 f2fs_msg(sb, KERN_INFO, 2275 "Corrupted extension count (%u + %u > %u)", 2276 le32_to_cpu(raw_super->extension_count), 2277 raw_super->hot_ext_count, 2278 F2FS_MAX_EXTENSION); 2279 return 1; 2280 } 2281 2282 if (le32_to_cpu(raw_super->cp_payload) > 2283 (blocks_per_seg - F2FS_CP_PACKS)) { 2284 f2fs_msg(sb, KERN_INFO, 2285 "Insane cp_payload (%u > %u)", 2286 le32_to_cpu(raw_super->cp_payload), 2287 blocks_per_seg - F2FS_CP_PACKS); 2288 return 1; 2289 } 2290 2291 /* check reserved ino info */ 2292 if (le32_to_cpu(raw_super->node_ino) != 1 || 2293 le32_to_cpu(raw_super->meta_ino) != 2 || 2294 le32_to_cpu(raw_super->root_ino) != 3) { 2295 f2fs_msg(sb, KERN_INFO, 2296 "Invalid Fs Meta Ino: node(%u) meta(%u) root(%u)", 2297 le32_to_cpu(raw_super->node_ino), 2298 le32_to_cpu(raw_super->meta_ino), 2299 le32_to_cpu(raw_super->root_ino)); 2300 return 1; 2301 } 2302 2303 /* check CP/SIT/NAT/SSA/MAIN_AREA area boundary */ 2304 if (sanity_check_area_boundary(sbi, bh)) 2305 return 1; 2306 2307 return 0; 2308 } 2309 2310 int f2fs_sanity_check_ckpt(struct f2fs_sb_info *sbi) 2311 { 2312 unsigned int total, fsmeta; 2313 struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi); 2314 struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi); 2315 unsigned int ovp_segments, reserved_segments; 2316 unsigned int main_segs, blocks_per_seg; 2317 unsigned int sit_segs, nat_segs; 2318 unsigned int sit_bitmap_size, nat_bitmap_size; 2319 unsigned int log_blocks_per_seg; 2320 unsigned int segment_count_main; 2321 unsigned int cp_pack_start_sum, cp_payload; 2322 block_t user_block_count; 2323 int i; 2324 2325 total = le32_to_cpu(raw_super->segment_count); 2326 fsmeta = le32_to_cpu(raw_super->segment_count_ckpt); 2327 sit_segs = le32_to_cpu(raw_super->segment_count_sit); 2328 fsmeta += sit_segs; 2329 nat_segs = le32_to_cpu(raw_super->segment_count_nat); 2330 fsmeta += nat_segs; 2331 fsmeta += le32_to_cpu(ckpt->rsvd_segment_count); 2332 fsmeta += le32_to_cpu(raw_super->segment_count_ssa); 2333 2334 if (unlikely(fsmeta >= total)) 2335 return 1; 2336 2337 ovp_segments = le32_to_cpu(ckpt->overprov_segment_count); 2338 reserved_segments = le32_to_cpu(ckpt->rsvd_segment_count); 2339 2340 if (unlikely(fsmeta < F2FS_MIN_SEGMENTS || 2341 ovp_segments == 0 || reserved_segments == 0)) { 2342 f2fs_msg(sbi->sb, KERN_ERR, 2343 "Wrong layout: check mkfs.f2fs version"); 2344 return 1; 2345 } 2346 2347 user_block_count = le64_to_cpu(ckpt->user_block_count); 2348 segment_count_main = le32_to_cpu(raw_super->segment_count_main); 2349 log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg); 2350 if (!user_block_count || user_block_count >= 2351 segment_count_main << log_blocks_per_seg) { 2352 f2fs_msg(sbi->sb, KERN_ERR, 2353 "Wrong user_block_count: %u", user_block_count); 2354 return 1; 2355 } 2356 2357 main_segs = le32_to_cpu(raw_super->segment_count_main); 2358 blocks_per_seg = sbi->blocks_per_seg; 2359 2360 for (i = 0; i < NR_CURSEG_NODE_TYPE; i++) { 2361 if (le32_to_cpu(ckpt->cur_node_segno[i]) >= main_segs || 2362 le16_to_cpu(ckpt->cur_node_blkoff[i]) >= blocks_per_seg) 2363 return 1; 2364 } 2365 for (i = 0; i < NR_CURSEG_DATA_TYPE; i++) { 2366 if (le32_to_cpu(ckpt->cur_data_segno[i]) >= main_segs || 2367 le16_to_cpu(ckpt->cur_data_blkoff[i]) >= blocks_per_seg) 2368 return 1; 2369 } 2370 2371 sit_bitmap_size = le32_to_cpu(ckpt->sit_ver_bitmap_bytesize); 2372 nat_bitmap_size = le32_to_cpu(ckpt->nat_ver_bitmap_bytesize); 2373 2374 if (sit_bitmap_size != ((sit_segs / 2) << log_blocks_per_seg) / 8 || 2375 nat_bitmap_size != ((nat_segs / 2) << log_blocks_per_seg) / 8) { 2376 f2fs_msg(sbi->sb, KERN_ERR, 2377 "Wrong bitmap size: sit: %u, nat:%u", 2378 sit_bitmap_size, nat_bitmap_size); 2379 return 1; 2380 } 