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/f2fs_fs.h> 26 #include <linux/sysfs.h> 27 28 #include "f2fs.h" 29 #include "node.h" 30 #include "segment.h" 31 #include "xattr.h" 32 #include "gc.h" 33 #include "trace.h" 34 35 #define CREATE_TRACE_POINTS 36 #include <trace/events/f2fs.h> 37 38 static struct proc_dir_entry *f2fs_proc_root; 39 static struct kmem_cache *f2fs_inode_cachep; 40 static struct kset *f2fs_kset; 41 42 /* f2fs-wide shrinker description */ 43 static struct shrinker f2fs_shrinker_info = { 44 .scan_objects = f2fs_shrink_scan, 45 .count_objects = f2fs_shrink_count, 46 .seeks = DEFAULT_SEEKS, 47 }; 48 49 enum { 50 Opt_gc_background, 51 Opt_disable_roll_forward, 52 Opt_norecovery, 53 Opt_discard, 54 Opt_noheap, 55 Opt_user_xattr, 56 Opt_nouser_xattr, 57 Opt_acl, 58 Opt_noacl, 59 Opt_active_logs, 60 Opt_disable_ext_identify, 61 Opt_inline_xattr, 62 Opt_inline_data, 63 Opt_inline_dentry, 64 Opt_flush_merge, 65 Opt_nobarrier, 66 Opt_fastboot, 67 Opt_extent_cache, 68 Opt_noextent_cache, 69 Opt_noinline_data, 70 Opt_err, 71 }; 72 73 static match_table_t f2fs_tokens = { 74 {Opt_gc_background, "background_gc=%s"}, 75 {Opt_disable_roll_forward, "disable_roll_forward"}, 76 {Opt_norecovery, "norecovery"}, 77 {Opt_discard, "discard"}, 78 {Opt_noheap, "no_heap"}, 79 {Opt_user_xattr, "user_xattr"}, 80 {Opt_nouser_xattr, "nouser_xattr"}, 81 {Opt_acl, "acl"}, 82 {Opt_noacl, "noacl"}, 83 {Opt_active_logs, "active_logs=%u"}, 84 {Opt_disable_ext_identify, "disable_ext_identify"}, 85 {Opt_inline_xattr, "inline_xattr"}, 86 {Opt_inline_data, "inline_data"}, 87 {Opt_inline_dentry, "inline_dentry"}, 88 {Opt_flush_merge, "flush_merge"}, 89 {Opt_nobarrier, "nobarrier"}, 90 {Opt_fastboot, "fastboot"}, 91 {Opt_extent_cache, "extent_cache"}, 92 {Opt_noextent_cache, "noextent_cache"}, 93 {Opt_noinline_data, "noinline_data"}, 94 {Opt_err, NULL}, 95 }; 96 97 /* Sysfs support for f2fs */ 98 enum { 99 GC_THREAD, /* struct f2fs_gc_thread */ 100 SM_INFO, /* struct f2fs_sm_info */ 101 NM_INFO, /* struct f2fs_nm_info */ 102 F2FS_SBI, /* struct f2fs_sb_info */ 103 }; 104 105 struct f2fs_attr { 106 struct attribute attr; 107 ssize_t (*show)(struct f2fs_attr *, struct f2fs_sb_info *, char *); 108 ssize_t (*store)(struct f2fs_attr *, struct f2fs_sb_info *, 109 const char *, size_t); 110 int struct_type; 111 int offset; 112 }; 113 114 static unsigned char *__struct_ptr(struct f2fs_sb_info *sbi, int struct_type) 115 { 116 if (struct_type == GC_THREAD) 117 return (unsigned char *)sbi->gc_thread; 118 else if (struct_type == SM_INFO) 119 return (unsigned char *)SM_I(sbi); 120 else if (struct_type == NM_INFO) 121 return (unsigned char *)NM_I(sbi); 122 else if (struct_type == F2FS_SBI) 123 return (unsigned char *)sbi; 124 return NULL; 125 } 126 127 static ssize_t f2fs_sbi_show(struct f2fs_attr *a, 128 struct f2fs_sb_info *sbi, char *buf) 129 { 130 unsigned char *ptr = NULL; 131 unsigned int *ui; 132 133 ptr = __struct_ptr(sbi, a->struct_type); 134 if (!ptr) 135 return -EINVAL; 136 137 ui = (unsigned int *)(ptr + a->offset); 138 139 return snprintf(buf, PAGE_SIZE, "%u\n", *ui); 140 } 141 142 static ssize_t f2fs_sbi_store(struct f2fs_attr *a, 143 struct f2fs_sb_info *sbi, 144 const char *buf, size_t count) 145 { 146 unsigned char *ptr; 147 unsigned long t; 148 unsigned int *ui; 149 ssize_t ret; 150 151 ptr = __struct_ptr(sbi, a->struct_type); 152 if (!ptr) 153 return -EINVAL; 154 155 ui = (unsigned int *)(ptr + a->offset); 156 157 ret = kstrtoul(skip_spaces(buf), 0, &t); 158 if (ret < 0) 159 return ret; 160 *ui = t; 161 return count; 162 } 163 164 static ssize_t f2fs_attr_show(struct kobject *kobj, 165 struct attribute *attr, char *buf) 166 { 167 struct f2fs_sb_info *sbi = container_of(kobj, struct f2fs_sb_info, 168 s_kobj); 169 struct f2fs_attr *a = container_of(attr, struct f2fs_attr, attr); 170 171 return a->show ? a->show(a, sbi, buf) : 0; 172 } 173 174 static ssize_t f2fs_attr_store(struct kobject *kobj, struct attribute *attr, 175 const char *buf, size_t len) 176 { 177 struct f2fs_sb_info *sbi = container_of(kobj, struct f2fs_sb_info, 178 s_kobj); 179 struct f2fs_attr *a = container_of(attr, struct f2fs_attr, attr); 180 181 return a->store ? a->store(a, sbi, buf, len) : 0; 182 } 183 184 static void f2fs_sb_release(struct kobject *kobj) 185 { 186 struct f2fs_sb_info *sbi = container_of(kobj, struct f2fs_sb_info, 187 s_kobj); 188 complete(&sbi->s_kobj_unregister); 189 } 190 191 #define F2FS_ATTR_OFFSET(_struct_type, _name, _mode, _show, _store, _offset) \ 192 static struct f2fs_attr f2fs_attr_##_name = { \ 193 .attr = {.name = __stringify(_name), .mode = _mode }, \ 194 .show = _show, \ 195 .store = _store, \ 196 .struct_type = _struct_type, \ 197 .offset = _offset \ 198 } 199 200 #define F2FS_RW_ATTR(struct_type, struct_name, name, elname) \ 201 F2FS_ATTR_OFFSET(struct_type, name, 0644, \ 202 f2fs_sbi_show, f2fs_sbi_store, \ 203 offsetof(struct struct_name, elname)) 204 205 F2FS_RW_ATTR(GC_THREAD, f2fs_gc_kthread, gc_min_sleep_time, min_sleep_time); 206 F2FS_RW_ATTR(GC_THREAD, f2fs_gc_kthread, gc_max_sleep_time, max_sleep_time); 207 F2FS_RW_ATTR(GC_THREAD, f2fs_gc_kthread, gc_no_gc_sleep_time, no_gc_sleep_time); 208 F2FS_RW_ATTR(GC_THREAD, f2fs_gc_kthread, gc_idle, gc_idle); 209 F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, reclaim_segments, rec_prefree_segments); 210 F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, max_small_discards, max_discards); 211 F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, batched_trim_sections, trim_sections); 212 F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, ipu_policy, ipu_policy); 213 F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, min_ipu_util, min_ipu_util); 214 F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, min_fsync_blocks, min_fsync_blocks); 215 F2FS_RW_ATTR(NM_INFO, f2fs_nm_info, ram_thresh, ram_thresh); 216 F2FS_RW_ATTR(NM_INFO, f2fs_nm_info, ra_nid_pages, ra_nid_pages); 217 