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 *fault_name[FAULT_MAX] = { 45 [FAULT_KMALLOC] = "kmalloc", 46 [FAULT_PAGE_ALLOC] = "page alloc", 47 [FAULT_PAGE_GET] = "page get", 48 [FAULT_ALLOC_BIO] = "alloc bio", 49 [FAULT_ALLOC_NID] = "alloc nid", 50 [FAULT_ORPHAN] = "orphan", 51 [FAULT_BLOCK] = "no more block", 52 [FAULT_DIR_DEPTH] = "too big dir depth", 53 [FAULT_EVICT_INODE] = "evict_inode fail", 54 [FAULT_TRUNCATE] = "truncate fail", 55 [FAULT_IO] = "IO error", 56 [FAULT_CHECKPOINT] = "checkpoint error", 57 }; 58 59 static void f2fs_build_fault_attr(struct f2fs_sb_info *sbi, 60 unsigned int rate) 61 { 62 struct f2fs_fault_info *ffi = &sbi->fault_info; 63 64 if (rate) { 65 atomic_set(&ffi->inject_ops, 0); 66 ffi->inject_rate = rate; 67 ffi->inject_type = (1 << FAULT_MAX) - 1; 68 } else { 69 memset(ffi, 0, sizeof(struct f2fs_fault_info)); 70 } 71 } 72 #endif 73 74 /* f2fs-wide shrinker description */ 75 static struct shrinker f2fs_shrinker_info = { 76 .scan_objects = f2fs_shrink_scan, 77 .count_objects = f2fs_shrink_count, 78 .seeks = DEFAULT_SEEKS, 79 }; 80 81 enum { 82 Opt_gc_background, 83 Opt_disable_roll_forward, 84 Opt_norecovery, 85 Opt_discard, 86 Opt_nodiscard, 87 Opt_noheap, 88 Opt_heap, 89 Opt_user_xattr, 90 Opt_nouser_xattr, 91 Opt_acl, 92 Opt_noacl, 93 Opt_active_logs, 94 Opt_disable_ext_identify, 95 Opt_inline_xattr, 96 Opt_noinline_xattr, 97 Opt_inline_xattr_size, 98 Opt_inline_data, 99 Opt_inline_dentry, 100 Opt_noinline_dentry, 101 Opt_flush_merge, 102 Opt_noflush_merge, 103 Opt_nobarrier, 104 Opt_fastboot, 105 Opt_extent_cache, 106 Opt_noextent_cache, 107 Opt_noinline_data, 108 Opt_data_flush, 109 Opt_mode, 110 Opt_io_size_bits, 111 Opt_fault_injection, 112 Opt_lazytime, 113 Opt_nolazytime, 114 Opt_quota, 115 Opt_noquota, 116 Opt_usrquota, 117 Opt_grpquota, 118 Opt_prjquota, 119 Opt_usrjquota, 120 Opt_grpjquota, 121 Opt_prjjquota, 122 Opt_offusrjquota, 123 Opt_offgrpjquota, 124 Opt_offprjjquota, 125 Opt_jqfmt_vfsold, 126 Opt_jqfmt_vfsv0, 127 Opt_jqfmt_vfsv1, 128 Opt_err, 129 }; 130 131 static match_table_t f2fs_tokens = { 132 {Opt_gc_background, "background_gc=%s"}, 133 {Opt_disable_roll_forward, "disable_roll_forward"}, 134 {Opt_norecovery, "norecovery"}, 135 {Opt_discard, "discard"}, 136 {Opt_nodiscard, "nodiscard"}, 137 {Opt_noheap, "no_heap"}, 138 {Opt_heap, "heap"}, 139 {Opt_user_xattr, "user_xattr"}, 140 {Opt_nouser_xattr, "nouser_xattr"}, 141 {Opt_acl, "acl"}, 142 {Opt_noacl, "noacl"}, 143 {Opt_active_logs, "active_logs=%u"}, 144 {Opt_disable_ext_identify, "disable_ext_identify"}, 145 {Opt_inline_xattr, "inline_xattr"}, 146 {Opt_noinline_xattr, "noinline_xattr"}, 147 {Opt_inline_xattr_size, "inline_xattr_size=%u"}, 148 {Opt_inline_data, "inline_data"}, 149 {Opt_inline_dentry, "inline_dentry"}, 150 {Opt_noinline_dentry, "noinline_dentry"}, 151 {Opt_flush_merge, "flush_merge"}, 152 {Opt_noflush_merge, "noflush_merge"}, 153 {Opt_nobarrier, "nobarrier"}, 154 {Opt_fastboot, "fastboot"}, 155 {Opt_extent_cache, "extent_cache"}, 156 {Opt_noextent_cache, "noextent_cache"}, 157 {Opt_noinline_data, "noinline_data"}, 158 {Opt_data_flush, "data_flush"}, 159 {Opt_mode, "mode=%s"}, 160 {Opt_io_size_bits, "io_bits=%u"}, 161 {Opt_fault_injection, "fault_injection=%u"}, 162 {Opt_lazytime, "lazytime"}, 163 {Opt_nolazytime, "nolazytime"}, 164 {Opt_quota, "quota"}, 165 {Opt_noquota, "noquota"}, 166 {Opt_usrquota, "usrquota"}, 167 {Opt_grpquota, "grpquota"}, 168 {Opt_prjquota, "prjquota"}, 169 {Opt_usrjquota, "usrjquota=%s"}, 170 {Opt_grpjquota, "grpjquota=%s"}, 171 {Opt_prjjquota, "prjjquota=%s"}, 172 {Opt_offusrjquota, "usrjquota="}, 173 {Opt_offgrpjquota, "grpjquota="}, 174 {Opt_offprjjquota, "prjjquota="}, 175 {Opt_jqfmt_vfsold, "jqfmt=vfsold"}, 176 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"}, 177 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"}, 178 {Opt_err, NULL}, 179 }; 180 181 void f2fs_msg(struct super_block *sb, const char *level, const char *fmt, ...) 182 { 183 struct va_format vaf; 184 va_list args; 185 186 va_start(args, fmt); 187 vaf.fmt = fmt; 188 vaf.va = &args; 189 printk_ratelimited("%sF2FS-fs (%s): %pV\n", level, sb->s_id, &vaf); 190 va_end(args); 191 } 192 193 static void init_once(void *foo) 194 { 195 struct f2fs_inode_info *fi = (struct f2fs_inode_info *) foo; 196 197 inode_init_once(&fi->vfs_inode); 198 } 199 200 #ifdef CONFIG_QUOTA 201 static const char * const quotatypes[] = INITQFNAMES; 202 #define QTYPE2NAME(t) (quotatypes[t]) 203 static int f2fs_set_qf_name(struct super_block *sb, int qtype, 204 substring_t *args) 205 { 206 struct f2fs_sb_info *sbi = F2FS_SB(sb); 207 char *qname; 208 int ret = -EINVAL; 209 210 if (sb_any_quota_loaded(sb) && !sbi->s_qf_names[qtype]) { 211 f2fs_msg(sb, KERN_ERR, 212 "Cannot change journaled " 213 "quota options when quota turned on"); 214 return -EINVAL; 215 } 216 if (f2fs_sb_has_quota_ino(sb)) { 217 f2fs_msg(sb, KERN_INFO, 218 "QUOTA feature is enabled, so ignore qf_name"); 219 return 0; 220 } 221 222 qname = match_strdup(args); 223 if (!qname) { 224 f2fs_msg(sb, KERN_ERR, 225 "Not enough memory for storing quotafile name"); 226 return -EINVAL; 227 } 228 if (sbi->s_qf_names[qtype]) { 229 if (strcmp(sbi->s_qf_names[qtype], qname) == 0) 230 ret = 0; 231 else 232 f2fs_msg(sb, KERN_ERR, 233 "%s quota file already specified", 234 QTYPE2NAME(qtype)); 235 goto errout; 236 } 237 if (strchr(qname, '/')) { 238 f2fs_msg(sb, KERN_ERR, 239 "quotafile must be on filesystem root"); 240 goto errout; 241 } 242 sbi->s_qf_names[qtype] = qname; 243 set_opt(sbi, QUOTA); 244 return 0; 245 errout: 246 kfree(qname); 247 return ret; 248 } 249 250 static int f2fs_clear_qf_name(struct super_block *sb, int qtype) 251 { 252 struct f2fs_sb_info *sbi = F2FS_SB(sb); 253 254 if (sb_any_quota_loaded(sb) && sbi->s_qf_names[qtype]) { 255 f2fs_msg(sb, KERN_ERR, "Cannot change journaled quota options" 256 " when quota turned on"); 257 return -EINVAL; 258 } 259 kfree(sbi->s_qf_names[qtype]); 260 sbi->s_qf_names[qtype] = NULL; 261 return 0; 262 } 263 264 static int f2fs_check_quota_options(struct f2fs_sb_info *sbi) 265 { 266 /* 267 * We do the test below only for project quotas. 'usrquota' and 268 * 'grpquota' mount options are allowed even without quota feature 269 * to support legacy quotas in quota files. 270 */ 271 if (test_opt(sbi, PRJQUOTA) && !f2fs_sb_has_project_quota(sbi->sb)) { 272 f2fs_msg(sbi->sb, KERN_ERR, "Project quota feature not enabled. " 273 "Cannot enable project quota enforcement."); 274 return -1; 275 } 276 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA] || 277 sbi->s_qf_names[PRJQUOTA]) { 278 if (test_opt(sbi, USRQUOTA) && sbi->s_qf_names[USRQUOTA]) 279 clear_opt(sbi, USRQUOTA); 280 281 if (test_opt(sbi, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA]) 282 clear_opt(sbi, GRPQUOTA); 283 284 if (test_opt(sbi, PRJQUOTA) && sbi->s_qf_names[PRJQUOTA]) 285 clear_opt(sbi, PRJQUOTA); 286 287 if (test_opt(sbi, GRPQUOTA) || test_opt(sbi, USRQUOTA) || 288 test_opt(sbi, PRJQUOTA)) { 289 f2fs_msg(sbi->sb, KERN_ERR, "old and new quota " 290 "format mixing"); 291 return -1; 292 } 293 294 if (!sbi->s_jquota_fmt) { 295 f2fs_msg(sbi->sb, KERN_ERR, "journaled quota format " 296 "not specified"); 297 return -1; 298 } 299 } 300 301 if (f2fs_sb_has_quota_ino(sbi->sb) && sbi->s_jquota_fmt) { 302 f2fs_msg(sbi->sb, KERN_INFO, 303 "QUOTA feature is enabled, so ignore jquota_fmt"); 304 sbi->s_jquota_fmt = 0; 305 } 306 if (f2fs_sb_has_quota_ino(sbi->sb) && sb_rdonly(sbi->sb)) { 307 f2fs_msg(sbi->sb, KERN_INFO, 308 "Filesystem with quota feature cannot be mounted RDWR " 309 "without CONFIG_QUOTA"); 310 return -1; 311 } 312 return 0; 313 } 314 #endif 315 316 static int parse_options(struct super_block *sb, char *options) 317 { 318 struct f2fs_sb_info *sbi = F2FS_SB(sb); 319 struct request_queue *q; 320 substring_t args[MAX_OPT_ARGS]; 321 char *p, *name; 322 int arg = 0; 323 #ifdef CONFIG_QUOTA 324 int ret; 325 #endif 326 327 if (!options) 328 return 0; 329 330 while ((p = strsep(&options, ",")) != NULL) { 331 int token; 332 if (!*p) 333 continue; 334 /* 335 * Initialize args struct so we know whether arg was 336 * found; some options take optional arguments. 337 */ 338 args[0].to = args[0].from = NULL; 339 token = match_token(p, f2fs_tokens, args); 340 341 switch (token) { 342 case Opt_gc_background: 343 name = match_strdup(&args[0]); 344 345 if (!name) 346 return -ENOMEM; 347 if (strlen(name) == 2 && !strncmp(name, "on", 2)) { 348 set_opt(sbi, BG_GC); 349 clear_opt(sbi, FORCE_FG_GC); 350 } else if (strlen(name) == 3 && !strncmp(name, "off", 3)) { 351 clear_opt(sbi, BG_GC); 352 clear_opt(sbi, FORCE_FG_GC); 353 } else if (strlen(name) == 4 && !strncmp(name, "sync", 4)) { 354 set_opt(sbi, BG_GC); 355 set_opt(sbi, FORCE_FG_GC); 356 } else { 357 kfree(name); 358 return -EINVAL; 359 } 360 kfree(name); 361 break; 362 case Opt_disable_roll_forward: 363 set_opt(sbi, DISABLE_ROLL_FORWARD); 364 break; 365 case Opt_norecovery: 366 /* this option mounts f2fs with ro */ 367 set_opt(sbi, DISABLE_ROLL_FORWARD); 368 if (!f2fs_readonly(sb)) 369 return -EINVAL; 370 break; 371 case Opt_discard: 372 q = bdev_get_queue(sb->s_bdev); 373 if (blk_queue_discard(q)) { 374 set_opt(sbi, DISCARD); 375 } else if (!f2fs_sb_mounted_blkzoned(sb)) { 376 f2fs_msg(sb, KERN_WARNING, 377 "mounting with \"discard\" option, but " 378 "the device does not support discard"); 379 } 380 break; 381 case Opt_nodiscard: 382 if (f2fs_sb_mounted_blkzoned(sb)) { 383 f2fs_msg(sb, KERN_WARNING, 384 "discard is required for zoned block devices"); 385 return -EINVAL; 386 } 387 clear_opt(sbi, DISCARD); 388 break; 389 case Opt_noheap: 390 set_opt(sbi, NOHEAP); 391 break; 392 case Opt_heap: 393 clear_opt(sbi, NOHEAP); 394 break; 395 #ifdef CONFIG_F2FS_FS_XATTR 396 case Opt_user_xattr: 397 set_opt(sbi, XATTR_USER); 398 break; 399 case Opt_nouser_xattr: 400 clear_opt(sbi, XATTR_USER); 401 break; 402 case Opt_inline_xattr: 403 set_opt(sbi, INLINE_XATTR); 404 break; 405 case Opt_noinline_xattr: 406 clear_opt(sbi, INLINE_XATTR); 407 break; 408 case Opt_inline_xattr_size: 409 if (args->from && match_int(args, &arg)) 410 return -EINVAL; 411 