2381 2382 cp_pack_start_sum = __start_sum_addr(sbi); 2383 cp_payload = __cp_payload(sbi); 2384 if (cp_pack_start_sum < cp_payload + 1 || 2385 cp_pack_start_sum > blocks_per_seg - 1 - 2386 NR_CURSEG_TYPE) { 2387 f2fs_msg(sbi->sb, KERN_ERR, 2388 "Wrong cp_pack_start_sum: %u", 2389 cp_pack_start_sum); 2390 return 1; 2391 } 2392 2393 if (unlikely(f2fs_cp_error(sbi))) { 2394 f2fs_msg(sbi->sb, KERN_ERR, "A bug case: need to run fsck"); 2395 return 1; 2396 } 2397 return 0; 2398 } 2399 2400 static void init_sb_info(struct f2fs_sb_info *sbi) 2401 { 2402 struct f2fs_super_block *raw_super = sbi->raw_super; 2403 int i, j; 2404 2405 sbi->log_sectors_per_block = 2406 le32_to_cpu(raw_super->log_sectors_per_block); 2407 sbi->log_blocksize = le32_to_cpu(raw_super->log_blocksize); 2408 sbi->blocksize = 1 << sbi->log_blocksize; 2409 sbi->log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg); 2410 sbi->blocks_per_seg = 1 << sbi->log_blocks_per_seg; 2411 sbi->segs_per_sec = le32_to_cpu(raw_super->segs_per_sec); 2412 sbi->secs_per_zone = le32_to_cpu(raw_super->secs_per_zone); 2413 sbi->total_sections = le32_to_cpu(raw_super->section_count); 2414 sbi->total_node_count = 2415 (le32_to_cpu(raw_super->segment_count_nat) / 2) 2416 * sbi->blocks_per_seg * NAT_ENTRY_PER_BLOCK; 2417 sbi->root_ino_num = le32_to_cpu(raw_super->root_ino); 2418 sbi->node_ino_num = le32_to_cpu(raw_super->node_ino); 2419 sbi->meta_ino_num = le32_to_cpu(raw_super->meta_ino); 2420 sbi->cur_victim_sec = NULL_SECNO; 2421 sbi->max_victim_search = DEF_MAX_VICTIM_SEARCH; 2422 2423 sbi->dir_level = DEF_DIR_LEVEL; 2424 sbi->interval_time[CP_TIME] = DEF_CP_INTERVAL; 2425 sbi->interval_time[REQ_TIME] = DEF_IDLE_INTERVAL; 2426 clear_sbi_flag(sbi, SBI_NEED_FSCK); 2427 2428 for (i = 0; i < NR_COUNT_TYPE; i++) 2429 atomic_set(&sbi->nr_pages[i], 0); 2430 2431 for (i = 0; i < META; i++) 2432 atomic_set(&sbi->wb_sync_req[i], 0); 2433 2434 INIT_LIST_HEAD(&sbi->s_list); 2435 mutex_init(&sbi->umount_mutex); 2436 for (i = 0; i < NR_PAGE_TYPE - 1; i++) 2437 for (j = HOT; j < NR_TEMP_TYPE; j++) 2438 mutex_init(&sbi->wio_mutex[i][j]); 2439 init_rwsem(&sbi->io_order_lock); 2440 spin_lock_init(&sbi->cp_lock); 2441 2442 sbi->dirty_device = 0; 2443 spin_lock_init(&sbi->dev_lock); 2444 2445 init_rwsem(&sbi->sb_lock); 2446 } 2447 2448 static int init_percpu_info(struct f2fs_sb_info *sbi) 2449 { 2450 int err; 2451 2452 err = percpu_counter_init(&sbi->alloc_valid_block_count, 0, GFP_KERNEL); 2453 if (err) 2454 return err; 2455 2456 return percpu_counter_init(&sbi->total_valid_inode_count, 0, 2457 GFP_KERNEL); 2458 } 2459 2460 #ifdef CONFIG_BLK_DEV_ZONED 2461 static int init_blkz_info(struct f2fs_sb_info *sbi, int devi) 2462 { 2463 struct block_device *bdev = FDEV(devi).bdev; 2464 sector_t nr_sectors = bdev->bd_part->nr_sects; 2465 sector_t sector = 0; 2466 struct blk_zone *zones; 2467 unsigned int i, nr_zones; 2468 unsigned int n = 0; 2469 int err = -EIO; 2470 2471 if (!f2fs_sb_has_blkzoned(sbi->sb)) 2472 return 0; 2473 2474 if (sbi->blocks_per_blkz && sbi->blocks_per_blkz != 2475 SECTOR_TO_BLOCK(bdev_zone_sectors(bdev))) 2476 return -EINVAL; 2477 sbi->blocks_per_blkz = SECTOR_TO_BLOCK(bdev_zone_sectors(bdev)); 2478 if (sbi->log_blocks_per_blkz && sbi->log_blocks_per_blkz != 2479 __ilog2_u32(sbi->blocks_per_blkz)) 2480 return -EINVAL; 2481 sbi->log_blocks_per_blkz = __ilog2_u32(sbi->blocks_per_blkz); 2482 FDEV(devi).nr_blkz = SECTOR_TO_BLOCK(nr_sectors) >> 2483 sbi->log_blocks_per_blkz; 2484 if (nr_sectors & (bdev_zone_sectors(bdev) - 1)) 2485 FDEV(devi).nr_blkz++; 2486 2487 FDEV(devi).blkz_type = f2fs_kmalloc(sbi, FDEV(devi).nr_blkz, 2488 GFP_KERNEL); 2489 if (!FDEV(devi).blkz_type) 2490 return -ENOMEM; 2491 2492 #define F2FS_REPORT_NR_ZONES 