F2FS_RW_ATTR(F2FS_SBI, f2fs_sb_info, max_victim_search, max_victim_search); 218 F2FS_RW_ATTR(F2FS_SBI, f2fs_sb_info, dir_level, dir_level); 219 F2FS_RW_ATTR(F2FS_SBI, f2fs_sb_info, cp_interval, cp_interval); 220 221 #define ATTR_LIST(name) (&f2fs_attr_##name.attr) 222 static struct attribute *f2fs_attrs[] = { 223 ATTR_LIST(gc_min_sleep_time), 224 ATTR_LIST(gc_max_sleep_time), 225 ATTR_LIST(gc_no_gc_sleep_time), 226 ATTR_LIST(gc_idle), 227 ATTR_LIST(reclaim_segments), 228 ATTR_LIST(max_small_discards), 229 ATTR_LIST(batched_trim_sections), 230 ATTR_LIST(ipu_policy), 231 ATTR_LIST(min_ipu_util), 232 ATTR_LIST(min_fsync_blocks), 233 ATTR_LIST(max_victim_search), 234 ATTR_LIST(dir_level), 235 ATTR_LIST(ram_thresh), 236 ATTR_LIST(ra_nid_pages), 237 ATTR_LIST(cp_interval), 238 NULL, 239 }; 240 241 static const struct sysfs_ops f2fs_attr_ops = { 242 .show = f2fs_attr_show, 243 .store = f2fs_attr_store, 244 }; 245 246 static struct kobj_type f2fs_ktype = { 247 .default_attrs = f2fs_attrs, 248 .sysfs_ops = &f2fs_attr_ops, 249 .release = f2fs_sb_release, 250 }; 251 252 void f2fs_msg(struct super_block *sb, const char *level, const char *fmt, ...) 253 { 254 struct va_format vaf; 255 va_list args; 256 257 va_start(args, fmt); 258 vaf.fmt = fmt; 259 vaf.va = &args; 260 printk("%sF2FS-fs (%s): %pV\n", level, sb->s_id, &vaf); 261 va_end(args); 262 } 263 264 static void init_once(void *foo) 265 { 266 struct f2fs_inode_info *fi = (struct f2fs_inode_info *) foo; 267 268 inode_init_once(&fi->vfs_inode); 269 } 270 271 static int parse_options(struct super_block *sb, char *options) 272 { 273 struct f2fs_sb_info *sbi = F2FS_SB(sb); 274 struct request_queue *q; 275 substring_t args[MAX_OPT_ARGS]; 276 char *p, *name; 277 int arg = 0; 278 279 if (!options) 280 return 0; 281 282 while ((p = strsep(&options, ",")) != NULL) { 283 int token; 284 if (!*p) 285 continue; 286 /* 287 * Initialize args struct so we know whether arg was 288 * found; some options take optional arguments. 289 */ 290 args[0].to = args[0].from = NULL; 291 token = match_token(p, f2fs_tokens, args); 292 293 switch (token) { 294 case Opt_gc_background: 295 name = match_strdup(&args[0]); 296 297 if (!name) 298 return -ENOMEM; 299 if (strlen(name) == 2 && !strncmp(name, "on", 2)) { 300 set_opt(sbi, BG_GC); 301 clear_opt(sbi, FORCE_FG_GC); 302 } else if (strlen(name) == 3 && !strncmp(name, "off", 3)) { 303 clear_opt(sbi, BG_GC); 304 clear_opt(sbi, FORCE_FG_GC); 305 } else if (strlen(name) == 4 && !strncmp(name, "sync", 4)) { 306 set_opt(sbi, BG_GC); 307 set_opt(sbi, FORCE_FG_GC); 308 } else { 309 kfree(name); 310 return -EINVAL; 311 } 312 kfree(name); 313 break; 314 case Opt_disable_roll_forward: 315 set_opt(sbi, DISABLE_ROLL_FORWARD); 316 break; 317 case Opt_norecovery: 318 /* this option mounts f2fs with ro */ 319 set_opt(sbi, DISABLE_ROLL_FORWARD); 320 if (!f2fs_readonly(sb)) 321 return -EINVAL; 322 break; 323 case Opt_discard: 324 q = bdev_get_queue(sb->s_bdev); 325 if (blk_queue_discard(q)) { 326 set_opt(sbi, DISCARD); 327 } else { 328 f2fs_msg(sb, KERN_WARNING, 329 "mounting with \"discard\" option, but " 330 "the device does not support discard"); 331 } 332 break; 333 case Opt_noheap: 334 set_opt(sbi, NOHEAP); 335 break; 336 #ifdef CONFIG_F2FS_FS_XATTR 337 case Opt_user_xattr: 338 set_opt(sbi, XATTR_USER); 339 break; 340 case Opt_nouser_xattr: 341 clear_opt(sbi, XATTR_USER); 342 break; 343 case Opt_inline_xattr: 344 set_opt(sbi, INLINE_XATTR); 345 break; 346 #else 347 case Opt_user_xattr: 348 f2fs_msg(sb, KERN_INFO, 349 "user_xattr options not supported"); 350 break; 351 case Opt_nouser_xattr: 352 f2fs_msg(sb, KERN_INFO, 353 "nouser_xattr options not supported"); 354 break; 355 case Opt_inline_xattr: 356 f2fs_msg(sb, KERN_INFO, 357 "inline_xattr options not supported"); 358 break; 359 #endif 360 #ifdef CONFIG_F2FS_FS_POSIX_ACL 361 case Opt_acl: 362 set_opt(sbi, POSIX_ACL); 363 break; 364 case Opt_noacl: 365 clear_opt(sbi, POSIX_ACL); 366 break; 367 #else 368 case Opt_acl: 369 f2fs_msg(sb, KERN_INFO, "acl options not supported"); 370 break; 371 case Opt_noacl: 372 f2fs_msg(sb, KERN_INFO, "noacl options not supported"); 373 break; 374 #endif 375 case Opt_active_logs: 376 if (args->from && match_int(args, &arg)) 377 return -EINVAL; 378 if (arg != 2 && arg != 4 && arg != NR_CURSEG_TYPE) 379 return -EINVAL; 380 sbi->active_logs = arg; 381 break; 382 case Opt_disable_ext_identify: 383 set_opt(sbi, DISABLE_EXT_IDENTIFY); 384 break; 385 case Opt_inline_data: 386 set_opt(sbi, INLINE_DATA); 387 break; 388 case Opt_inline_dentry: 389 set_opt(sbi, INLINE_DENTRY); 390 break; 391 case Opt_flush_merge: 392 set_opt(sbi, FLUSH_MERGE); 393 break; 394 case Opt_nobarrier: 395 set_opt(sbi, NOBARRIER); 396 break; 397 case Opt_fastboot: 398 set_opt(sbi, FASTBOOT); 399 break; 400 case Opt_extent_cache: 401 set_opt(sbi, EXTENT_CACHE); 402 break; 403 case Opt_noextent_cache: 404 clear_opt(sbi, EXTENT_CACHE); 405 break; 406 case Opt_noinline_data: 407 clear_opt(sbi, INLINE_DATA); 408 break; 409 default: 410 f2fs_msg(sb, KERN_ERR, 411 "Unrecognized mount option \"%s\" or missing value", 412 p); 413 return -EINVAL; 414 } 415 } 416 return 0; 417 } 418 419 static struct inode *f2fs_alloc_inode(struct super_block *sb) 420 { 421 struct f2fs_inode_info *fi; 422 423 fi = kmem_cache_alloc(f2fs_inode_cachep, GFP_F2FS_ZERO); 424 if (!fi) 425 return NULL; 426 427 init_once((void *) fi); 428 429 /* Initialize f2fs-specific inode info */ 430 fi->vfs_inode.i_version = 1; 431 atomic_set(&fi->dirty_pages, 0); 432 fi->i_current_depth = 1; 433 fi->i_advise = 0; 434 init_rwsem(&fi->i_sem); 435 INIT_LIST_HEAD(&fi->dirty_list); 436 INIT_LIST_HEAD(&fi->inmem_pages); 437 mutex_init(&fi->inmem_lock); 438 439 set_inode_flag(fi, FI_NEW_INODE); 440 441 if (test_opt(F2FS_SB(sb), INLINE_XATTR)) 442 set_inode_flag(fi, FI_INLINE_XATTR); 443 444 /* Will be used by directory only */ 445 fi->i_dir_level = F2FS_SB(sb)->dir_level; 446 447 #ifdef CONFIG_F2FS_FS_ENCRYPTION 448 fi->i_crypt_info = NULL; 449 #endif 450 return &fi->vfs_inode; 451 } 452 453 static int f2fs_drop_inode(struct inode *inode) 454 { 455 /* 456 * This is to avoid a deadlock condition like below. 