set_opt(sbi, INLINE_XATTR_SIZE); 412 sbi->inline_xattr_size = arg; 413 break; 414 #else 415 case Opt_user_xattr: 416 f2fs_msg(sb, KERN_INFO, 417 "user_xattr options not supported"); 418 break; 419 case Opt_nouser_xattr: 420 f2fs_msg(sb, KERN_INFO, 421 "nouser_xattr options not supported"); 422 break; 423 case Opt_inline_xattr: 424 f2fs_msg(sb, KERN_INFO, 425 "inline_xattr options not supported"); 426 break; 427 case Opt_noinline_xattr: 428 f2fs_msg(sb, KERN_INFO, 429 "noinline_xattr options not supported"); 430 break; 431 #endif 432 #ifdef CONFIG_F2FS_FS_POSIX_ACL 433 case Opt_acl: 434 set_opt(sbi, POSIX_ACL); 435 break; 436 case Opt_noacl: 437 clear_opt(sbi, POSIX_ACL); 438 break; 439 #else 440 case Opt_acl: 441 f2fs_msg(sb, KERN_INFO, "acl options not supported"); 442 break; 443 case Opt_noacl: 444 f2fs_msg(sb, KERN_INFO, "noacl options not supported"); 445 break; 446 #endif 447 case Opt_active_logs: 448 if (args->from && match_int(args, &arg)) 449 return -EINVAL; 450 if (arg != 2 && arg != 4 && arg != NR_CURSEG_TYPE) 451 return -EINVAL; 452 sbi->active_logs = arg; 453 break; 454 case Opt_disable_ext_identify: 455 set_opt(sbi, DISABLE_EXT_IDENTIFY); 456 break; 457 case Opt_inline_data: 458 set_opt(sbi, INLINE_DATA); 459 break; 460 case Opt_inline_dentry: 461 set_opt(sbi, INLINE_DENTRY); 462 break; 463 case Opt_noinline_dentry: 464 clear_opt(sbi, INLINE_DENTRY); 465 break; 466 case Opt_flush_merge: 467 set_opt(sbi, FLUSH_MERGE); 468 break; 469 case Opt_noflush_merge: 470 clear_opt(sbi, FLUSH_MERGE); 471 break; 472 case Opt_nobarrier: 473 set_opt(sbi, NOBARRIER); 474 break; 475 case Opt_fastboot: 476 set_opt(sbi, FASTBOOT); 477 break; 478 case Opt_extent_cache: 479 set_opt(sbi, EXTENT_CACHE); 480 break; 481 case Opt_noextent_cache: 482 clear_opt(sbi, EXTENT_CACHE); 483 break; 484 case Opt_noinline_data: 485 clear_opt(sbi, INLINE_DATA); 486 break; 487 case Opt_data_flush: 488 set_opt(sbi, DATA_FLUSH); 489 break; 490 case Opt_mode: 491 name = match_strdup(&args[0]); 492 493 if (!name) 494 return -ENOMEM; 495 if (strlen(name) == 8 && 496 !strncmp(name, "adaptive", 8)) { 497 if (f2fs_sb_mounted_blkzoned(sb)) { 498 f2fs_msg(sb, KERN_WARNING, 499 "adaptive mode is not allowed with " 500 "zoned block device feature"); 501 kfree(name); 502 return -EINVAL; 503 } 504 set_opt_mode(sbi, F2FS_MOUNT_ADAPTIVE); 505 } else if (strlen(name) == 3 && 506 !strncmp(name, "lfs", 3)) { 507 set_opt_mode(sbi, F2FS_MOUNT_LFS); 508 } else { 509 kfree(name); 510 return -EINVAL; 511 } 512 kfree(name); 513 break; 514 case Opt_io_size_bits: 515 if (args->from && match_int(args, &arg)) 516 return -EINVAL; 517 if (arg > __ilog2_u32(BIO_MAX_PAGES)) { 518 f2fs_msg(sb, KERN_WARNING, 519 "Not support %d, larger than %d", 520 1 << arg, BIO_MAX_PAGES); 521 return -EINVAL; 522 } 523 sbi->write_io_size_bits = arg; 524 break; 525 case Opt_fault_injection: 526 if (args->from && match_int(args, &arg)) 527 return -EINVAL; 528 #ifdef CONFIG_F2FS_FAULT_INJECTION 529 f2fs_build_fault_attr(sbi, arg); 530 set_opt(sbi, FAULT_INJECTION); 531 #else 532 f2fs_msg(sb, KERN_INFO, 533 "FAULT_INJECTION was not selected"); 534 #endif 535 break; 536 case Opt_lazytime: 537 sb->s_flags |= SB_LAZYTIME; 538 break; 539 case Opt_nolazytime: 540 sb->s_flags &= ~SB_LAZYTIME; 541 break; 542 #ifdef CONFIG_QUOTA 543 case Opt_quota: 544 case Opt_usrquota: 545 set_opt(sbi, USRQUOTA); 546 break; 547 case Opt_grpquota: 548 set_opt(sbi, GRPQUOTA); 549 break; 550 case Opt_prjquota: 551 set_opt(sbi, PRJQUOTA); 552 break; 553 case Opt_usrjquota: 554 ret = f2fs_set_qf_name(sb, USRQUOTA, &args[0]); 555 if (ret) 556 return ret; 557 break; 558 case Opt_grpjquota: 559 ret = f2fs_set_qf_name(sb, GRPQUOTA, &args[0]); 560 if (ret) 561 return ret; 562 break; 563 case Opt_prjjquota: 564 ret = f2fs_set_qf_name(sb, PRJQUOTA, &args[0]); 565 if (ret) 566 return ret; 567 break; 568 case Opt_offusrjquota: 569 ret = f2fs_clear_qf_name(sb, USRQUOTA); 570 if (ret) 571 return ret; 572 break; 573 case Opt_offgrpjquota: 574 ret = f2fs_clear_qf_name(sb, GRPQUOTA); 575 if (ret) 576 return ret; 577 break; 578 case Opt_offprjjquota: 579 ret = f2fs_clear_qf_name(sb, PRJQUOTA); 580 if (ret) 581 return ret; 582 break; 583 case Opt_jqfmt_vfsold: 584 sbi->s_jquota_fmt = QFMT_VFS_OLD; 585 break; 586 case Opt_jqfmt_vfsv0: 587 sbi->s_jquota_fmt = QFMT_VFS_V0; 588 break; 589 case Opt_jqfmt_vfsv1: 590 sbi->s_jquota_fmt = QFMT_VFS_V1; 591 break; 592 case Opt_noquota: 593 clear_opt(sbi, QUOTA); 594 clear_opt(sbi, USRQUOTA); 595 clear_opt(sbi, GRPQUOTA); 596 clear_opt(sbi, PRJQUOTA); 597 break; 598 #else 599 case Opt_quota: 600 case Opt_usrquota: 601 case Opt_grpquota: 602 case Opt_prjquota: 603 case Opt_usrjquota: 604 case Opt_grpjquota: 605 case Opt_prjjquota: 606 case Opt_offusrjquota: 607 case Opt_offgrpjquota: 608 case Opt_offprjjquota: 609 case Opt_jqfmt_vfsold: 610 case Opt_jqfmt_vfsv0: 611 case Opt_jqfmt_vfsv1: 612 case Opt_noquota: 613 f2fs_msg(sb, KERN_INFO, 614 "quota operations not supported"); 615 break; 616 #endif 617 default: 618 f2fs_msg(sb, KERN_ERR, 619 "Unrecognized mount option \"%s\" or missing value", 620 p); 621 return -EINVAL; 622 } 623 } 624 #ifdef CONFIG_QUOTA 625 if (f2fs_check_quota_options(sbi)) 626 return -EINVAL; 627 #endif 628 629 if (F2FS_IO_SIZE_BITS(sbi) && !test_opt(sbi, LFS)) { 630 f2fs_msg(sb, KERN_ERR, 631 "Should set mode=lfs with %uKB-sized IO", 632 F2FS_IO_SIZE_KB(sbi)); 633 return -EINVAL; 634 } 635 636 if (test_opt(sbi, INLINE_XATTR_SIZE)) { 637 if (!test_opt(sbi, INLINE_XATTR)) { 638 f2fs_msg(sb, KERN_ERR, 639 "inline_xattr_size option should be " 640 "set with inline_xattr option"); 641 return -EINVAL; 642 } 643 if (!sbi->inline_xattr_size || 644 sbi->inline_xattr_size >= DEF_ADDRS_PER_INODE - 645 F2FS_TOTAL_EXTRA_ATTR_SIZE - 646 DEF_INLINE_RESERVED_SIZE - 647 DEF_MIN_INLINE_SIZE) { 648 f2fs_msg(sb, KERN_ERR, 649 "inline xattr size is out of range"); 650 return -EINVAL; 651 } 652 } 653 return 0; 654 } 655 656 static struct inode *f2fs_alloc_inode(struct super_block *sb) 657 { 658 struct f2fs_inode_info *fi; 659 660 fi = kmem_cache_alloc(f2fs_inode_cachep, GFP_F2FS_ZERO); 661 if (!fi) 662 return NULL; 663 664 init_once((void *) fi); 665 666 /* Initialize f2fs-specific inode info */ 667 atomic_set(&fi->dirty_pages, 0); 668 fi->i_current_depth = 1; 669 fi->i_advise = 0; 670 init_rwsem(&fi->i_sem); 671 INIT_LIST_HEAD(&fi->dirty_list); 672 INIT_LIST_HEAD(&fi->gdirty_list); 673 INIT_LIST_HEAD(&fi->inmem_ilist); 674 INIT_LIST_HEAD(&fi->inmem_pages); 675 mutex_init(&fi->inmem_lock); 676 init_rwsem(&fi->dio_rwsem[READ]); 677 init_rwsem(&fi->dio_rwsem[WRITE]); 678 init_rwsem(&fi->i_mmap_sem); 679 init_rwsem(&fi->i_xattr_sem); 680 681 #ifdef CONFIG_QUOTA 682 memset(&fi->i_dquot, 0, sizeof(fi->i_dquot)); 683 fi->i_reserved_quota = 0; 684 #endif 685 /* Will be used by directory only */ 686 fi->i_dir_level = F2FS_SB(sb)->dir_level; 687 688 return &fi->vfs_inode; 689 } 690 691 static int f2fs_drop_inode(struct inode *inode) 692 { 693 int ret; 694 /* 695 * This is to avoid a deadlock condition like below. 696 * writeback_single_inode(inode) 697 * - f2fs_write_data_page 698 * - f2fs_gc -> iput -> evict 699 * - inode_wait_for_writeback(inode) 700 */ 701 if ((!inode_unhashed(inode) && inode->i_state & I_SYNC)) { 702 if (!inode->i_nlink && !is_bad_inode(inode)) { 703 /* to avoid evict_inode call simultaneously */ 704 atomic_inc(&inode->i_count); 705 spin_unlock(&inode->i_lock); 706 707 /* some remained atomic pages should discarded */ 708 if (f2fs_is_atomic_file(inode)) 709 drop_inmem_pages(inode); 710 711 /* should remain fi->extent_tree for writepage */ 712 f2fs_destroy_extent_node(inode); 713 714 sb_start_intwrite(inode->i_sb); 715 f2fs_i_size_write(inode, 0); 716 717 if (F2FS_HAS_BLOCKS(inode)) 718 f2fs_truncate(inode); 719 720 sb_end_intwrite(inode->i_sb); 721 722 spin_lock(&inode->i_lock); 723 atomic_dec(&inode->i_count); 724 } 725 trace_f2fs_drop_inode(inode, 0); 726 return 0; 727 } 728 ret = generic_drop_inode(inode); 729 trace_f2fs_drop_inode(inode, ret); 730 return ret; 731 } 732 733 int f2fs_inode_dirtied(struct inode *inode, bool sync) 734 { 735 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 736 int ret = 0; 737 738 spin_lock(&sbi->inode_lock[DIRTY_META]); 739 if (is_inode_flag_set(inode, FI_DIRTY_INODE)) { 740 ret = 1; 741 } else { 742 set_inode_flag(inode, FI_DIRTY_INODE); 743 stat_inc_dirty_inode(sbi, DIRTY_META); 744 } 745 if (sync && list_empty(&F2FS_I(inode)->gdirty_list)) { 746 list_add_tail(&F2FS_I(inode)->gdirty_list, 747 &sbi->inode_list[DIRTY_META]); 748 inc_page_count(sbi, F2FS_DIRTY_IMETA); 749 } 750 spin_unlock(&sbi->inode_lock[DIRTY_META]); 751 return ret; 752 } 753 754 void f2fs_inode_synced(struct inode *inode) 755 { 756 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 757 758 spin_lock(&sbi->inode_lock[DIRTY_META]); 759 if (!is_inode_flag_set(inode, FI_DIRTY_INODE)) { 760 spin_unlock(&sbi->inode_lock[DIRTY_META]); 761 return; 762 } 763 if (!list_empty(&F2FS_I(inode)->gdirty_list)) { 764 list_del_init(&F2FS_I(inode)->gdirty_list); 765 dec_page_count(sbi, F2FS_DIRTY_IMETA); 766 } 767 clear_inode_flag(inode, FI_DIRTY_INODE); 768 clear_inode_flag(inode, FI_AUTO_RECOVER); 769 stat_dec_dirty_inode(F2FS_I_SB(inode), DIRTY_META); 770 spin_unlock(&sbi->inode_lock[DIRTY_META]); 771 } 772 773 /* 774 * f2fs_dirty_inode() is called from __mark_inode_dirty() 775 * 776 * We should call set_dirty_inode to write the dirty inode through write_inode. 