4096 2493 2494 zones = f2fs_kzalloc(sbi, 2495 array_size(F2FS_REPORT_NR_ZONES, 2496 sizeof(struct blk_zone)), 2497 GFP_KERNEL); 2498 if (!zones) 2499 return -ENOMEM; 2500 2501 /* Get block zones type */ 2502 while (zones && sector < nr_sectors) { 2503 2504 nr_zones = F2FS_REPORT_NR_ZONES; 2505 err = blkdev_report_zones(bdev, sector, 2506 zones, &nr_zones, 2507 GFP_KERNEL); 2508 if (err) 2509 break; 2510 if (!nr_zones) { 2511 err = -EIO; 2512 break; 2513 } 2514 2515 for (i = 0; i < nr_zones; i++) { 2516 FDEV(devi).blkz_type[n] = zones[i].type; 2517 sector += zones[i].len; 2518 n++; 2519 } 2520 } 2521 2522 kfree(zones); 2523 2524 return err; 2525 } 2526 #endif 2527 2528 /* 2529 * Read f2fs raw super block. 2530 * Because we have two copies of super block, so read both of them 2531 * to get the first valid one. If any one of them is broken, we pass 2532 * them recovery flag back to the caller. 2533 */ 2534 static int read_raw_super_block(struct f2fs_sb_info *sbi, 2535 struct f2fs_super_block **raw_super, 2536 int *valid_super_block, int *recovery) 2537 { 2538 struct super_block *sb = sbi->sb; 2539 int block; 2540 struct buffer_head *bh; 2541 struct f2fs_super_block *super; 2542 int err = 0; 2543 2544 super = kzalloc(sizeof(struct f2fs_super_block), GFP_KERNEL); 2545 if (!super) 2546 return -ENOMEM; 2547 2548 for (block = 0; block < 2; block++) { 2549 bh = sb_bread(sb, block); 2550 if (!bh) { 2551 f2fs_msg(sb, KERN_ERR, "Unable to read %dth superblock", 2552 block + 1); 2553 err = -EIO; 2554 continue; 2555 } 2556 2557 /* sanity checking of raw super */ 2558 if (sanity_check_raw_super(sbi, bh)) { 2559 f2fs_msg(sb, KERN_ERR, 2560 "Can't find valid F2FS filesystem in %dth superblock", 2561 block + 1); 2562 err = -EINVAL; 2563 brelse(bh); 2564 continue; 2565 } 2566 2567 if (!*raw_super) { 2568 memcpy(super, bh->b_data + F2FS_SUPER_OFFSET, 2569 sizeof(*super)); 2570 *valid_super_block = block; 2571 *raw_super = super; 2572 } 2573 brelse(bh); 2574 } 2575 2576 /* Fail to read any one of the superblocks*/ 2577 if (err < 0) 2578 *recovery = 1; 2579 2580 /* No valid superblock */ 2581 if (!*raw_super) 2582 kfree(super); 2583 else 2584 err = 0; 2585 2586 return err; 2587 } 2588 2589 int f2fs_commit_super(struct f2fs_sb_info *sbi, bool recover) 2590 { 2591 struct buffer_head *bh; 2592 int err; 2593 2594 if ((recover && f2fs_readonly(sbi->sb)) || 2595 bdev_read_only(sbi->sb->s_bdev)) { 2596 set_sbi_flag(sbi, SBI_NEED_SB_WRITE); 2597 return -EROFS; 2598 } 2599 2600 /* write back-up superblock first */ 2601 bh = sb_bread(sbi->sb, sbi->valid_super_block ? 0 : 1); 2602 if (!bh) 2603 return -EIO; 2604 err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi)); 2605 brelse(bh); 2606 2607 /* if we are in recovery path, skip writing valid superblock */ 2608 if (recover || err) 2609 return err; 2610 2611 /* write current valid superblock */ 2612 bh = sb_bread(sbi->sb, sbi->valid_super_block); 2613 if (!bh) 2614 return -EIO; 2615 err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi)); 2616 brelse(bh); 2617 return err; 2618 } 2619 2620 static int f2fs_scan_devices(struct f2fs_sb_info *sbi) 2621 { 2622 struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi); 2623 unsigned int max_devices = MAX_DEVICES; 2624 int i; 2625 2626 /* Initialize single device information */ 2627 if (!RDEV(0).path[0]) { 2628 if (!bdev_is_zoned(sbi->sb->s_bdev)) 2629 return 0; 2630 max_devices = 1; 2631 } 2632 2633 /* 2634 * Initialize multiple devices information, or single 2635 * zoned block device information. 