457 * writeback_single_inode(inode) 458 * - f2fs_write_data_page 459 * - f2fs_gc -> iput -> evict 460 * - inode_wait_for_writeback(inode) 461 */ 462 if (!inode_unhashed(inode) && inode->i_state & I_SYNC) { 463 if (!inode->i_nlink && !is_bad_inode(inode)) { 464 /* to avoid evict_inode call simultaneously */ 465 atomic_inc(&inode->i_count); 466 spin_unlock(&inode->i_lock); 467 468 /* some remained atomic pages should discarded */ 469 if (f2fs_is_atomic_file(inode)) 470 commit_inmem_pages(inode, true); 471 472 /* should remain fi->extent_tree for writepage */ 473 f2fs_destroy_extent_node(inode); 474 475 sb_start_intwrite(inode->i_sb); 476 i_size_write(inode, 0); 477 478 if (F2FS_HAS_BLOCKS(inode)) 479 f2fs_truncate(inode, true); 480 481 sb_end_intwrite(inode->i_sb); 482 483 #ifdef CONFIG_F2FS_FS_ENCRYPTION 484 if (F2FS_I(inode)->i_crypt_info) 485 f2fs_free_encryption_info(inode, 486 F2FS_I(inode)->i_crypt_info); 487 #endif 488 spin_lock(&inode->i_lock); 489 atomic_dec(&inode->i_count); 490 } 491 return 0; 492 } 493 return generic_drop_inode(inode); 494 } 495 496 /* 497 * f2fs_dirty_inode() is called from __mark_inode_dirty() 498 * 499 * We should call set_dirty_inode to write the dirty inode through write_inode. 500 */ 501 static void f2fs_dirty_inode(struct inode *inode, int flags) 502 { 503 set_inode_flag(F2FS_I(inode), FI_DIRTY_INODE); 504 } 505 506 static void f2fs_i_callback(struct rcu_head *head) 507 { 508 struct inode *inode = container_of(head, struct inode, i_rcu); 509 kmem_cache_free(f2fs_inode_cachep, F2FS_I(inode)); 510 } 511 512 static void f2fs_destroy_inode(struct inode *inode) 513 { 514 call_rcu(&inode->i_rcu, f2fs_i_callback); 515 } 516 517 static void f2fs_put_super(struct super_block *sb) 518 { 519 struct f2fs_sb_info *sbi = F2FS_SB(sb); 520 521 if (sbi->s_proc) { 522 remove_proc_entry("segment_info", sbi->s_proc); 523 remove_proc_entry(sb->s_id, f2fs_proc_root); 524 } 525 kobject_del(&sbi->s_kobj); 526 527 stop_gc_thread(sbi); 528 529 /* prevent remaining shrinker jobs */ 530 mutex_lock(&sbi->umount_mutex); 531 532 /* 533 * We don't need to do checkpoint when superblock is clean. 534 * But, the previous checkpoint was not done by umount, it needs to do 535 * clean checkpoint again. 536 */ 537 if (is_sbi_flag_set(sbi, SBI_IS_DIRTY) || 538 !is_set_ckpt_flags(F2FS_CKPT(sbi), CP_UMOUNT_FLAG)) { 539 struct cp_control cpc = { 540 .reason = CP_UMOUNT, 541 }; 542 write_checkpoint(sbi, &cpc); 543 } 544 545 /* write_checkpoint can update stat informaion */ 546 f2fs_destroy_stats(sbi); 547 548 /* 549 * normally superblock is clean, so we need to release this. 550 * In addition, EIO will skip do checkpoint, we need this as well. 551 */ 552 release_ino_entry(sbi); 553 release_discard_addrs(sbi); 554 555 f2fs_leave_shrinker(sbi); 556 mutex_unlock(&sbi->umount_mutex); 557 558 iput(sbi->node_inode); 