777 */ 778 static void f2fs_dirty_inode(struct inode *inode, int flags) 779 { 780 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 781 782 if (inode->i_ino == F2FS_NODE_INO(sbi) || 783 inode->i_ino == F2FS_META_INO(sbi)) 784 return; 785 786 if (flags == I_DIRTY_TIME) 787 return; 788 789 if (is_inode_flag_set(inode, FI_AUTO_RECOVER)) 790 clear_inode_flag(inode, FI_AUTO_RECOVER); 791 792 f2fs_inode_dirtied(inode, false); 793 } 794 795 static void f2fs_i_callback(struct rcu_head *head) 796 { 797 struct inode *inode = container_of(head, struct inode, i_rcu); 798 kmem_cache_free(f2fs_inode_cachep, F2FS_I(inode)); 799 } 800 801 static void f2fs_destroy_inode(struct inode *inode) 802 { 803 call_rcu(&inode->i_rcu, f2fs_i_callback); 804 } 805 806 static void destroy_percpu_info(struct f2fs_sb_info *sbi) 807 { 808 percpu_counter_destroy(&sbi->alloc_valid_block_count); 809 percpu_counter_destroy(&sbi->total_valid_inode_count); 810 } 811 812 static void destroy_device_list(struct f2fs_sb_info *sbi) 813 { 814 int i; 815 816 for (i = 0; i < sbi->s_ndevs; i++) { 817 blkdev_put(FDEV(i).bdev, FMODE_EXCL); 818 #ifdef CONFIG_BLK_DEV_ZONED 819 kfree(FDEV(i).blkz_type); 820 #endif 821 } 822 kfree(sbi->devs); 823 } 824 825 static void f2fs_put_super(struct super_block *sb) 826 { 827 struct f2fs_sb_info *sbi = F2FS_SB(sb); 828 int i; 829 bool dropped; 830 831 f2fs_quota_off_umount(sb); 832 833 /* prevent remaining shrinker jobs */ 834 mutex_lock(&sbi->umount_mutex); 835 836 /* 837 * We don't need to do checkpoint when superblock is clean. 838 * But, the previous checkpoint was not done by umount, it needs to do 839 * clean checkpoint again. 840 */ 841 if (is_sbi_flag_set(sbi, SBI_IS_DIRTY) || 842 !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG)) { 843 struct cp_control cpc = { 844 .reason = CP_UMOUNT, 845 }; 846 write_checkpoint(sbi, &cpc); 847 } 848 849 /* be sure to wait for any on-going discard commands */ 850 dropped = f2fs_wait_discard_bios(sbi); 851 852 if (f2fs_discard_en(sbi) && !sbi->discard_blks && !dropped) { 853 struct cp_control cpc = { 854 .reason = CP_UMOUNT | CP_TRIMMED, 855 }; 856 write_checkpoint(sbi, &cpc); 857 } 858 859 /* write_checkpoint can update stat informaion */ 860 f2fs_destroy_stats(sbi); 861 862 /* 863 * normally superblock is clean, so we need to release this. 864 * In addition, EIO will skip do checkpoint, we need this as well. 865 */ 866 release_ino_entry(sbi, true); 867 868 f2fs_leave_shrinker(sbi); 869 mutex_unlock(&sbi->umount_mutex); 870 871 /* our cp_error case, we can wait for any writeback page */ 872 f2fs_flush_merged_writes(sbi); 873 874 iput(sbi->node_inode); 875 iput(sbi->meta_inode); 876 877 /* destroy f2fs internal modules */ 878 destroy_node_manager(sbi); 879 destroy_segment_manager(sbi); 880 881 kfree(sbi->ckpt); 882 883 f2fs_unregister_sysfs(sbi); 884 885 sb->s_fs_info = NULL; 886 if (sbi->s_chksum_driver) 887 crypto_free_shash(sbi->s_chksum_driver); 888 kfree(sbi->raw_super); 889 890 destroy_device_list(sbi); 891 mempool_destroy(sbi->write_io_dummy); 892 #ifdef CONFIG_QUOTA 893 for (i = 0; i < MAXQUOTAS; i++) 894 kfree(sbi->s_qf_names[i]); 895 #endif 896 destroy_percpu_info(sbi); 897 for (i = 0; i < NR_PAGE_TYPE; i++) 898 kfree(sbi->write_io[i]); 899 kfree(sbi); 900 } 901 902 int f2fs_sync_fs(struct super_block *sb, int sync) 903 { 904 struct f2fs_sb_info *sbi = F2FS_SB(sb); 905 int err = 0; 906 907 if (unlikely(f2fs_cp_error(sbi))) 908 return 0; 909 910 trace_f2fs_sync_fs(sb, sync); 911 912 if (unlikely(is_sbi_flag_set(sbi, SBI_POR_DOING))) 913 return -EAGAIN; 914 915 if (sync) { 916 struct cp_control cpc; 917 918 cpc.reason = __get_cp_reason(sbi); 919 920 mutex_lock(&sbi->gc_mutex); 921 err = write_checkpoint(sbi, &cpc); 922 mutex_unlock(&sbi->gc_mutex); 923 } 924 f2fs_trace_ios(NULL, 1); 925 926 return err; 927 } 928 929 static int f2fs_freeze(struct super_block *sb) 930 { 931 if (f2fs_readonly(sb)) 932 return 0; 933 934 /* IO error happened before */ 935 if (unlikely(f2fs_cp_error(F2FS_SB(sb)))) 936 return -EIO; 937 938 /* must be clean, since sync_filesystem() was already called */ 939 if (is_sbi_flag_set(F2FS_SB(sb), SBI_IS_DIRTY)) 940 return -EINVAL; 941 return 0; 942 } 943 944 static int f2fs_unfreeze(struct super_block *sb) 945 { 946 return 0; 947 } 948 949 #ifdef CONFIG_QUOTA 950 static int f2fs_statfs_project(struct super_block *sb, 951 kprojid_t projid, struct kstatfs *buf) 952 { 953 struct kqid qid; 954 struct dquot *dquot; 955 u64 limit; 956 u64 curblock; 957 958 qid = make_kqid_projid(projid); 959 dquot = dqget(sb, qid); 960 if (IS_ERR(dquot)) 961 return PTR_ERR(dquot); 962 spin_lock(&dq_data_lock); 963 964 limit = (dquot->dq_dqb.dqb_bsoftlimit ? 965 dquot->dq_dqb.dqb_bsoftlimit : 966 dquot->dq_dqb.dqb_bhardlimit) >> sb->s_blocksize_bits; 967 if (limit && buf->f_blocks > limit) { 968 curblock = dquot->dq_dqb.dqb_curspace >> sb->s_blocksize_bits; 969 buf->f_blocks = limit; 970 buf->f_bfree = buf->f_bavail = 971 (buf->f_blocks > curblock) ? 972 (buf->f_blocks - curblock) : 0; 973 } 974 975 limit = dquot->dq_dqb.dqb_isoftlimit ? 976 dquot->dq_dqb.dqb_isoftlimit : 977 dquot->dq_dqb.dqb_ihardlimit; 978 if (limit && buf->f_files > limit) { 979 buf->f_files = limit; 980 buf->f_ffree = 981 (buf->f_files > dquot->dq_dqb.dqb_curinodes) ? 982 (buf->f_files - dquot->dq_dqb.dqb_curinodes) : 0; 983 } 984 985 spin_unlock(&dq_data_lock); 986 dqput(dquot); 987 return 0; 988 } 989 #endif 990 991 static int f2fs_statfs(struct dentry *dentry, struct kstatfs *buf) 992 { 993 struct super_block *sb = dentry->d_sb; 994 struct f2fs_sb_info *sbi = F2FS_SB(sb); 995 u64 id = huge_encode_dev(sb->s_bdev->bd_dev); 996 block_t total_count, user_block_count, start_count, ovp_count; 997 u64 avail_node_count; 998 999 total_count = le64_to_cpu(sbi->raw_super->block_count); 1000 user_block_count = sbi->user_block_count; 1001 start_count = le32_to_cpu(sbi->raw_super->segment0_blkaddr); 1002 ovp_count = SM_I(sbi)->ovp_segments << sbi->log_blocks_per_seg; 1003 buf->f_type = F2FS_SUPER_MAGIC; 1004 buf->f_bsize = sbi->blocksize; 1005 1006 buf->f_blocks = total_count - start_count; 1007 buf->f_bfree = user_block_count - valid_user_blocks(sbi) + ovp_count; 1008 buf->f_bavail = user_block_count - valid_user_blocks(sbi) - 1009 sbi->current_reserved_blocks; 1010 1011 avail_node_count = sbi->total_node_count - F2FS_RESERVED_NODE_NUM; 1012 1013 if (avail_node_count > user_block_count) { 1014 buf->f_files = user_block_count; 1015 buf->f_ffree = buf->f_bavail; 1016 } else { 1017 buf->f_files = avail_node_count; 1018 buf->f_ffree = min(avail_node_count - valid_node_count(sbi), 1019 buf->f_bavail); 1020 } 1021 1022 buf->f_namelen = F2FS_NAME_LEN; 1023 buf->f_fsid.val[0] = (u32)id; 1024 buf->f_fsid.val[1] = (u32)(id >> 32); 1025 1026 #ifdef CONFIG_QUOTA 1027 if (is_inode_flag_set(dentry->d_inode, FI_PROJ_INHERIT) && 1028 sb_has_quota_limits_enabled(sb, PRJQUOTA)) { 1029 f2fs_statfs_project(sb, F2FS_I(dentry->d_inode)->i_projid, buf); 1030 } 1031 #endif 1032 return 0; 1033 } 1034 1035 static inline void f2fs_show_quota_options(struct seq_file *seq, 1036 struct super_block *sb) 1037 { 1038 #ifdef CONFIG_QUOTA 1039 struct f2fs_sb_info *sbi = F2FS_SB(sb); 1040 1041 if (sbi->s_jquota_fmt) { 1042 char *fmtname = ""; 1043 1044 switch (sbi->s_jquota_fmt) { 1045 case QFMT_VFS_OLD: 1046 fmtname = "vfsold"; 1047 break; 1048 case QFMT_VFS_V0: 1049 fmtname = "vfsv0"; 1050 break; 1051 case QFMT_VFS_V1: 1052 fmtname = "vfsv1"; 1053 break; 1054 } 1055 seq_printf(seq, ",jqfmt=%s", fmtname); 1056 } 1057 1058 if (sbi->s_qf_names[USRQUOTA]) 1059 seq_show_option(seq, "usrjquota", sbi->s_qf_names[USRQUOTA]); 1060 1061 if (sbi->s_qf_names[GRPQUOTA]) 1062 seq_show_option(seq, "grpjquota", sbi->s_qf_names[GRPQUOTA]); 1063 1064 if (sbi->s_qf_names[PRJQUOTA]) 1065 seq_show_option(seq, "prjjquota", sbi->s_qf_names[PRJQUOTA]); 1066 #endif 1067 } 1068 1069 static int f2fs_show_options(struct seq_file *seq, struct dentry *root) 1070 { 1071 struct f2fs_sb_info *sbi = F2FS_SB(root->d_sb); 1072 1073 if (!f2fs_readonly(sbi->sb) && test_opt(sbi, BG_GC)) { 1074 if (test_opt(sbi, FORCE_FG_GC)) 1075 seq_printf(seq, ",background_gc=%s", "sync"); 1076 else 1077 seq_printf(seq, ",background_gc=%s", "on"); 1078 } else { 1079 seq_printf(seq, ",background_gc=%s", "off"); 1080 } 1081 if (test_opt(sbi, DISABLE_ROLL_FORWARD)) 1082 seq_puts(seq, ",disable_roll_forward"); 1083 if (test_opt(sbi, DISCARD)) 1084 seq_puts(seq, ",discard"); 1085 if (test_opt(sbi, NOHEAP)) 1086 seq_puts(seq, ",no_heap"); 1087 else 1088 seq_puts(seq, ",heap"); 1089 #ifdef CONFIG_F2FS_FS_XATTR 1090 if (test_opt(sbi, XATTR_USER)) 1091 seq_puts(seq, ",user_xattr"); 1092 else 1093 seq_puts(seq, ",nouser_xattr"); 1094 if (test_opt(sbi, INLINE_XATTR)) 1095 seq_puts(seq, ",inline_xattr"); 1096 else 1097 seq_puts(seq, ",noinline_xattr"); 1098 if (test_opt(sbi, INLINE_XATTR_SIZE)) 1099 seq_printf(seq, ",inline_xattr_size=%u", 1100 sbi->inline_xattr_size); 1101 #endif 1102 #ifdef CONFIG_F2FS_FS_POSIX_ACL 1103 if (test_opt(sbi, POSIX_ACL)) 1104 seq_puts(seq, ",acl"); 1105 else 1106 seq_puts(seq, ",noacl"); 1107 #endif 1108 if (test_opt(sbi, DISABLE_EXT_IDENTIFY)) 1109 seq_puts(seq, ",disable_ext_identify"); 1110 if (test_opt(sbi, INLINE_DATA)) 1111 seq_puts(seq, ",inline_data"); 1112 else 1113 seq_puts(seq, ",noinline_data"); 1114 if (test_opt(sbi, INLINE_DENTRY)) 1115 seq_puts(seq, ",inline_dentry"); 1116 else 1117 seq_puts(seq, ",noinline_dentry"); 1118 if (!f2fs_readonly(sbi->sb) && test_opt(sbi, FLUSH_MERGE)) 1119 seq_puts(seq, ",flush_merge"); 1120 if (test_opt(sbi, NOBARRIER)) 1121 seq_puts(seq, ",nobarrier"); 1122 if (test_opt(sbi, FASTBOOT)) 1123 seq_puts(seq, ",fastboot"); 1124 if (test_opt(sbi, EXTENT_CACHE)) 1125 seq_puts(seq, ",extent_cache"); 1126 else 1127 seq_puts(seq, ",noextent_cache"); 1128 if (test_opt(sbi, DATA_FLUSH)) 1129 seq_puts(seq, ",data_flush"); 1130 1131 seq_puts(seq, ",mode="); 1132 if (test_opt(sbi, ADAPTIVE)) 1133 seq_puts(seq, "adaptive"); 1134 else if (test_opt(sbi, LFS)) 1135 seq_puts(seq, "lfs"); 1136 seq_printf(seq, ",active_logs=%u", sbi->active_logs); 1137 if (F2FS_IO_SIZE_BITS(sbi)) 1138 seq_printf(seq, ",io_size=%uKB", F2FS_IO_SIZE_KB(sbi)); 1139 #ifdef CONFIG_F2FS_FAULT_INJECTION 1140 if (test_opt(sbi, FAULT_INJECTION)) 1141 seq_printf(seq, ",fault_injection=%u", 1142 sbi->fault_info.inject_rate); 1143 #endif 1144 #ifdef CONFIG_QUOTA 1145 if (test_opt(sbi, QUOTA)) 1146 seq_puts(seq, ",quota"); 1147 if (test_opt(sbi, USRQUOTA)) 1148 seq_puts(seq, ",usrquota"); 1149 if (test_opt(sbi, GRPQUOTA)) 1150 seq_puts(seq, ",grpquota"); 1151 if (test_opt(sbi, PRJQUOTA)) 1152 seq_puts(seq, ",prjquota"); 1153 #endif 1154 f2fs_show_quota_options(seq, sbi->sb); 1155 1156 return 0; 1157 } 1158 1159 static void default_options(struct f2fs_sb_info *sbi) 1160 { 