2636 */ 2637 sbi->devs = f2fs_kzalloc(sbi, 2638 array_size(max_devices, 2639 sizeof(struct f2fs_dev_info)), 2640 GFP_KERNEL); 2641 if (!sbi->devs) 2642 return -ENOMEM; 2643 2644 for (i = 0; i < max_devices; i++) { 2645 2646 if (i > 0 && !RDEV(i).path[0]) 2647 break; 2648 2649 if (max_devices == 1) { 2650 /* Single zoned block device mount */ 2651 FDEV(0).bdev = 2652 blkdev_get_by_dev(sbi->sb->s_bdev->bd_dev, 2653 sbi->sb->s_mode, sbi->sb->s_type); 2654 } else { 2655 /* Multi-device mount */ 2656 memcpy(FDEV(i).path, RDEV(i).path, MAX_PATH_LEN); 2657 FDEV(i).total_segments = 2658 le32_to_cpu(RDEV(i).total_segments); 2659 if (i == 0) { 2660 FDEV(i).start_blk = 0; 2661 FDEV(i).end_blk = FDEV(i).start_blk + 2662 (FDEV(i).total_segments << 2663 sbi->log_blocks_per_seg) - 1 + 2664 le32_to_cpu(raw_super->segment0_blkaddr); 2665 } else { 2666 FDEV(i).start_blk = FDEV(i - 1).end_blk + 1; 2667 FDEV(i).end_blk = FDEV(i).start_blk + 2668 (FDEV(i).total_segments << 2669 sbi->log_blocks_per_seg) - 1; 2670 } 2671 FDEV(i).bdev = blkdev_get_by_path(FDEV(i).path, 2672 sbi->sb->s_mode, sbi->sb->s_type); 2673 } 2674 if (IS_ERR(FDEV(i).bdev)) 2675 return PTR_ERR(FDEV(i).bdev); 2676 2677 /* to release errored devices */ 2678 sbi->s_ndevs = i + 1; 2679 2680 #ifdef CONFIG_BLK_DEV_ZONED 2681 if (bdev_zoned_model(FDEV(i).bdev) == BLK_ZONED_HM && 2682 !f2fs_sb_has_blkzoned(sbi->sb)) { 2683 f2fs_msg(sbi->sb, KERN_ERR, 2684 "Zoned block device feature not enabled\n"); 2685 return -EINVAL; 2686 } 2687 if (bdev_zoned_model(FDEV(i).bdev) != BLK_ZONED_NONE) { 2688 if (init_blkz_info(sbi, i)) { 2689 f2fs_msg(sbi->sb, KERN_ERR, 2690 "Failed to initialize F2FS blkzone information"); 2691 return -EINVAL; 2692 } 2693 if (max_devices == 1) 2694 break; 2695 f2fs_msg(sbi->sb, KERN_INFO, 2696 "Mount Device [%2d]: %20s, %8u, %8x - %8x (zone: %s)", 2697 i, FDEV(i).path, 2698 FDEV(i).total_segments, 2699 FDEV(i).start_blk, FDEV(i).end_blk, 2700 bdev_zoned_model(FDEV(i).bdev) == BLK_ZONED_HA ? 2701 "Host-aware" : "Host-managed"); 2702 continue; 2703 } 2704 #endif 2705 f2fs_msg(sbi->sb, KERN_INFO, 2706 "Mount Device [%2d]: %20s, %8u, %8x - %8x", 2707 i, FDEV(i).path, 2708 FDEV(i).total_segments, 2709 FDEV(i).start_blk, FDEV(i).end_blk); 2710 } 2711 f2fs_msg(sbi->sb, KERN_INFO, 2712 "IO Block Size: %8d KB", F2FS_IO_SIZE_KB(sbi)); 2713 return 0; 2714 } 2715 2716 static void f2fs_tuning_parameters(struct f2fs_sb_info *sbi) 2717 { 2718 struct f2fs_sm_info *sm_i = SM_I(sbi); 2719 2720 /* adjust parameters according to the volume size */ 2721 if (sm_i->main_segments <= SMALL_VOLUME_SEGMENTS) { 2722 F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_REUSE; 2723 sm_i->dcc_info->discard_granularity = 1; 2724 sm_i->ipu_policy = 1 << F2FS_IPU_FORCE; 2725 } 2726 2727 sbi->readdir_ra = 1; 2728 } 2729 2730 static int f2fs_fill_super(struct super_block *sb, void *data, int silent) 2731 { 2732 struct f2fs_sb_info *sbi; 2733 struct f2fs_super_block *raw_super; 2734 struct inode *root; 2735 int err; 2736 bool retry = true, need_fsck = false; 2737 char *options = NULL; 2738 int recovery, i, valid_super_block; 2739 struct curseg_info *seg_i; 2740 2741 try_onemore: 2742 err = -EINVAL; 2743 raw_super = NULL; 2744 valid_super_block = -1; 2745 recovery = 0; 2746 2747 /* allocate memory for f2fs-specific super block info */ 2748 sbi = kzalloc(sizeof(struct f2fs_sb_info), GFP_KERNEL); 2749 if (!sbi) 2750 return -ENOMEM; 2751 2752 sbi->sb = sb; 2753 2754 /* Load the checksum driver */ 2755 sbi->s_chksum_driver = crypto_alloc_shash("crc32", 0, 0); 2756 if (IS_ERR(sbi->s_chksum_driver)) { 2757 f2fs_msg(sb, KERN_ERR, "Cannot load crc32 driver."); 2758 err = PTR_ERR(sbi->s_chksum_driver); 2759 sbi->s_chksum_driver = NULL; 2760 goto free_sbi; 2761 } 2762 2763 /* set a block size */ 2764 if (unlikely(!sb_set_blocksize(sb, F2FS_BLKSIZE))) { 2765 f2fs_msg(sb, KERN_ERR, "unable to set blocksize"); 2766 goto free_sbi; 2767 } 2768 2769 