1161 /* init some FS parameters */ 1162 sbi->active_logs = NR_CURSEG_TYPE; 1163 sbi->inline_xattr_size = DEFAULT_INLINE_XATTR_ADDRS; 1164 1165 set_opt(sbi, BG_GC); 1166 set_opt(sbi, INLINE_XATTR); 1167 set_opt(sbi, INLINE_DATA); 1168 set_opt(sbi, INLINE_DENTRY); 1169 set_opt(sbi, EXTENT_CACHE); 1170 set_opt(sbi, NOHEAP); 1171 sbi->sb->s_flags |= SB_LAZYTIME; 1172 set_opt(sbi, FLUSH_MERGE); 1173 if (f2fs_sb_mounted_blkzoned(sbi->sb)) { 1174 set_opt_mode(sbi, F2FS_MOUNT_LFS); 1175 set_opt(sbi, DISCARD); 1176 } else { 1177 set_opt_mode(sbi, F2FS_MOUNT_ADAPTIVE); 1178 } 1179 1180 #ifdef CONFIG_F2FS_FS_XATTR 1181 set_opt(sbi, XATTR_USER); 1182 #endif 1183 #ifdef CONFIG_F2FS_FS_POSIX_ACL 1184 set_opt(sbi, POSIX_ACL); 1185 #endif 1186 1187 #ifdef CONFIG_F2FS_FAULT_INJECTION 1188 f2fs_build_fault_attr(sbi, 0); 1189 #endif 1190 } 1191 1192 #ifdef CONFIG_QUOTA 1193 static int f2fs_enable_quotas(struct super_block *sb); 1194 #endif 1195 static int f2fs_remount(struct super_block *sb, int *flags, char *data) 1196 { 1197 struct f2fs_sb_info *sbi = F2FS_SB(sb); 1198 struct f2fs_mount_info org_mount_opt; 1199 unsigned long old_sb_flags; 1200 int err, active_logs; 1201 bool need_restart_gc = false; 1202 bool need_stop_gc = false; 1203 bool no_extent_cache = !test_opt(sbi, EXTENT_CACHE); 1204 #ifdef CONFIG_F2FS_FAULT_INJECTION 1205 struct f2fs_fault_info ffi = sbi->fault_info; 1206 #endif 1207 #ifdef CONFIG_QUOTA 1208 int s_jquota_fmt; 1209 char *s_qf_names[MAXQUOTAS]; 1210 int i, j; 1211 #endif 1212 1213 /* 1214 * Save the old mount options in case we 1215 * need to restore them. 1216 */ 1217 org_mount_opt = sbi->mount_opt; 1218 old_sb_flags = sb->s_flags; 1219 active_logs = sbi->active_logs; 1220 1221 #ifdef CONFIG_QUOTA 1222 s_jquota_fmt = sbi->s_jquota_fmt; 1223 for (i = 0; i < MAXQUOTAS; i++) { 1224 if (sbi->s_qf_names[i]) { 1225 s_qf_names[i] = kstrdup(sbi->s_qf_names[i], 1226 GFP_KERNEL); 1227 if (!s_qf_names[i]) { 1228 for (j = 0; j < i; j++) 1229 kfree(s_qf_names[j]); 1230 return -ENOMEM; 1231 } 1232 } else { 1233 s_qf_names[i] = NULL; 1234 } 1235 } 1236 #endif 1237 1238 /* recover superblocks we couldn't write due to previous RO mount */ 1239 if (!(*flags & SB_RDONLY) && is_sbi_flag_set(sbi, SBI_NEED_SB_WRITE)) { 1240 err = f2fs_commit_super(sbi, false); 1241 f2fs_msg(sb, KERN_INFO, 1242 "Try to recover all the superblocks, ret: %d", err); 1243 if (!err) 1244 clear_sbi_flag(sbi, SBI_NEED_SB_WRITE); 1245 } 1246 1247 default_options(sbi); 1248 1249 /* parse mount options */ 1250 err = parse_options(sb, data); 1251 if (err) 1252 goto restore_opts; 1253 1254 /* 1255 * Previous and new state of filesystem is RO, 1256 * so skip checking GC and FLUSH_MERGE conditions. 1257 */ 1258 if (f2fs_readonly(sb) && (*flags & SB_RDONLY)) 1259 goto skip; 1260 1261 #ifdef CONFIG_QUOTA 1262 if (!f2fs_readonly(sb) && (*flags & SB_RDONLY)) { 1263 err = dquot_suspend(sb, -1); 1264 if (err < 0) 1265 goto restore_opts; 1266 } else { 1267 /* dquot_resume needs RW */ 1268 sb->s_flags &= ~SB_RDONLY; 1269 if (sb_any_quota_suspended(sb)) { 1270 dquot_resume(sb, -1); 1271 } else if (f2fs_sb_has_quota_ino(sb)) { 1272 err = f2fs_enable_quotas(sb); 1273 if (err) 1274 goto restore_opts; 1275 } 1276 } 1277 #endif 1278 /* disallow enable/disable extent_cache dynamically */ 1279 if (no_extent_cache == !!test_opt(sbi, EXTENT_CACHE)) { 1280 err = -EINVAL; 1281 f2fs_msg(sbi->sb, KERN_WARNING, 1282 "switch extent_cache option is not allowed"); 1283 goto restore_opts; 1284 } 1285 1286 /* 1287 * We stop the GC thread if FS is mounted as RO 1288 * or if background_gc = off is passed in mount 1289 * option. Also sync the filesystem. 1290 */ 1291 if ((*flags & SB_RDONLY) || !test_opt(sbi, BG_GC)) { 1292 if (sbi->gc_thread) { 1293 stop_gc_thread(sbi); 1294 need_restart_gc = true; 1295 } 1296 } else if (!sbi->gc_thread) { 1297 err = start_gc_thread(sbi); 1298 if (err) 1299 goto restore_opts; 1300 need_stop_gc = true; 1301 } 1302 1303 if (*flags & SB_RDONLY) { 1304 writeback_inodes_sb(sb, WB_REASON_SYNC); 1305 sync_inodes_sb(sb); 1306 1307 set_sbi_flag(sbi, SBI_IS_DIRTY); 1308 set_sbi_flag(sbi, SBI_IS_CLOSE); 1309 f2fs_sync_fs(sb, 1); 1310 clear_sbi_flag(sbi, SBI_IS_CLOSE); 1311 } 1312 1313 /* 1314 * We stop issue flush thread if FS is mounted as RO 1315 * or if flush_merge is not passed in mount option. 1316 */ 1317 if ((*flags & SB_RDONLY) || !test_opt(sbi, FLUSH_MERGE)) { 1318 clear_opt(sbi, FLUSH_MERGE); 1319 destroy_flush_cmd_control(sbi, false); 1320 } else { 1321 err = create_flush_cmd_control(sbi); 1322 if (err) 1323 goto restore_gc; 1324 } 1325 skip: 1326 #ifdef CONFIG_QUOTA 1327 /* Release old quota file names */ 1328 for (i = 0; i < MAXQUOTAS; i++) 1329 kfree(s_qf_names[i]); 1330 #endif 1331 /* Update the POSIXACL Flag */ 1332 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) | 1333 (test_opt(sbi, POSIX_ACL) ? SB_POSIXACL : 0); 1334 1335 return 0; 1336 restore_gc: 1337 if (need_restart_gc) { 1338 if (start_gc_thread(sbi)) 1339 f2fs_msg(sbi->sb, KERN_WARNING, 1340 "background gc thread has stopped"); 1341 } else if (need_stop_gc) { 1342 stop_gc_thread(sbi); 1343 } 1344 restore_opts: 1345 #ifdef CONFIG_QUOTA 1346 sbi->s_jquota_fmt = s_jquota_fmt; 1347 for (i = 0; i < MAXQUOTAS; i++) { 1348 kfree(sbi->s_qf_names[i]); 1349 sbi->s_qf_names[i] = s_qf_names[i]; 1350 } 1351 #endif 1352 sbi->mount_opt = org_mount_opt; 1353 sbi->active_logs = active_logs; 1354 sb->s_flags = old_sb_flags; 1355 #ifdef CONFIG_F2FS_FAULT_INJECTION 1356 sbi->fault_info = ffi; 1357 #endif 1358 return err; 1359 } 1360 1361 #ifdef CONFIG_QUOTA 1362 /* Read data from quotafile */ 1363 static ssize_t f2fs_quota_read(struct super_block *sb, int type, char *data, 1364 size_t len, loff_t off) 1365 { 1366 struct inode *inode = sb_dqopt(sb)->files[type]; 1367 struct address_space *mapping = inode->i_mapping; 1368 block_t blkidx = F2FS_BYTES_TO_BLK(off); 1369 int offset = off & (sb->s_blocksize - 1); 1370 int tocopy; 1371 size_t toread; 1372 loff_t i_size = i_size_read(inode); 1373 struct page *page; 1374 char *kaddr; 1375 1376 if (off > i_size) 1377 return 0; 1378 1379 if (off + len > i_size) 1380 len = i_size - off; 1381 toread = len; 1382 while (toread > 0) { 1383 tocopy = min_t(unsigned long, sb->s_blocksize - offset, toread); 1384 repeat: 1385 page = read_mapping_page(mapping, blkidx, NULL); 1386 if (IS_ERR(page)) { 1387 if (PTR_ERR(page) == -ENOMEM) { 1388 congestion_wait(BLK_RW_ASYNC, HZ/50); 1389 goto repeat; 1390 } 1391 return PTR_ERR(page); 1392 } 1393 1394 lock_page(page); 1395 1396 if (unlikely(page->mapping != mapping)) { 1397 f2fs_put_page(page, 1); 1398 goto repeat; 1399 } 1400 if (unlikely(!PageUptodate(page))) { 1401 f2fs_put_page(page, 1); 1402 return -EIO; 1403 } 1404 1405 kaddr = kmap_atomic(page); 1406 memcpy(data, kaddr + offset, tocopy); 1407 kunmap_atomic(kaddr); 1408 f2fs_put_page(page, 1); 1409 1410 offset = 0; 1411 toread -= tocopy; 1412 data += tocopy; 1413 blkidx++; 1414 } 1415 return len; 1416 } 1417 1418 /* Write to quotafile */ 1419 static ssize_t f2fs_quota_write(struct super_block *sb, int type, 1420 const char *data, size_t len, loff_t off) 1421 { 1422 struct inode *inode = sb_dqopt(sb)->files[type]; 1423 struct address_space *mapping = inode->i_mapping; 1424 const struct address_space_operations *a_ops = mapping->a_ops; 1425 int offset = off & (sb->s_blocksize - 1); 1426 size_t towrite = len; 1427 struct page *page; 1428 char *kaddr; 1429 int err = 0; 1430 int tocopy; 1431 1432 while (towrite > 0) { 1433 tocopy = min_t(unsigned long, sb->s_blocksize - offset, 1434 towrite); 1435 retry: 1436 err = a_ops->write_begin(NULL, mapping, off, tocopy, 0, 1437 &page, NULL); 1438 if (unlikely(err)) { 1439 if (err == -ENOMEM) { 1440 congestion_wait(BLK_RW_ASYNC, HZ/50); 1441 goto retry; 1442 } 1443 break; 1444 } 1445 1446 kaddr = kmap_atomic(page); 1447 memcpy(kaddr + offset, data, tocopy); 1448 kunmap_atomic(kaddr); 1449 flush_dcache_page(page); 1450 1451 a_ops->write_end(NULL, mapping, off, tocopy, tocopy, 1452 page, NULL); 1453 offset = 0; 1454 towrite -= tocopy; 1455 off += tocopy; 1456 data += tocopy; 1457 cond_resched(); 1458 } 1459 1460 if (len == towrite) 1461 return err; 1462 inode->i_mtime = inode->i_ctime = current_time(inode); 1463 f2fs_mark_inode_dirty_sync(inode, false); 1464 return len - towrite; 1465 } 1466 1467 static struct dquot **f2fs_get_dquots(struct inode *inode) 1468 { 1469 return F2FS_I(inode)->i_dquot; 1470 } 1471 1472 static qsize_t *f2fs_get_reserved_space(struct inode *inode) 1473 { 1474 return &F2FS_I(inode)->i_reserved_quota; 1475 } 1476 1477 static int f2fs_quota_on_mount(struct f2fs_sb_info *sbi, int type) 1478 { 1479 return dquot_quota_on_mount(sbi->sb, sbi->s_qf_names[type], 1480 sbi->s_jquota_fmt, type); 1481 } 1482 1483 int f2fs_enable_quota_files(struct f2fs_sb_info *sbi, bool rdonly) 1484 { 1485 int enabled = 0; 1486 int i, err; 1487 1488 if (f2fs_sb_has_quota_ino(sbi->sb) && rdonly) { 1489 err = f2fs_enable_quotas(sbi->sb); 1490 if (err) { 1491 f2fs_msg(sbi->sb, KERN_ERR, 1492 "Cannot turn on quota_ino: %d", err); 1493 return 0; 1494 } 1495 return 1; 1496 } 1497 1498 for (i = 0; i < MAXQUOTAS; i++) { 1499 if (sbi->s_qf_names[i]) { 1500 err = f2fs_quota_on_mount(sbi, i); 1501 if (!err) { 1502 enabled = 1; 1503 continue; 1504 } 1505 f2fs_msg(sbi->sb, KERN_ERR, 1506 "Cannot turn on quotas: %d on %d", err, i); 1507 } 1508 } 1509 return enabled; 1510 } 1511 1512 static int f2fs_quota_enable(struct super_block *sb, int type, int format_id, 1513 unsigned int flags) 1514 { 1515 struct inode *qf_inode; 1516 unsigned long qf_inum; 1517 int err; 1518 1519 BUG_ON(!f2fs_sb_has_quota_ino(sb)); 1520 1521 qf_inum = f2fs_qf_ino(sb, type); 1522 if (!qf_inum) 1523 return -EPERM; 1524 1525 qf_inode = f2fs_iget(sb, qf_inum); 1526 if (IS_ERR(qf_inode)) { 1527 f2fs_msg(sb, KERN_ERR, 1528 "Bad quota inode %u:%lu", type, qf_inum); 1529 return PTR_ERR(qf_inode); 1530 } 1531 1532 /* Don't account quota for quota files to avoid recursion */ 1533 qf_inode->i_flags |= S_NOQUOTA; 1534 err = dquot_enable(qf_inode, type, format_id, flags); 1535 iput(qf_inode); 1536 return err; 1537 } 1538 1539 static int f2fs_enable_quotas(struct super_block *sb) 1540 { 1541 int type, err = 0; 1542 unsigned long qf_inum; 1543 bool quota_mopt[MAXQUOTAS] = { 1544 test_opt(F2FS_SB(sb), USRQUOTA), 1545 test_opt(F2FS_SB(sb), GRPQUOTA), 1546 test_opt(F2FS_SB(sb), PRJQUOTA), 1547 }; 1548 1549 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE | DQUOT_NOLIST_DIRTY; 1550 for (type = 0; type < MAXQUOTAS; type++) { 1551 qf_inum = f2fs_qf_ino(sb, type); 1552 if (qf_inum) { 1553 err = f2fs_quota_enable(sb, type, QFMT_VFS_V1, 1554 DQUOT_USAGE_ENABLED | 1555 (quota_mopt[type] ? DQUOT_LIMITS_ENABLED : 0)); 1556 if (err) { 1557 f2fs_msg(sb, KERN_ERR, 1558 "Failed to enable quota tracking " 1559 "(type=%d, err=%d). Please run " 1560 "fsck to fix.", type, err); 1561 for (type--; type >= 0; type--) 1562 dquot_quota_off(sb, type); 1563 return err; 1564 } 1565 } 1566 } 1567 return 0; 1568 } 1569 1570 static int f2fs_quota_sync(struct super_block *sb, int type) 1571 { 1572 struct quota_info *dqopt = sb_dqopt(sb); 1573 int cnt; 1574 int ret; 1575 1576 ret = dquot_writeback_dquots(sb, type); 1577 if (ret) 1578 return ret; 1579 1580 /* 1581 * Now when everything is written we can discard the pagecache so 1582 * that userspace sees the changes. 1583 */ 1584 for (cnt = 0; cnt < MAXQUOTAS; cnt++) { 1585 if (type != -1 && cnt != type) 1586 continue; 1587 if (!sb_has_quota_active(sb, cnt)) 1588 continue; 1589 1590 ret = filemap_write_and_wait(dqopt->files[cnt]->i_mapping); 1591 if (ret) 1592 return ret; 1593 1594 inode_lock(dqopt->files[cnt]); 1595 truncate_inode_pages(&dqopt->files[cnt]->i_data, 0); 1596 inode_unlock(dqopt->files[cnt]); 1597 } 1598 return 0; 1599 } 1600 1601 static int f2fs_quota_on(struct super_block *sb, int type, int format_id, 1602 const struct path *path) 1603 { 1604 struct inode *inode; 1605 int err; 1606 1607 err = f2fs_quota_sync(sb, type); 1608 if (err) 1609 return err; 1610 1611 err = dquot_quota_on(sb, type, format_id, path); 1612 if (err) 1613 return err; 1614 1615 inode = d_inode(path->dentry); 1616 1617 inode_lock(inode); 1618 F2FS_I(inode)->i_flags |= FS_NOATIME_FL | FS_IMMUTABLE_FL; 1619 inode_set_flags(inode, S_NOATIME | S_IMMUTABLE, 1620 S_NOATIME | S_IMMUTABLE); 1621 inode_unlock(inode); 1622 f2fs_mark_inode_dirty_sync(inode, false); 1623 1624 return 0; 1625 } 1626 1627 static int f2fs_quota_off(struct super_block *sb, int type) 1628 { 1629 struct inode *inode = sb_dqopt(sb)->files[type]; 1630 int err; 1631 1632 if (!inode || !igrab(inode)) 1633 return dquot_quota_off(sb, type); 1634 1635 f2fs_quota_sync(sb, type); 1636 1637 err = dquot_quota_off(sb, type); 1638 if (err || f2fs_sb_has_quota_ino(sb)) 1639 goto out_put; 1640 1641 inode_lock(inode); 1642 F2FS_I(inode)->i_flags &= ~(FS_NOATIME_FL | FS_IMMUTABLE_FL); 1643 inode_set_flags(inode, 0, S_NOATIME | S_IMMUTABLE); 1644 inode_unlock(inode); 1645 f2fs_mark_inode_dirty_sync(inode, false); 1646 out_put: 1647 iput(inode); 1648 return err; 1649 } 1650 1651 void f2fs_quota_off_umount(struct super_block *sb) 1652 { 1653 int type; 1654 1655 for (type = 0; type < MAXQUOTAS; type++) 1656 f2fs_quota_off(sb, type); 1657 } 1658 1659 int f2fs_get_projid(struct inode *inode, kprojid_t *projid) 1660 { 1661 *projid = F2FS_I(inode)->i_projid; 1662 return 0; 1663 } 1664 1665 static const struct dquot_operations f2fs_quota_operations = { 1666 .get_reserved_space = f2fs_get_reserved_space, 1667 .write_dquot = dquot_commit, 1668 .acquire_dquot = dquot_acquire, 1669 .release_dquot = dquot_release, 1670 .mark_dirty = dquot_mark_dquot_dirty, 1671 .write_info = dquot_commit_info, 1672 .alloc_dquot = dquot_alloc, 1673 .destroy_dquot = dquot_destroy, 1674 .get_projid = f2fs_get_projid, 1675 .get_next_id = dquot_get_next_id, 1676 }; 1677 1678 static const struct quotactl_ops f2fs_quotactl_ops = { 1679 .quota_on = f2fs_quota_on, 1680 .quota_off = f2fs_quota_off, 1681 .quota_sync = f2fs_quota_sync, 1682 .get_state = dquot_get_state, 1683 .set_info = dquot_set_dqinfo, 1684 .get_dqblk = dquot_get_dqblk, 1685 .set_dqblk = dquot_set_dqblk, 1686 .get_nextdqblk = dquot_get_next_dqblk, 1687 }; 1688 #else 1689 void f2fs_quota_off_umount(struct super_block *sb) 1690 { 1691 } 1692 #endif 1693 1694 static const struct super_operations f2fs_sops = { 1695 .alloc_inode = f2fs_alloc_inode, 1696 .drop_inode = f2fs_drop_inode, 1697 .destroy_inode = f2fs_destroy_inode, 1698 .write_inode = f2fs_write_inode, 1699 .dirty_inode = f2fs_dirty_inode, 1700 .show_options = f2fs_show_options, 1701 #ifdef CONFIG_QUOTA 1702 .quota_read = f2fs_quota_read, 1703 .quota_write = f2fs_quota_write, 1704 .get_dquots = f2fs_get_dquots, 1705 #endif 1706 .evict_inode = f2fs_evict_inode, 1707 .put_super = f2fs_put_super, 1708 .sync_fs = f2fs_sync_fs, 1709 .freeze_fs = f2fs_freeze, 1710 .unfreeze_fs = f2fs_unfreeze, 1711 .statfs = f2fs_statfs, 1712 .remount_fs = f2fs_remount, 1713 }; 1714 1715 #ifdef CONFIG_F2FS_FS_ENCRYPTION 1716 static int f2fs_get_context(struct inode *inode, void *ctx, size_t len) 1717 { 1718 return f2fs_getxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION, 1719 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT, 1720 ctx, len, NULL); 1721 } 1722 1723 static int f2fs_set_context(struct inode *inode, const void *ctx, size_t len, 1724 void *fs_data) 1725 { 1726 return f2fs_setxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION, 1727 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT, 1728 ctx, len, fs_data, XATTR_CREATE); 1729 } 1730 1731 static unsigned f2fs_max_namelen(struct inode *inode) 1732 { 1733 return S_ISLNK(inode->i_mode) ? 1734 inode->i_sb->s_blocksize : F2FS_NAME_LEN; 1735 } 1736 1737 static const struct fscrypt_operations f2fs_cryptops = { 1738 .key_prefix = "f2fs:", 1739 .get_context = f2fs_get_context, 1740 .set_context = f2fs_set_context, 1741 .empty_dir = f2fs_empty_dir, 1742 .max_namelen = f2fs_max_namelen, 1743 }; 1744 #endif 1745 1746 static struct inode *f2fs_nfs_get_inode(struct super_block *sb, 1747 u64 ino, u32 generation) 1748 { 1749 struct f2fs_sb_info *sbi = F2FS_SB(sb); 1750 struct inode *inode; 1751 1752 if (check_nid_range(sbi, ino)) 1753 return ERR_PTR(-ESTALE); 1754 1755 /* 1756 * f2fs_iget isn't quite right if the inode is currently unallocated! 1757 * However f2fs_iget currently does appropriate checks to handle stale 1758 * inodes so everything is OK. 1759 */ 1760 inode = f2fs_iget(sb, ino); 1761 if (IS_ERR(inode)) 1762 return ERR_CAST(inode); 1763 if (unlikely(generation && inode->i_generation != generation)) { 1764 /* we didn't find the right inode.. */ 1765 iput(inode); 1766 return ERR_PTR(-ESTALE); 1767 } 1768 return inode; 1769 } 1770 1771 static struct dentry *f2fs_fh_to_dentry(struct super_block *sb, struct fid *fid, 1772 int fh_len, int fh_type) 1773 { 1774 return generic_fh_to_dentry(sb, fid, fh_len, fh_type, 1775 f2fs_nfs_get_inode); 1776 } 1777 1778 static struct dentry *f2fs_fh_to_parent(struct super_block *sb, struct fid *fid, 1779 int fh_len, int fh_type) 1780 { 1781 return generic_fh_to_parent(sb, fid, fh_len, fh_type, 1782 f2fs_nfs_get_inode); 1783 } 1784 1785 static const struct export_operations f2fs_export_ops = { 1786 .fh_to_dentry = f2fs_fh_to_dentry, 1787 .fh_to_parent = f2fs_fh_to_parent, 1788 .get_parent = f2fs_get_parent, 1789 }; 1790 1791 static loff_t max_file_blocks(void) 1792 { 1793 loff_t result = 0; 1794 loff_t leaf_count = ADDRS_PER_BLOCK; 1795 1796 /* 1797 * note: previously, result is equal to (DEF_ADDRS_PER_INODE - 1798 * DEFAULT_INLINE_XATTR_ADDRS), but now f2fs try to reserve more 1799 * space in inode.i_addr, it will be more safe to reassign 1800 * result as zero. 1801 */ 1802 1803 /* two direct node blocks */ 1804 result += (leaf_count * 2); 1805 1806 /* two indirect node blocks */ 1807 leaf_count *= NIDS_PER_BLOCK; 1808 result += (leaf_count * 2); 1809 1810 /* one double indirect node block */ 1811 leaf_count *= NIDS_PER_BLOCK; 1812 result += leaf_count; 1813 1814 return result; 1815 } 1816 1817 static int __f2fs_commit_super(struct buffer_head *bh, 1818 struct f2fs_super_block *super) 1819 { 1820 lock_buffer(bh); 1821 if (super) 1822 memcpy(bh->b_data + F2FS_SUPER_OFFSET, super, sizeof(*super)); 1823 set_buffer_uptodate(bh); 1824 set_buffer_dirty(bh); 1825 unlock_buffer(bh); 1826 1827 /* it's rare case, we can do fua all the time */ 1828 return __sync_dirty_buffer(bh, REQ_SYNC | REQ_PREFLUSH | REQ_FUA); 1829 } 1830 1831 static inline bool sanity_check_area_boundary(struct f2fs_sb_info *sbi, 1832 struct buffer_head *bh) 1833 { 1834 struct f2fs_super_block *raw_super = (struct f2fs_super_block *) 1835 (bh->b_data + F2FS_SUPER_OFFSET); 1836 struct super_block *sb = sbi->sb; 1837 u32 segment0_blkaddr = le32_to_cpu(raw_super->segment0_blkaddr); 1838 u32 cp_blkaddr = le32_to_cpu(raw_super->cp_blkaddr); 1839 u32 sit_blkaddr = le32_to_cpu(raw_super->sit_blkaddr); 1840 u32 nat_blkaddr = le32_to_cpu(raw_super->nat_blkaddr); 1841 u32 ssa_blkaddr = le32_to_cpu(raw_super->ssa_blkaddr); 1842 u32 main_blkaddr = le32_to_cpu(raw_super->main_blkaddr); 1843 u32 segment_count_ckpt = le32_to_cpu(raw_super->segment_count_ckpt); 1844 u32 segment_count_sit = le32_to_cpu(raw_super->segment_count_sit); 1845 u32 segment_count_nat = le32_to_cpu(raw_super->segment_count_nat); 1846 u32 segment_count_ssa = le32_to_cpu(raw_super->segment_count_ssa); 1847 u32 segment_count_main = le32_to_cpu(raw_super->segment_count_main); 1848 u32 segment_count = le32_to_cpu(raw_super->segment_count); 1849 u32 log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg); 1850 u64 main_end_blkaddr = main_blkaddr + 1851 (segment_count_main << log_blocks_per_seg); 1852 u64 seg_end_blkaddr = segment0_blkaddr + 1853 (segment_count << log_blocks_per_seg); 1854 1855 if (segment0_blkaddr != cp_blkaddr) { 1856 f2fs_msg(sb, KERN_INFO, 1857 "Mismatch start address, segment0(%u) cp_blkaddr(%u)", 1858 segment0_blkaddr, cp_blkaddr); 1859 return true; 1860 } 1861 1862 if (cp_blkaddr + (segment_count_ckpt << log_blocks_per_seg) != 1863 sit_blkaddr) { 1864 f2fs_msg(sb, KERN_INFO, 1865 "Wrong CP boundary, start(%u) end(%u) blocks(%u)", 1866 cp_blkaddr, sit_blkaddr, 1867 segment_count_ckpt << log_blocks_per_seg); 