err = read_raw_super_block(sbi, &raw_super, &valid_super_block, 2770 &recovery); 2771 if (err) 2772 goto free_sbi; 2773 2774 sb->s_fs_info = sbi; 2775 sbi->raw_super = raw_super; 2776 2777 /* precompute checksum seed for metadata */ 2778 if (f2fs_sb_has_inode_chksum(sb)) 2779 sbi->s_chksum_seed = f2fs_chksum(sbi, ~0, raw_super->uuid, 2780 sizeof(raw_super->uuid)); 2781 2782 /* 2783 * The BLKZONED feature indicates that the drive was formatted with 2784 * zone alignment optimization. This is optional for host-aware 2785 * devices, but mandatory for host-managed zoned block devices. 2786 */ 2787 #ifndef CONFIG_BLK_DEV_ZONED 2788 if (f2fs_sb_has_blkzoned(sb)) { 2789 f2fs_msg(sb, KERN_ERR, 2790 "Zoned block device support is not enabled\n"); 2791 err = -EOPNOTSUPP; 2792 goto free_sb_buf; 2793 } 2794 #endif 2795 default_options(sbi); 2796 /* parse mount options */ 2797 options = kstrdup((const char *)data, GFP_KERNEL); 2798 if (data && !options) { 2799 err = -ENOMEM; 2800 goto free_sb_buf; 2801 } 2802 2803 err = parse_options(sb, options); 2804 if (err) 2805 goto free_options; 2806 2807 sbi->max_file_blocks = max_file_blocks(); 2808 sb->s_maxbytes = sbi->max_file_blocks << 2809 le32_to_cpu(raw_super->log_blocksize); 2810 sb->s_max_links = F2FS_LINK_MAX; 2811 get_random_bytes(&sbi->s_next_generation, sizeof(u32)); 2812 2813 #ifdef CONFIG_QUOTA 2814 sb->dq_op = &f2fs_quota_operations; 2815 if (f2fs_sb_has_quota_ino(sb)) 2816 sb->s_qcop = &dquot_quotactl_sysfile_ops; 2817 else 2818 sb->s_qcop = &f2fs_quotactl_ops; 2819 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ; 2820 2821 if (f2fs_sb_has_quota_ino(sbi->sb)) { 2822 for (i = 0; i < MAXQUOTAS; i++) { 2823 if (f2fs_qf_ino(sbi->sb, i)) 2824 sbi->nquota_files++; 2825 } 2826 } 2827 #endif 2828 2829 sb->s_op = &f2fs_sops; 2830 #ifdef CONFIG_F2FS_FS_ENCRYPTION 2831 sb->s_cop = &f2fs_cryptops; 2832 #endif 2833 sb->s_xattr = f2fs_xattr_handlers; 2834 sb->s_export_op = &f2fs_export_ops; 2835 sb->s_magic = F2FS_SUPER_MAGIC; 2836 sb->s_time_gran = 1; 2837 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) | 2838 (test_opt(sbi, POSIX_ACL) ? SB_POSIXACL : 0); 2839 memcpy(&sb->s_uuid, raw_super->uuid, sizeof(raw_super->uuid)); 2840 sb->s_iflags |= SB_I_CGROUPWB; 2841 2842 /* init f2fs-specific super block info */ 2843 sbi->valid_super_block = valid_super_block; 2844 mutex_init(&sbi->gc_mutex); 2845 mutex_init(&sbi->writepages); 2846 mutex_init(&sbi->cp_mutex); 2847 init_rwsem(&sbi->node_write); 2848 init_rwsem(&sbi->node_change); 2849 2850 /* disallow all the data/node/meta page writes */ 2851 set_sbi_flag(sbi, SBI_POR_DOING); 2852 spin_lock_init(&sbi->stat_lock); 2853 2854 /* init iostat info */ 2855 spin_lock_init(&sbi->iostat_lock); 2856 sbi->iostat_enable = false; 2857 2858 for (i = 0; i < NR_PAGE_TYPE; i++) { 2859 int n = (i == META) ? 1: NR_TEMP_TYPE; 2860 int j; 2861 2862 sbi->write_io[i] = 2863 f2fs_kmalloc(sbi, 2864 array_size(n, 2865 sizeof(struct f2fs_bio_info)), 2866 GFP_KERNEL); 2867 if (!sbi->write_io[i]) { 2868 err = -ENOMEM; 2869 goto free_options; 2870 } 2871 2872 for (j = HOT; j < n; j++) { 2873 init_rwsem(&sbi->write_io[i][j].io_rwsem); 2874 sbi->write_io[i][j].sbi = sbi; 2875 sbi->write_io[i][j].bio = NULL; 2876 spin_lock_init(&sbi->write_io[i][j].io_lock); 2877 INIT_LIST_HEAD(&sbi->write_io[i][j].io_list); 2878 } 2879 } 2880 2881 init_rwsem(&sbi->cp_rwsem); 2882 init_waitqueue_head(&sbi->cp_wait); 2883 init_sb_info(sbi); 2884 2885 err = init_percpu_info(sbi); 2886 if (err) 2887 goto free_bio_info; 2888 2889 if (F2FS_IO_SIZE(sbi) > 1) { 2890 sbi->write_io_dummy = 2891 mempool_create_page_pool(2 * (F2FS_IO_SIZE(sbi) - 1), 0); 2892 if (!sbi->write_io_dummy) { 2893 err = -ENOMEM; 2894 goto free_percpu; 2895 } 2896 } 2897 2898 /* get an inode for meta space */ 2899 sbi->meta_inode = f2fs_iget(sb, F2FS_META_INO(sbi)); 2900 if (IS_ERR(sbi->meta_inode)) { 2901 f2fs_msg(sb, KERN_ERR, "Failed to read F2FS meta data inode"); 2902 err = PTR_ERR(sbi->meta_inode); 2903 goto free_io_dummy; 2904 } 2905 2906 err = f2fs_get_valid_checkpoint(sbi); 2907 if (err) { 2908 f2fs_msg(sb, KERN_ERR, "Failed to get valid F2FS checkpoint"); 2909 goto free_meta_inode; 2910 } 2911 2912 /* Initialize device list */ 2913 err = f2fs_scan_devices(sbi); 2914 if (err) { 2915 f2fs_msg(sb, KERN_ERR, "Failed to find devices"); 2916 goto free_devices; 2917 } 2918 2919 sbi->total_valid_node_count = 2920 le32_to_cpu(sbi->ckpt->valid_node_count); 2921 percpu_counter_set(&sbi->total_valid_inode_count, 2922 le32_to_cpu(sbi->ckpt->valid_inode_count)); 2923 sbi->user_block_count = le64_to_cpu(sbi->ckpt->user_block_count); 2924 sbi->total_valid_block_count = 2925 le64_to_cpu(sbi->ckpt->valid_block_count); 2926 sbi->last_valid_block_count = sbi->total_valid_block_count; 2927 sbi->reserved_blocks = 0; 2928 sbi->current_reserved_blocks = 0; 2929 limit_reserve_root(sbi); 2930 2931 for (i = 0; i < NR_INODE_TYPE; i++) { 2932 INIT_LIST_HEAD(&sbi->inode_list[i]); 2933 spin_lock_init(&sbi->inode_lock[i]); 2934 } 2935 2936 f2fs_init_extent_cache_info(sbi); 2937 2938 f2fs_init_ino_entry_info(sbi); 2939 2940 f2fs_init_fsync_node_info(sbi); 2941 2942 /* setup f2fs internal modules */ 2943 err = f2fs_build_segment_manager(sbi); 2944 if (err) { 2945 f2fs_msg(sb, KERN_ERR, 2946 "Failed to initialize F2FS segment manager"); 2947 goto free_sm; 2948 } 2949 err = f2fs_build_node_manager(sbi); 2950 if (err) { 2951 f2fs_msg(sb, KERN_ERR, 2952 "Failed to initialize F2FS node manager"); 2953 goto free_nm; 2954 } 2955 2956 /* For write statistics */ 2957 if (sb->s_bdev->bd_part) 2958 sbi->sectors_written_start = 2959 (u64)part_stat_read(sb->s_bdev->bd_part, 2960 sectors[STAT_WRITE]); 2961 2962 /* Read accumulated write IO statistics if exists */ 2963 seg_i = CURSEG_I(sbi, CURSEG_HOT_NODE); 2964 if (__exist_node_summaries(sbi)) 2965 sbi->kbytes_written = 2966 le64_to_cpu(seg_i->journal->info.kbytes_written); 2967 2968 f2fs_build_gc_manager(sbi); 2969 2970 /* get an inode for node space */ 2971 sbi->node_inode = f2fs_iget(sb, F2FS_NODE_INO(sbi)); 2972 if (IS_ERR(sbi->node_inode)) { 2973 f2fs_msg(sb, KERN_ERR, "Failed to read node inode"); 2974 err = PTR_ERR(sbi->node_inode); 2975 goto free_nm; 2976 } 2977 2978 err = f2fs_build_stats(sbi); 2979 if (err) 2980 goto free_node_inode; 2981 2982 /* read root inode and dentry */ 2983 root = f2fs_iget(sb, F2FS_ROOT_INO(sbi)); 2984 if (IS_ERR(root)) { 2985 f2fs_msg(sb, KERN_ERR, "Failed to read root inode"); 2986 err = PTR_ERR(root); 2987 goto free_stats; 2988 } 2989 if (!S_ISDIR(root->i_mode) || !root->i_blocks || 2990 !root->i_size || !root->i_nlink) { 2991 iput(root); 2992 err = -EINVAL; 2993 goto free_stats; 2994 } 2995 2996 sb->s_root = d_make_root(root); /* allocate root dentry */ 2997 if (!sb->s_root) { 2998 err = -ENOMEM; 2999 goto free_root_inode; 3000 } 3001 3002 err = f2fs_register_sysfs(sbi); 3003 if (err) 3004 goto free_root_inode; 3005 3006 #ifdef CONFIG_QUOTA 3007 /* Enable quota usage during mount */ 3008 if (f2fs_sb_has_quota_ino(sb) && !f2fs_readonly(sb)) { 3009 err = f2fs_enable_quotas(sb); 3010 if (err) { 3011 f2fs_msg(sb, KERN_ERR, 3012 "Cannot turn on quotas: error %d", err); 3013 goto free_sysfs; 3014 } 3015 } 3016 #endif 3017 /* if there are nt orphan nodes free them */ 3018 err = f2fs_recover_orphan_inodes(sbi); 3019 if (err) 3020 goto free_meta; 3021 3022 /* recover fsynced data */ 3023 if (!test_opt(sbi, DISABLE_ROLL_FORWARD)) { 3024 /* 3025 * mount should be failed, when device has readonly mode, and 3026 * previous checkpoint was not done by clean system shutdown. 