1868 return true; 1869 } 1870 1871 if (sit_blkaddr + (segment_count_sit << log_blocks_per_seg) != 1872 nat_blkaddr) { 1873 f2fs_msg(sb, KERN_INFO, 1874 "Wrong SIT boundary, start(%u) end(%u) blocks(%u)", 1875 sit_blkaddr, nat_blkaddr, 1876 segment_count_sit << log_blocks_per_seg); 1877 return true; 1878 } 1879 1880 if (nat_blkaddr + (segment_count_nat << log_blocks_per_seg) != 1881 ssa_blkaddr) { 1882 f2fs_msg(sb, KERN_INFO, 1883 "Wrong NAT boundary, start(%u) end(%u) blocks(%u)", 1884 nat_blkaddr, ssa_blkaddr, 1885 segment_count_nat << log_blocks_per_seg); 1886 return true; 1887 } 1888 1889 if (ssa_blkaddr + (segment_count_ssa << log_blocks_per_seg) != 1890 main_blkaddr) { 1891 f2fs_msg(sb, KERN_INFO, 1892 "Wrong SSA boundary, start(%u) end(%u) blocks(%u)", 1893 ssa_blkaddr, main_blkaddr, 1894 segment_count_ssa << log_blocks_per_seg); 1895 return true; 1896 } 1897 1898 if (main_end_blkaddr > seg_end_blkaddr) { 1899 f2fs_msg(sb, KERN_INFO, 1900 "Wrong MAIN_AREA boundary, start(%u) end(%u) block(%u)", 1901 main_blkaddr, 1902 segment0_blkaddr + 1903 (segment_count << log_blocks_per_seg), 1904 segment_count_main << log_blocks_per_seg); 1905 return true; 1906 } else if (main_end_blkaddr < seg_end_blkaddr) { 1907 int err = 0; 1908 char *res; 1909 1910 /* fix in-memory information all the time */ 1911 raw_super->segment_count = cpu_to_le32((main_end_blkaddr - 1912 segment0_blkaddr) >> log_blocks_per_seg); 1913 1914 if (f2fs_readonly(sb) || bdev_read_only(sb->s_bdev)) { 1915 set_sbi_flag(sbi, SBI_NEED_SB_WRITE); 1916 res = "internally"; 1917 } else { 1918 err = __f2fs_commit_super(bh, NULL); 1919 res = err ? "failed" : "done"; 1920 } 1921 f2fs_msg(sb, KERN_INFO, 1922 "Fix alignment : %s, start(%u) end(%u) block(%u)", 1923 res, main_blkaddr, 1924 segment0_blkaddr + 1925 (segment_count << log_blocks_per_seg), 1926 segment_count_main << log_blocks_per_seg); 1927 if (err) 1928 return true; 1929 } 1930 return false; 1931 } 1932 1933 static int sanity_check_raw_super(struct f2fs_sb_info *sbi, 1934 struct buffer_head *bh) 1935 { 1936 struct f2fs_super_block *raw_super = (struct f2fs_super_block *) 1937 (bh->b_data + F2FS_SUPER_OFFSET); 1938 struct super_block *sb = sbi->sb; 1939 unsigned int blocksize; 1940 1941 if (F2FS_SUPER_MAGIC != le32_to_cpu(raw_super->magic)) { 1942 f2fs_msg(sb, KERN_INFO, 1943 "Magic Mismatch, valid(0x%x) - read(0x%x)", 1944 F2FS_SUPER_MAGIC, le32_to_cpu(raw_super->magic)); 1945 return 1; 1946 } 1947 1948 /* Currently, support only 4KB page cache size */ 1949 if (F2FS_BLKSIZE != PAGE_SIZE) { 1950 f2fs_msg(sb, KERN_INFO, 1951 "Invalid page_cache_size (%lu), supports only 4KB\n", 1952 PAGE_SIZE); 1953 return 1; 1954 } 1955 1956 /* Currently, support only 4KB block size */ 1957 blocksize = 1 << le32_to_cpu(raw_super->log_blocksize); 1958 if (blocksize != F2FS_BLKSIZE) { 1959 f2fs_msg(sb, KERN_INFO, 1960 "Invalid blocksize (%u), supports only 4KB\n", 1961 blocksize); 1962 return 1; 1963 } 1964 1965 /* check log blocks per segment */ 1966 if (le32_to_cpu(raw_super->log_blocks_per_seg) != 9) { 1967 f2fs_msg(sb, KERN_INFO, 1968 "Invalid log blocks per segment (%u)\n", 1969 le32_to_cpu(raw_super->log_blocks_per_seg)); 1970 return 1; 1971 } 1972 1973 /* Currently, support 512/1024/2048/4096 bytes sector size */ 1974 if (le32_to_cpu(raw_super->log_sectorsize) > 1975 F2FS_MAX_LOG_SECTOR_SIZE || 1976 le32_to_cpu(raw_super->log_sectorsize) < 1977 F2FS_MIN_LOG_SECTOR_SIZE) { 1978 f2fs_msg(sb, KERN_INFO, "Invalid log sectorsize (%u)", 1979 le32_to_cpu(raw_super->log_sectorsize)); 1980 return 1; 1981 } 1982 if (le32_to_cpu(raw_super->log_sectors_per_block) + 1983 le32_to_cpu(raw_super->log_sectorsize) != 1984 F2FS_MAX_LOG_SECTOR_SIZE) { 1985 f2fs_msg(sb, KERN_INFO, 1986 "Invalid log sectors per block(%u) log sectorsize(%u)", 1987 le32_to_cpu(raw_super->log_sectors_per_block), 1988 le32_to_cpu(raw_super->log_sectorsize)); 1989 return 1; 1990 } 1991 1992 /* check reserved ino info */ 1993 if (le32_to_cpu(raw_super->node_ino) != 1 || 1994 le32_to_cpu(raw_super->meta_ino) != 2 || 1995 le32_to_cpu(raw_super->root_ino) != 3) { 1996 f2fs_msg(sb, KERN_INFO, 1997 "Invalid Fs Meta Ino: node(%u) meta(%u) root(%u)", 1998 le32_to_cpu(raw_super->node_ino), 1999 le32_to_cpu(raw_super->meta_ino), 2000 le32_to_cpu(raw_super->root_ino)); 2001 return 1; 2002 } 2003 2004 if (le32_to_cpu(raw_super->segment_count) > F2FS_MAX_SEGMENT) { 2005 f2fs_msg(sb, KERN_INFO, 2006 "Invalid segment count (%u)", 2007 le32_to_cpu(raw_super->segment_count)); 2008 return 1; 2009 } 2010 2011 /* check CP/SIT/NAT/SSA/MAIN_AREA area boundary */ 2012 if (sanity_check_area_boundary(sbi, bh)) 2013 return 1; 2014 2015 return 0; 2016 } 2017 2018 int sanity_check_ckpt(struct f2fs_sb_info *sbi) 2019 { 2020 unsigned int total, fsmeta; 2021 struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi); 2022 struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi); 2023 unsigned int ovp_segments, reserved_segments; 2024 unsigned int main_segs, blocks_per_seg; 2025 int i; 2026 2027 total = le32_to_cpu(raw_super->segment_count); 2028 fsmeta = le32_to_cpu(raw_super->segment_count_ckpt); 2029 fsmeta += le32_to_cpu(raw_super->segment_count_sit); 2030 fsmeta += le32_to_cpu(raw_super->segment_count_nat); 2031 fsmeta += le32_to_cpu(ckpt->rsvd_segment_count); 2032 fsmeta += le32_to_cpu(raw_super->segment_count_ssa); 2033 2034 if (unlikely(fsmeta >= total)) 2035 return 1; 2036 2037 ovp_segments = le32_to_cpu(ckpt->overprov_segment_count); 2038 reserved_segments = le32_to_cpu(ckpt->rsvd_segment_count); 2039 2040 if (unlikely(fsmeta < F2FS_MIN_SEGMENTS || 2041 ovp_segments == 0 || reserved_segments == 0)) { 2042 f2fs_msg(sbi->sb, KERN_ERR, 2043 "Wrong layout: check mkfs.f2fs version"); 2044 return 1; 2045 } 2046 2047 main_segs = le32_to_cpu(raw_super->segment_count_main); 2048 blocks_per_seg = sbi->blocks_per_seg; 2049 2050 for (i = 0; i < NR_CURSEG_NODE_TYPE; i++) { 2051 if (le32_to_cpu(ckpt->cur_node_segno[i]) >= main_segs || 2052 le16_to_cpu(ckpt->cur_node_blkoff[i]) >= blocks_per_seg) 2053 return 1; 2054 } 2055 for (i = 0; i < NR_CURSEG_DATA_TYPE; i++) { 2056 if (le32_to_cpu(ckpt->cur_data_segno[i]) >= main_segs || 2057 le16_to_cpu(ckpt->cur_data_blkoff[i]) >= blocks_per_seg) 2058 return 1; 2059 } 2060 2061 if (unlikely(f2fs_cp_error(sbi))) { 2062 f2fs_msg(sbi->sb, KERN_ERR, "A bug case: need to run fsck"); 2063 return 1; 2064 } 2065 return 0; 2066 } 2067 2068 static void init_sb_info(struct f2fs_sb_info *sbi) 2069 { 2070 struct f2fs_super_block *raw_super = sbi->raw_super; 2071 int i, j; 2072 2073 sbi->log_sectors_per_block = 2074 le32_to_cpu(raw_super->log_sectors_per_block); 2075 sbi->log_blocksize = le32_to_cpu(raw_super->log_blocksize); 2076 sbi->blocksize = 1 << sbi->log_blocksize; 2077 sbi->log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg); 2078 sbi->blocks_per_seg = 1 << sbi->log_blocks_per_seg; 2079 sbi->segs_per_sec = le32_to_cpu(raw_super->segs_per_sec); 2080 sbi->secs_per_zone = le32_to_cpu(raw_super->secs_per_zone); 2081 sbi->total_sections = le32_to_cpu(raw_super->section_count); 2082 sbi->total_node_count = 2083 (le32_to_cpu(raw_super->segment_count_nat) / 2) 2084 * sbi->blocks_per_seg * NAT_ENTRY_PER_BLOCK; 2085 sbi->root_ino_num = le32_to_cpu(raw_super->root_ino); 2086 sbi->node_ino_num = le32_to_cpu(raw_super->node_ino); 2087 sbi->meta_ino_num = le32_to_cpu(raw_super->meta_ino); 2088 sbi->cur_victim_sec = NULL_SECNO; 2089 sbi->max_victim_search = DEF_MAX_VICTIM_SEARCH; 2090 2091 sbi->dir_level = DEF_DIR_LEVEL; 2092 sbi->interval_time[CP_TIME] = DEF_CP_INTERVAL; 2093 sbi->interval_time[REQ_TIME] = DEF_IDLE_INTERVAL; 2094 clear_sbi_flag(sbi, SBI_NEED_FSCK); 2095 2096 for (i = 0; i < NR_COUNT_TYPE; i++) 2097 atomic_set(&sbi->nr_pages[i], 0); 2098 2099 atomic_set(&sbi->wb_sync_req, 0); 2100 2101 INIT_LIST_HEAD(&sbi->s_list); 2102 mutex_init(&sbi->umount_mutex); 2103 for (i = 0; i < NR_PAGE_TYPE - 1; i++) 2104 for (j = HOT; j < NR_TEMP_TYPE; j++) 2105 mutex_init(&sbi->wio_mutex[i][j]); 2106 spin_lock_init(&sbi->cp_lock); 2107 2108 sbi->dirty_device = 0; 2109 spin_lock_init(&sbi->dev_lock); 2110 } 2111 2112 static int init_percpu_info(struct f2fs_sb_info *sbi) 2113 { 2114 int err; 2115 2116 err = percpu_counter_init(&sbi->alloc_valid_block_count, 0, GFP_KERNEL); 2117 if (err) 2118 return err; 2119 2120 return percpu_counter_init(&sbi->total_valid_inode_count, 0, 2121 GFP_KERNEL); 2122 } 2123 2124 #ifdef CONFIG_BLK_DEV_ZONED 2125 static int init_blkz_info(struct f2fs_sb_info *sbi, int devi) 2126 { 2127 struct block_device *bdev = FDEV(devi).bdev; 2128 sector_t nr_sectors = bdev->bd_part->nr_sects; 2129 sector_t sector = 0; 2130 struct blk_zone *zones; 2131 unsigned int i, nr_zones; 2132 unsigned int n = 0; 2133 int err = -EIO; 2134 2135 if (!f2fs_sb_mounted_blkzoned(sbi->sb)) 2136 return 0; 2137 2138 if (sbi->blocks_per_blkz && sbi->blocks_per_blkz != 2139 SECTOR_TO_BLOCK(bdev_zone_sectors(bdev))) 2140 return -EINVAL; 2141 sbi->blocks_per_blkz = SECTOR_TO_BLOCK(bdev_zone_sectors(bdev)); 2142 if (sbi->log_blocks_per_blkz && sbi->log_blocks_per_blkz != 2143 __ilog2_u32(sbi->blocks_per_blkz)) 2144 return -EINVAL; 2145 sbi->log_blocks_per_blkz = __ilog2_u32(sbi->blocks_per_blkz); 2146 FDEV(devi).nr_blkz = SECTOR_TO_BLOCK(nr_sectors) >> 2147 sbi->log_blocks_per_blkz; 2148 if (nr_sectors & (bdev_zone_sectors(bdev) - 1)) 2149 FDEV(devi).nr_blkz++; 2150 2151 FDEV(devi).blkz_type = kmalloc(FDEV(devi).nr_blkz, GFP_KERNEL); 2152 if (!FDEV(devi).blkz_type) 2153 return -ENOMEM; 2154 2155 #define F2FS_REPORT_NR_ZONES 4096 2156 2157 zones = kcalloc(F2FS_REPORT_NR_ZONES, sizeof(struct blk_zone), 2158 GFP_KERNEL); 2159 if (!zones) 2160 return -ENOMEM; 2161 2162 /* Get block zones type */ 2163 while (zones && sector < nr_sectors) { 2164 2165 nr_zones = F2FS_REPORT_NR_ZONES; 2166 err = blkdev_report_zones(bdev, sector, 2167 zones, &nr_zones, 2168 GFP_KERNEL); 2169 if (err) 2170 break; 2171 if (!nr_zones) { 2172 err = -EIO; 2173 break; 2174 } 2175 2176 for (i = 0; i < nr_zones; i++) { 2177 FDEV(devi).blkz_type[n] = zones[i].type; 2178 sector += zones[i].len; 2179 n++; 2180 } 2181 } 2182 2183 kfree(zones); 2184 2185 return err; 2186 } 2187 #endif 2188 2189 /* 2190 * Read f2fs raw super block. 2191 * Because we have two copies of super block, so read both of them 2192 * to get the first valid one. If any one of them is broken, we pass 2193 * them recovery flag back to the caller. 2194 */ 2195 static int read_raw_super_block(struct f2fs_sb_info *sbi, 2196 struct f2fs_super_block **raw_super, 2197 int *valid_super_block, int *recovery) 2198 { 2199 struct super_block *sb = sbi->sb; 2200 int block; 2201 struct buffer_head *bh; 2202 struct f2fs_super_block *super; 2203 int err = 0; 2204 2205 super = kzalloc(sizeof(struct f2fs_super_block), GFP_KERNEL); 2206 if (!super) 2207 return -ENOMEM; 2208 2209 for (block = 0; block < 2; block++) { 2210 bh = sb_bread(sb, block); 2211 if (!bh) { 2212 f2fs_msg(sb, KERN_ERR, "Unable to read %dth superblock", 2213 block + 1); 2214 err = -EIO; 2215 continue; 2216 } 2217 2218 /* sanity checking of raw super */ 2219 if (sanity_check_raw_super(sbi, bh)) { 2220 f2fs_msg(sb, KERN_ERR, 2221 "Can't find valid F2FS filesystem in %dth superblock", 2222 block + 1); 2223 err = -EINVAL; 2224 brelse(bh); 2225 continue; 2226 } 2227 2228 if (!*raw_super) { 2229 memcpy(super, bh->b_data + F2FS_SUPER_OFFSET, 2230 sizeof(*super)); 2231 *valid_super_block = block; 2232 *raw_super = super; 2233 } 2234 brelse(bh); 2235 } 2236 2237 /* Fail to read any one of the superblocks*/ 2238 if (err < 0) 2239 *recovery = 1; 2240 2241 /* No valid superblock */ 2242 if (!*raw_super) 2243 kfree(super); 2244 else 2245 err = 0; 2246 2247 return err; 2248 } 2249 2250 int f2fs_commit_super(struct f2fs_sb_info *sbi, bool recover) 2251 { 2252 struct buffer_head *bh; 2253 int err; 2254 2255 if ((recover && f2fs_readonly(sbi->sb)) || 2256 bdev_read_only(sbi->sb->s_bdev)) { 2257 set_sbi_flag(sbi, SBI_NEED_SB_WRITE); 2258 return -EROFS; 2259 } 2260 2261 /* write back-up superblock first */ 2262 bh = sb_getblk(sbi->sb, sbi->valid_super_block ? 0: 1); 2263 if (!bh) 2264 return -EIO; 2265 err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi)); 2266 brelse(bh); 2267 2268 /* if we are in recovery path, skip writing valid superblock */ 2269 if (recover || err) 2270 return err; 2271 2272 /* write current valid superblock */ 2273 bh = sb_getblk(sbi->sb, sbi->valid_super_block); 2274 if (!bh) 2275 return -EIO; 2276 err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi)); 2277 brelse(bh); 2278 return err; 2279 } 2280 2281 static int f2fs_scan_devices(struct f2fs_sb_info *sbi) 2282 { 2283 struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi); 2284 unsigned int max_devices = MAX_DEVICES; 2285 int i; 2286 2287 /* Initialize single device information */ 2288 if (!RDEV(0).path[0]) { 2289 if (!bdev_is_zoned(sbi->sb->s_bdev)) 2290 return 0; 2291 max_devices = 1; 2292 } 2293 2294 /* 2295 * Initialize multiple devices information, or single 2296 * zoned block device information. 2297 */ 2298 sbi->devs = kcalloc(max_devices, sizeof(struct f2fs_dev_info), 2299 GFP_KERNEL); 2300 if (!sbi->devs) 2301 return -ENOMEM; 2302 2303 for (i = 0; i < max_devices; i++) { 2304 2305 if (i > 0 && !RDEV(i).path[0]) 2306 break; 2307 2308 if (max_devices == 1) { 2309 /* Single zoned block device mount */ 2310 FDEV(0).bdev = 2311 blkdev_get_by_dev(sbi->sb->s_bdev->bd_dev, 2312 sbi->sb->s_mode, sbi->sb->s_type); 2313 } else { 2314 /* Multi-device mount */ 2315 memcpy(FDEV(i).path, RDEV(i).path, MAX_PATH_LEN); 2316 FDEV(i).total_segments = 2317 le32_to_cpu(RDEV(i).total_segments); 2318 if (i == 0) { 2319 FDEV(i).start_blk = 0; 2320 FDEV(i).end_blk = FDEV(i).start_blk + 2321 (FDEV(i).total_segments << 2322 sbi->log_blocks_per_seg) - 1 + 2323 le32_to_cpu(raw_super->segment0_blkaddr); 2324 } else { 2325 FDEV(i).start_blk = FDEV(i - 1).end_blk + 1; 2326 FDEV(i).end_blk = FDEV(i).start_blk + 2327 (FDEV(i).total_segments << 2328 sbi->log_blocks_per_seg) - 1; 2329 } 2330 FDEV(i).bdev = blkdev_get_by_path(FDEV(i).path, 2331 sbi->sb->s_mode, sbi->sb->s_type); 2332 } 2333 if (IS_ERR(FDEV(i).bdev)) 2334 return PTR_ERR(FDEV(i).bdev); 2335 2336 /* to release errored devices */ 2337 sbi->s_ndevs = i + 1; 2338 2339 #ifdef CONFIG_BLK_DEV_ZONED 2340 if (bdev_zoned_model(FDEV(i).bdev) == BLK_ZONED_HM && 2341 !f2fs_sb_mounted_blkzoned(sbi->sb)) { 2342 f2fs_msg(sbi->sb, KERN_ERR, 2343 "Zoned block device feature not enabled\n"); 2344 return -EINVAL; 2345 } 2346 if (bdev_zoned_model(FDEV(i).bdev) != BLK_ZONED_NONE) { 2347 if (init_blkz_info(sbi, i)) { 2348 f2fs_msg(sbi->sb, KERN_ERR, 2349 "Failed to initialize F2FS blkzone information"); 2350 return -EINVAL; 2351 } 2352 if (max_devices == 1) 2353 break; 2354 f2fs_msg(sbi->sb, KERN_INFO, 2355 "Mount Device [%2d]: %20s, %8u, %8x - %8x (zone: %s)", 2356 i, FDEV(i).path, 2357 FDEV(i).total_segments, 2358 FDEV(i).start_blk, FDEV(i).end_blk, 2359 bdev_zoned_model(FDEV(i).bdev) == BLK_ZONED_HA ? 2360 "Host-aware" : "Host-managed"); 2361 continue; 2362 } 2363 #endif 2364 f2fs_msg(sbi->sb, KERN_INFO, 2365 "Mount Device [%2d]: %20s, %8u, %8x - %8x", 2366 i, FDEV(i).path, 2367 FDEV(i).total_segments, 2368 FDEV(i).start_blk, FDEV(i).end_blk); 2369 } 2370 f2fs_msg(sbi->sb, KERN_INFO, 2371 "IO Block Size: %8d KB", F2FS_IO_SIZE_KB(sbi)); 2372 return 0; 2373 } 2374 2375 static int f2fs_fill_super(struct super_block *sb, void *data, int silent) 2376 { 2377 struct f2fs_sb_info *sbi; 2378 struct f2fs_super_block *raw_super; 2379 struct inode *root; 2380 int err; 2381 bool retry = true, need_fsck = false; 2382 char *options = NULL; 2383 int recovery, i, valid_super_block; 2384 struct curseg_info *seg_i; 2385 2386 try_onemore: 2387 err = -EINVAL; 2388 raw_super = NULL; 2389 valid_super_block = -1; 2390 recovery = 0; 2391 2392 /* allocate memory for f2fs-specific super block info */ 2393 sbi = kzalloc(sizeof(struct f2fs_sb_info), GFP_KERNEL); 2394 if (!sbi) 2395 return -ENOMEM; 2396 2397 sbi->sb = sb; 2398 2399 /* Load the checksum driver */ 2400 sbi->s_chksum_driver = crypto_alloc_shash("crc32", 0, 0); 2401 if (IS_ERR(sbi->s_chksum_driver)) { 2402 f2fs_msg(sb, KERN_ERR, "Cannot load crc32 driver."); 2403 err = PTR_ERR(sbi->s_chksum_driver); 2404 sbi->s_chksum_driver = NULL; 2405 goto free_sbi; 2406 } 2407 2408 /* set a block size */ 2409 if (unlikely(!sb_set_blocksize(sb, F2FS_BLKSIZE))) { 2410 f2fs_msg(sb, KERN_ERR, "unable to set blocksize"); 2411 goto free_sbi; 2412 } 2413 2414 err = read_raw_super_block(sbi, &raw_super, &valid_super_block, 2415 &recovery); 2416 if (err) 2417 goto free_sbi; 2418 2419 sb->s_fs_info = sbi; 2420 sbi->raw_super = raw_super; 2421 2422 /* precompute checksum seed for metadata */ 2423 if (f2fs_sb_has_inode_chksum(sb)) 2424 sbi->s_chksum_seed = f2fs_chksum(sbi, ~0, raw_super->uuid, 2425 sizeof(raw_super->uuid)); 2426 2427 /* 2428 * The BLKZONED feature indicates that the drive was formatted with 2429 * zone alignment optimization. This is optional for host-aware 2430 * devices, but mandatory for host-managed zoned block devices. 2431 */ 2432 #ifndef CONFIG_BLK_DEV_ZONED 2433 if (f2fs_sb_mounted_blkzoned(sb)) { 2434 f2fs_msg(sb, KERN_ERR, 2435 "Zoned block device support is not enabled\n"); 2436 err = -EOPNOTSUPP; 2437 goto free_sb_buf; 2438 } 2439 #endif 2440 default_options(sbi); 2441 /* parse mount options */ 2442 options = kstrdup((const char *)data, GFP_KERNEL); 2443 if (data && !options) { 2444 err = -ENOMEM; 2445 goto free_sb_buf; 2446 } 2447 2448 err = parse_options(sb, options); 2449 if (err) 2450 goto free_options; 2451 2452 sbi->max_file_blocks = max_file_blocks(); 2453 sb->s_maxbytes = sbi->max_file_blocks << 2454 le32_to_cpu(raw_super->log_blocksize); 2455 sb->s_max_links = F2FS_LINK_MAX; 2456 get_random_bytes(&sbi->s_next_generation, sizeof(u32)); 2457 2458 #ifdef CONFIG_QUOTA 2459 sb->dq_op = &f2fs_quota_operations; 2460 if (f2fs_sb_has_quota_ino(sb)) 2461 sb->s_qcop = &dquot_quotactl_sysfile_ops; 2462 else 2463 sb->s_qcop = &f2fs_quotactl_ops; 2464 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ; 2465 #endif 2466 2467 sb->s_op = &f2fs_sops; 2468 #ifdef CONFIG_F2FS_FS_ENCRYPTION 2469 sb->s_cop = &f2fs_cryptops; 2470 #endif 2471 sb->s_xattr = f2fs_xattr_handlers; 2472 sb->s_export_op = &f2fs_export_ops; 2473 sb->s_magic = F2FS_SUPER_MAGIC; 2474 sb->s_time_gran = 1; 2475 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) | 2476 (test_opt(sbi, POSIX_ACL) ? SB_POSIXACL : 0); 2477 memcpy(&sb->s_uuid, raw_super->uuid, sizeof(raw_super->uuid)); 2478 2479 /* init f2fs-specific super block info */ 2480 sbi->valid_super_block = valid_super_block; 2481 mutex_init(&sbi->gc_mutex); 2482 mutex_init(&sbi->cp_mutex); 2483 init_rwsem(&sbi->node_write); 2484 init_rwsem(&sbi->node_change); 2485 2486 /* disallow all the data/node/meta page writes */ 2487 set_sbi_flag(sbi, SBI_POR_DOING); 2488 spin_lock_init(&sbi->stat_lock); 2489 2490 /* init iostat info */ 2491 spin_lock_init(&sbi->iostat_lock); 2492 sbi->iostat_enable = false; 2493 2494 for (i = 0; i < NR_PAGE_TYPE; i++) { 2495 int n = (i == META) ? 