3027 */ 3028 if (bdev_read_only(sb->s_bdev) && 3029 !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG)) { 3030 err = -EROFS; 3031 goto free_meta; 3032 } 3033 3034 if (need_fsck) 3035 set_sbi_flag(sbi, SBI_NEED_FSCK); 3036 3037 if (!retry) 3038 goto skip_recovery; 3039 3040 err = f2fs_recover_fsync_data(sbi, false); 3041 if (err < 0) { 3042 need_fsck = true; 3043 f2fs_msg(sb, KERN_ERR, 3044 "Cannot recover all fsync data errno=%d", err); 3045 goto free_meta; 3046 } 3047 } else { 3048 err = f2fs_recover_fsync_data(sbi, true); 3049 3050 if (!f2fs_readonly(sb) && err > 0) { 3051 err = -EINVAL; 3052 f2fs_msg(sb, KERN_ERR, 3053 "Need to recover fsync data"); 3054 goto free_meta; 3055 } 3056 } 3057 skip_recovery: 3058 /* f2fs_recover_fsync_data() cleared this already */ 3059 clear_sbi_flag(sbi, SBI_POR_DOING); 3060 3061 /* 3062 * If filesystem is not mounted as read-only then 3063 * do start the gc_thread. 3064 */ 3065 if (test_opt(sbi, BG_GC) && !f2fs_readonly(sb)) { 3066 /* After POR, we can run background GC thread.*/ 3067 err = f2fs_start_gc_thread(sbi); 3068 if (err) 3069 goto free_meta; 3070 } 3071 kfree(options); 3072 3073 /* recover broken superblock */ 3074 if (recovery) { 3075 err = f2fs_commit_super(sbi, true); 3076 f2fs_msg(sb, KERN_INFO, 3077 "Try to recover %dth superblock, ret: %d", 3078 sbi->valid_super_block ? 1 : 2, err); 3079 } 3080 3081 f2fs_join_shrinker(sbi); 3082 3083 f2fs_tuning_parameters(sbi); 3084 3085 f2fs_msg(sbi->sb, KERN_NOTICE, "Mounted with checkpoint version = %llx", 3086 cur_cp_version(F2FS_CKPT(sbi))); 3087 f2fs_update_time(sbi, CP_TIME); 3088 f2fs_update_time(sbi, REQ_TIME); 3089 return 0; 3090 3091 free_meta: 3092 #ifdef CONFIG_QUOTA 3093 if (f2fs_sb_has_quota_ino(sb) && !f2fs_readonly(sb)) 3094 f2fs_quota_off_umount(sbi->sb); 3095 #endif 3096 f2fs_sync_inode_meta(sbi); 3097 /* 3098 * Some dirty meta pages can be produced by f2fs_recover_orphan_inodes() 3099 * failed by EIO. Then, iput(node_inode) can trigger balance_fs_bg() 3100 * followed by f2fs_write_checkpoint() through f2fs_write_node_pages(), which 3101 * falls into an infinite loop in f2fs_sync_meta_pages(). 3102 */ 3103 truncate_inode_pages_final(META_MAPPING(sbi)); 3104 #ifdef CONFIG_QUOTA 3105 free_sysfs: 3106 #endif 3107 f2fs_unregister_sysfs(sbi); 3108 free_root_inode: 3109 dput(sb->s_root); 3110 sb->s_root = NULL; 3111 free_stats: 3112 f2fs_destroy_stats(sbi); 3113 free_node_inode: 3114 f2fs_release_ino_entry(sbi, true); 3115 truncate_inode_pages_final(NODE_MAPPING(sbi)); 3116 iput(sbi->node_inode); 3117 free_nm: 3118 f2fs_destroy_node_manager(sbi); 3119 free_sm: 3120 f2fs_destroy_segment_manager(sbi); 3121 free_devices: 3122 destroy_device_list(sbi); 3123 kfree(sbi->ckpt); 3124 free_meta_inode: 3125 make_bad_inode(sbi->meta_inode); 3126 iput(sbi->meta_inode); 3127 free_io_dummy: 3128 mempool_destroy(sbi->write_io_dummy); 3129 free_percpu: 3130 destroy_percpu_info(sbi); 3131 free_bio_info: 3132 for (i = 0; i < NR_PAGE_TYPE; i++) 3133 kfree(sbi->write_io[i]); 3134 free_options: 3135 #ifdef CONFIG_QUOTA 3136 for (i = 0; i < MAXQUOTAS; i++) 3137 kfree(F2FS_OPTION(sbi).s_qf_names[i]); 3138 #endif 3139 kfree(options); 3140 free_sb_buf: 3141 kfree(raw_super); 3142 free_sbi: 3143 if (sbi->s_chksum_driver) 3144 crypto_free_shash(sbi->s_chksum_driver); 3145 kfree(sbi); 3146 3147 /* give only one another chance */ 3148 if (retry) { 3149 retry = false; 3150 shrink_dcache_sb(sb); 3151 goto try_onemore; 3152 } 3153 return err; 3154 } 3155 3156 static struct dentry *f2fs_mount(struct file_system_type *fs_type, int flags, 3157 const char *dev_name, void *data) 3158 { 3159 return mount_bdev(fs_type, flags, dev_name, data, f2fs_fill_super); 3160 } 3161 3162 static void kill_f2fs_super(struct super_block *sb) 3163 { 3164 if (sb->s_root) { 3165 struct f2fs_sb_info *sbi = F2FS_SB(sb); 3166 3167 set_sbi_flag(sbi, SBI_IS_CLOSE); 3168 f2fs_stop_gc_thread(sbi); 3169 f2fs_stop_discard_thread(sbi); 3170 3171 if (is_sbi_flag_set(sbi, SBI_IS_DIRTY) || 3172 !