1: NR_TEMP_TYPE; 2496 int j; 2497 2498 sbi->write_io[i] = kmalloc(n * sizeof(struct f2fs_bio_info), 2499 GFP_KERNEL); 2500 if (!sbi->write_io[i]) { 2501 err = -ENOMEM; 2502 goto free_options; 2503 } 2504 2505 for (j = HOT; j < n; j++) { 2506 init_rwsem(&sbi->write_io[i][j].io_rwsem); 2507 sbi->write_io[i][j].sbi = sbi; 2508 sbi->write_io[i][j].bio = NULL; 2509 spin_lock_init(&sbi->write_io[i][j].io_lock); 2510 INIT_LIST_HEAD(&sbi->write_io[i][j].io_list); 2511 } 2512 } 2513 2514 init_rwsem(&sbi->cp_rwsem); 2515 init_waitqueue_head(&sbi->cp_wait); 2516 init_sb_info(sbi); 2517 2518 err = init_percpu_info(sbi); 2519 if (err) 2520 goto free_options; 2521 2522 if (F2FS_IO_SIZE(sbi) > 1) { 2523 sbi->write_io_dummy = 2524 mempool_create_page_pool(2 * (F2FS_IO_SIZE(sbi) - 1), 0); 2525 if (!sbi->write_io_dummy) { 2526 err = -ENOMEM; 2527 goto free_options; 2528 } 2529 } 2530 2531 /* get an inode for meta space */ 2532 sbi->meta_inode = f2fs_iget(sb, F2FS_META_INO(sbi)); 2533 if (IS_ERR(sbi->meta_inode)) { 2534 f2fs_msg(sb, KERN_ERR, "Failed to read F2FS meta data inode"); 2535 err = PTR_ERR(sbi->meta_inode); 2536 goto free_io_dummy; 2537 } 2538 2539 err = get_valid_checkpoint(sbi); 2540 if (err) { 2541 f2fs_msg(sb, KERN_ERR, "Failed to get valid F2FS checkpoint"); 2542 goto free_meta_inode; 2543 } 2544 2545 /* Initialize device list */ 2546 err = f2fs_scan_devices(sbi); 2547 if (err) { 2548 f2fs_msg(sb, KERN_ERR, "Failed to find devices"); 2549 goto free_devices; 2550 } 2551 2552 sbi->total_valid_node_count = 2553 le32_to_cpu(sbi->ckpt->valid_node_count); 2554 percpu_counter_set(&sbi->total_valid_inode_count, 2555 le32_to_cpu(sbi->ckpt->valid_inode_count)); 2556 sbi->user_block_count = le64_to_cpu(sbi->ckpt->user_block_count); 2557 sbi->total_valid_block_count = 2558 le64_to_cpu(sbi->ckpt->valid_block_count); 2559 sbi->last_valid_block_count = sbi->total_valid_block_count; 2560 sbi->reserved_blocks = 0; 2561 sbi->current_reserved_blocks = 0; 2562 2563 for (i = 0; i < NR_INODE_TYPE; i++) { 2564 INIT_LIST_HEAD(&sbi->inode_list[i]); 2565 spin_lock_init(&sbi->inode_lock[i]); 2566 } 2567 2568 init_extent_cache_info(sbi); 2569 2570 init_ino_entry_info(sbi); 2571 2572 /* setup f2fs internal modules */ 2573 err = build_segment_manager(sbi); 2574 if (err) { 2575 f2fs_msg(sb, KERN_ERR, 2576 "Failed to initialize F2FS segment manager"); 2577 goto free_sm; 2578 } 2579 err = build_node_manager(sbi); 2580 if (err) { 2581 f2fs_msg(sb, KERN_ERR, 2582 "Failed to initialize F2FS node manager"); 2583 goto free_nm; 2584 } 2585 2586 /* For write statistics */ 2587 if (sb->s_bdev->bd_part) 2588 sbi->sectors_written_start = 2589 (u64)part_stat_read(sb->s_bdev->bd_part, sectors[1]); 2590 2591 /* Read accumulated write IO statistics if exists */ 2592 seg_i = CURSEG_I(sbi, CURSEG_HOT_NODE); 2593 if (__exist_node_summaries(sbi)) 2594 sbi->kbytes_written = 2595 le64_to_cpu(seg_i->journal->info.kbytes_written); 2596 2597 build_gc_manager(sbi); 2598 2599 /* get an inode for node space */ 2600 sbi->node_inode = f2fs_iget(sb, F2FS_NODE_INO(sbi)); 2601 if (IS_ERR(sbi->node_inode)) { 2602 f2fs_msg(sb, KERN_ERR, "Failed to read node inode"); 2603 err = PTR_ERR(sbi->node_inode); 2604 goto free_nm; 2605 } 2606 2607 f2fs_join_shrinker(sbi); 2608 2609 err = f2fs_build_stats(sbi); 2610 if (err) 2611 goto free_nm; 2612 2613 /* read root inode and dentry */ 2614 root = f2fs_iget(sb, F2FS_ROOT_INO(sbi)); 2615 if (IS_ERR(root)) { 2616 f2fs_msg(sb, KERN_ERR, "Failed to read root inode"); 2617 err = PTR_ERR(root); 2618 goto free_node_inode; 2619 } 2620 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) { 2621 iput(root); 2622 err = -EINVAL; 2623 goto free_node_inode; 2624 } 2625 2626 sb->s_root = d_make_root(root); /* allocate root dentry */ 2627 if (!sb->s_root) { 2628 err = -ENOMEM; 2629 goto free_root_inode; 2630 } 2631 2632 err = f2fs_register_sysfs(sbi); 2633 if (err) 2634 goto free_root_inode; 2635 2636 #ifdef CONFIG_QUOTA 2637 /* 2638 * Turn on quotas which were not enabled for read-only mounts if 2639 * filesystem has quota feature, so that they are updated correctly. 2640 */ 2641 if (f2fs_sb_has_quota_ino(sb) && !sb_rdonly(sb)) { 2642 err = f2fs_enable_quotas(sb); 2643 if (err) { 2644 f2fs_msg(sb, KERN_ERR, 2645 "Cannot turn on quotas: error %d", err); 2646 goto free_sysfs; 2647 } 2648 } 2649 #endif 2650 /* if there are nt orphan nodes free them */ 2651 err = recover_orphan_inodes(sbi); 2652 if (err) 2653 goto free_meta; 2654 2655 /* recover fsynced data */ 2656 if (!test_opt(sbi, DISABLE_ROLL_FORWARD)) { 2657 /* 2658 * mount should be failed, when device has readonly mode, and 2659 * previous checkpoint was not done by clean system shutdown. 2660 */ 2661 if (bdev_read_only(sb->s_bdev) && 2662 !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG)) { 2663 err = -EROFS; 2664 goto free_meta; 2665 } 2666 2667 if (need_fsck) 2668 set_sbi_flag(sbi, SBI_NEED_FSCK); 2669 2670 if (!retry) 2671 goto skip_recovery; 2672 2673 err = recover_fsync_data(sbi, false); 2674 if (err < 0) { 2675 need_fsck = true; 2676 f2fs_msg(sb, KERN_ERR, 2677 "Cannot recover all fsync data errno=%d", err); 2678 goto free_meta; 2679 } 2680 } else { 2681 err = recover_fsync_data(sbi, true); 2682 2683 if (!f2fs_readonly(sb) && err > 0) { 2684 err = -EINVAL; 2685 f2fs_msg(sb, KERN_ERR, 2686 "Need to recover fsync data"); 2687 goto free_meta; 2688 } 2689 } 2690 skip_recovery: 2691 /* recover_fsync_data() cleared this already */ 2692 clear_sbi_flag(sbi, SBI_POR_DOING); 2693 2694 /* 2695 * If filesystem is not mounted as read-only then 2696 * do start the gc_thread. 2697 */ 2698 if (test_opt(sbi, BG_GC) && !f2fs_readonly(sb)) { 2699 /* After POR, we can run background GC thread.*/ 2700 err = start_gc_thread(sbi); 2701 if (err) 2702 goto free_meta; 2703 } 2704 kfree(options); 2705 2706 /* recover broken superblock */ 2707 if (recovery) { 2708 err = f2fs_commit_super(sbi, true); 2709 f2fs_msg(sb, KERN_INFO, 2710 "Try to recover %dth superblock, ret: %d", 2711 sbi->valid_super_block ? 1 : 2, err); 2712 } 2713 2714 f2fs_msg(sbi->sb, KERN_NOTICE, "Mounted with checkpoint version = %llx", 2715 cur_cp_version(F2FS_CKPT(sbi))); 2716 f2fs_update_time(sbi, CP_TIME); 2717 f2fs_update_time(sbi, REQ_TIME); 2718 return 0; 2719 2720 free_meta: 2721 #ifdef CONFIG_QUOTA 2722 if (f2fs_sb_has_quota_ino(sb) && !sb_rdonly(sb)) 2723 f2fs_quota_off_umount(sbi->sb); 2724 #endif 2725 f2fs_sync_inode_meta(sbi); 2726 /* 2727 * Some dirty meta pages can be produced by recover_orphan_inodes() 2728 * failed by EIO. Then, iput(node_inode) can trigger balance_fs_bg() 2729 * followed by write_checkpoint() through f2fs_write_node_pages(), which 2730 * falls into an infinite loop in sync_meta_pages(). 2731 */ 2732 truncate_inode_pages_final(META_MAPPING(sbi)); 2733 #ifdef CONFIG_QUOTA 2734 free_sysfs: 2735 #endif 2736 f2fs_unregister_sysfs(sbi); 2737 free_root_inode: 2738 dput(sb->s_root); 2739 sb->s_root = NULL; 2740 free_node_inode: 2741 truncate_inode_pages_final(NODE_MAPPING(sbi)); 2742 mutex_lock(&sbi->umount_mutex); 2743 release_ino_entry(sbi, true); 2744 f2fs_leave_shrinker(sbi); 2745 iput(sbi->node_inode); 2746 mutex_unlock(&sbi->umount_mutex); 2747 f2fs_destroy_stats(sbi); 2748 free_nm: 2749 destroy_node_manager(sbi); 2750 free_sm: 2751 destroy_segment_manager(sbi); 2752 free_devices: 2753 destroy_device_list(sbi); 2754 kfree(sbi->ckpt); 2755 free_meta_inode: 2756 make_bad_inode(sbi->meta_inode); 2757 iput(sbi->meta_inode); 2758 free_io_dummy: 2759 mempool_destroy(sbi->write_io_dummy); 2760 free_options: 2761 for (i = 0; i < NR_PAGE_TYPE; i++) 2762 kfree(sbi->write_io[i]); 2763 destroy_percpu_info(sbi); 2764 #ifdef CONFIG_QUOTA 2765 for (i = 0; i < MAXQUOTAS; i++) 2766 kfree(sbi->s_qf_names[i]); 2767 #endif 2768 kfree(options); 2769 free_sb_buf: 2770 kfree(raw_super); 2771 free_sbi: 2772 if (sbi->s_chksum_driver) 2773 crypto_free_shash(sbi->s_chksum_driver); 2774 kfree(sbi); 2775 2776 /* give only one another chance */ 2777 if (retry) { 2778 retry = false; 2779 shrink_dcache_sb(sb); 2780 goto try_onemore; 2781 } 2782 return err; 2783 } 2784 2785 static struct dentry *f2fs_mount(struct file_system_type *fs_type, int flags, 2786 const char *dev_name, void *data) 2787 { 2788 return mount_bdev(fs_type, flags, dev_name, data, f2fs_fill_super); 2789 } 2790 2791 static void kill_f2fs_super(struct super_block *sb) 2792 { 2793 if (sb->s_root) { 2794 set_sbi_flag(F2FS_SB(sb), SBI_IS_CLOSE); 2795 stop_gc_thread(F2FS_SB(sb)); 2796 stop_discard_thread(F2FS_SB(sb)); 2797 } 2798 kill_block_super(sb); 2799 } 2800 2801 static struct file_system_type f2fs_fs_type = { 2802 .owner = THIS_MODULE, 2803 .name = "f2fs", 2804 .mount = f2fs_mount, 2805 .kill_sb = kill_f2fs_super, 2806 .fs_flags = FS_REQUIRES_DEV, 2807 }; 2808 MODULE_ALIAS_FS("f2fs"); 2809 2810 static int __init init_inodecache(void) 2811 { 2812 f2fs_inode_cachep = kmem_cache_create("f2fs_inode_cache", 2813 sizeof(struct f2fs_inode_info), 0, 2814 SLAB_RECLAIM_ACCOUNT|SLAB_ACCOUNT, NULL); 2815 if (!f2fs_inode_cachep) 2816 return -ENOMEM; 2817 return 0; 2818 } 2819 2820 static void destroy_inodecache(void) 2821 { 2822 /* 2823 * Make sure all delayed rcu free inodes are flushed before we 2824 * destroy cache. 2825 */ 2826 rcu_barrier(); 2827 kmem_cache_destroy(f2fs_inode_cachep); 2828 } 2829 2830 static int __init init_f2fs_fs(void) 2831 { 2832 int err; 2833 2834 f2fs_build_trace_ios(); 2835 2836 err = init_inodecache(); 2837 if (err) 2838 goto fail; 2839 err = create_node_manager_caches(); 2840 if (err) 2841 goto free_inodecache; 2842 err = create_segment_manager_caches(); 2843 if (err) 2844 goto free_node_manager_caches; 2845 err = create_checkpoint_caches(); 2846 if (err) 2847 goto free_segment_manager_caches; 2848 err = create_extent_cache(); 2849 if (err) 2850 goto free_checkpoint_caches; 2851 err = f2fs_init_sysfs(); 2852 if (err) 2853 goto free_extent_cache; 2854 err = register_shrinker(&f2fs_shrinker_info); 2855 if (err) 2856 goto free_sysfs; 2857 err = register_filesystem(&f2fs_fs_type); 2858 if (err) 2859 goto free_shrinker; 2860 err = f2fs_create_root_stats(); 2861 if (err) 2862 goto free_filesystem; 2863 return 0; 2864 2865 free_filesystem: 2866 unregister_filesystem(&f2fs_fs_type); 2867 free_shrinker: 2868 unregister_shrinker(&f2fs_shrinker_info); 2869 free_sysfs: 2870 f2fs_exit_sysfs(); 2871 free_extent_cache: 2872 destroy_extent_cache(); 2873 free_checkpoint_caches: 2874 destroy_checkpoint_caches(); 2875 free_segment_manager_caches: 2876 destroy_segment_manager_caches(); 2877 free_node_manager_caches: 2878 destroy_node_manager_caches(); 2879 free_inodecache: 2880 destroy_inodecache(); 2881 fail: 2882 return err; 2883 } 2884 2885 static void __exit exit_f2fs_fs(void) 2886 { 2887 f2fs_destroy_root_stats(); 2888 unregister_filesystem(&f2fs_fs_type); 2889 unregister_shrinker(&f2fs_shrinker_info); 2890 f2fs_exit_sysfs(); 2891 destroy_extent_cache(); 2892 destroy_checkpoint_caches(); 2893 destroy_segment_manager_caches(); 2894 destroy_node_manager_caches(); 2895 destroy_inodecache(); 2896 f2fs_destroy_trace_ios(); 2897 } 2898 2899 module_init(init_f2fs_fs) 2900 module_exit(exit_f2fs_fs) 2901 2902 MODULE_AUTHOR("Samsung Electronics's Praesto Team"); 2903 MODULE_DESCRIPTION("Flash Friendly File System"); 2904 MODULE_LICENSE("GPL"); 2905 2906