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG)) { 3173 struct cp_control cpc = { 3174 .reason = CP_UMOUNT, 3175 }; 3176 f2fs_write_checkpoint(sbi, &cpc); 3177 } 3178 } 3179 kill_block_super(sb); 3180 } 3181 3182 static struct file_system_type f2fs_fs_type = { 3183 .owner = THIS_MODULE, 3184 .name = "f2fs", 3185 .mount = f2fs_mount, 3186 .kill_sb = kill_f2fs_super, 3187 .fs_flags = FS_REQUIRES_DEV, 3188 }; 3189 MODULE_ALIAS_FS("f2fs"); 3190 3191 static int __init init_inodecache(void) 3192 { 3193 f2fs_inode_cachep = kmem_cache_create("f2fs_inode_cache", 3194 sizeof(struct f2fs_inode_info), 0, 3195 SLAB_RECLAIM_ACCOUNT|SLAB_ACCOUNT, NULL); 3196 if (!f2fs_inode_cachep) 3197 return -ENOMEM; 3198 return 0; 3199 } 3200 3201 static void destroy_inodecache(void) 3202 { 3203 /* 3204 * Make sure all delayed rcu free inodes are flushed before we 3205 * destroy cache. 3206 */ 3207 rcu_barrier(); 3208 kmem_cache_destroy(f2fs_inode_cachep); 3209 } 3210 3211 static int __init init_f2fs_fs(void) 3212 { 3213 int err; 3214 3215 if (PAGE_SIZE != F2FS_BLKSIZE) { 3216 printk("F2FS not supported on PAGE_SIZE(%lu) != %d\n", 3217 PAGE_SIZE, F2FS_BLKSIZE); 3218 return -EINVAL; 3219 } 3220 3221 f2fs_build_trace_ios(); 3222 3223 err = init_inodecache(); 3224 if (err) 3225 goto fail; 3226 err = f2fs_create_node_manager_caches(); 3227 if (err) 3228 goto free_inodecache; 3229 err = f2fs_create_segment_manager_caches(); 3230 if (err) 3231 goto free_node_manager_caches; 3232 err = f2fs_create_checkpoint_caches(); 3233 if (err) 3234 goto free_segment_manager_caches; 3235 err = f2fs_create_extent_cache(); 3236 if (err) 3237 goto free_checkpoint_caches; 3238 err = f2fs_init_sysfs(); 3239 if (err) 3240 goto free_extent_cache; 3241 err = register_shrinker(&f2fs_shrinker_info); 3242 if (err) 3243 goto free_sysfs; 3244 err = register_filesystem(&f2fs_fs_type); 3245 if (err) 3246 goto free_shrinker; 3247 err = f2fs_create_root_stats(); 3248 if (err) 3249 goto free_filesystem; 3250 err = f2fs_init_post_read_processing(); 3251 if (err) 3252 goto free_root_stats; 3253 return 0; 3254 3255 free_root_stats: 3256 f2fs_destroy_root_stats(); 3257 free_filesystem: 3258 unregister_filesystem(&f2fs_fs_type); 3259 free_shrinker: 3260 unregister_shrinker(&f2fs_shrinker_info); 3261 free_sysfs: 3262 f2fs_exit_sysfs(); 3263 free_extent_cache: 3264 f2fs_destroy_extent_cache(); 3265 free_checkpoint_caches: 3266 f2fs_destroy_checkpoint_caches(); 3267 free_segment_manager_caches: 3268 f2fs_destroy_segment_manager_caches(); 3269 free_node_manager_caches: 3270 f2fs_destroy_node_manager_caches(); 3271 free_inodecache: 3272 destroy_inodecache(); 3273 fail: 3274 return err; 3275 } 3276 3277 static void __exit exit_f2fs_fs(void) 3278 { 3279 f2fs_destroy_post_read_processing(); 3280 f2fs_destroy_root_stats(); 3281 unregister_filesystem(&f2fs_fs_type); 3282 unregister_shrinker(&f2fs_shrinker_info); 3283 f2fs_exit_sysfs(); 3284 f2fs_destroy_extent_cache(); 3285 f2fs_destroy_checkpoint_caches(); 3286 f2fs_destroy_segment_manager_caches(); 3287 f2fs_destroy_node_manager_caches(); 3288 destroy_inodecache(); 3289 f2fs_destroy_trace_ios(); 3290 } 3291 3292 module_init(init_f2fs_fs) 3293 module_exit(exit_f2fs_fs) 3294 3295 MODULE_AUTHOR("Samsung Electronics's Praesto Team"); 3296 MODULE_DESCRIPTION("Flash Friendly File System"); 3297 MODULE_LICENSE("GPL"); 3298 3299