559 iput(sbi->meta_inode); 560 561 /* destroy f2fs internal modules */ 562 destroy_node_manager(sbi); 563 destroy_segment_manager(sbi); 564 565 kfree(sbi->ckpt); 566 kobject_put(&sbi->s_kobj); 567 wait_for_completion(&sbi->s_kobj_unregister); 568 569 sb->s_fs_info = NULL; 570 kfree(sbi->raw_super); 571 kfree(sbi); 572 } 573 574 int f2fs_sync_fs(struct super_block *sb, int sync) 575 { 576 struct f2fs_sb_info *sbi = F2FS_SB(sb); 577 578 trace_f2fs_sync_fs(sb, sync); 579 580 if (sync) { 581 struct cp_control cpc; 582 583 cpc.reason = __get_cp_reason(sbi); 584 585 mutex_lock(&sbi->gc_mutex); 586 write_checkpoint(sbi, &cpc); 587 mutex_unlock(&sbi->gc_mutex); 588 } else { 589 f2fs_balance_fs(sbi); 590 } 591 f2fs_trace_ios(NULL, 1); 592 593 return 0; 594 } 595 596 static int f2fs_freeze(struct super_block *sb) 597 { 598 int err; 599 600 if (f2fs_readonly(sb)) 601 return 0; 602 603 err = f2fs_sync_fs(sb, 1); 604 return err; 605 } 606 607 static int f2fs_unfreeze(struct super_block *sb) 608 { 609 return 0; 610 } 611 612 static int f2fs_statfs(struct dentry *dentry, struct kstatfs *buf) 613 { 614 struct super_block *sb = dentry->d_sb; 615 struct f2fs_sb_info *sbi = F2FS_SB(sb); 616 u64 id = huge_encode_dev(sb->s_bdev->bd_dev); 617 block_t total_count, user_block_count, start_count, ovp_count; 618 619 total_count = le64_to_cpu(sbi->raw_super->block_count); 620 user_block_count = sbi->user_block_count; 621 start_count = le32_to_cpu(sbi->raw_super->segment0_blkaddr); 622 ovp_count = SM_I(sbi)->ovp_segments << sbi->log_blocks_per_seg; 623 buf->f_type = F2FS_SUPER_MAGIC; 624 buf->f_bsize = sbi->blocksize; 625 626 buf->f_blocks = total_count - start_count; 627 buf->f_bfree = buf->f_blocks - valid_user_blocks(sbi) - ovp_count; 628 buf->f_bavail = user_block_count - valid_user_blocks(sbi); 629 630 buf->f_files = sbi->total_node_count - F2FS_RESERVED_NODE_NUM; 631 buf->f_ffree = buf->f_files - valid_inode_count(sbi); 632 633 buf->f_namelen = F2FS_NAME_LEN; 634 buf->f_fsid.val[0] = (u32)id; 635 buf->f_fsid.val[1] = (u32)(id >> 32); 636 637 return 0; 638 } 639 640 static int f2fs_show_options(struct seq_file *seq, struct dentry *root) 641 { 642 struct f2fs_sb_info *sbi = F2FS_SB(root->d_sb); 643 644 if (!f2fs_readonly(sbi->sb) && test_opt(sbi, BG_GC)) { 645 if (test_opt(sbi, FORCE_FG_GC)) 646 seq_printf(seq, ",background_gc=%s", "sync"); 647 else 648 seq_printf(seq, ",background_gc=%s", "on"); 649 } else { 650 seq_printf(seq, ",background_gc=%s", "off"); 651 } 652 if (test_opt(sbi, DISABLE_ROLL_FORWARD)) 653 seq_puts(seq, ",disable_roll_forward"); 654 if (test_opt(sbi, DISCARD)) 655 seq_puts(seq, ",discard"); 656 if (test_opt(sbi, NOHEAP)) 657 seq_puts(seq, ",no_heap_alloc"); 658 #ifdef CONFIG_F2FS_FS_XATTR 659 if (test_opt(sbi, XATTR_USER)) 660 seq_puts(seq, ",user_xattr"); 661 else 662 seq_puts(seq, ",nouser_xattr"); 663 if (test_opt(sbi, INLINE_XATTR)) 664 seq_puts(seq, ",inline_xattr"); 665 #endif 666 #ifdef CONFIG_F2FS_FS_POSIX_ACL 667 if (test_opt(sbi, POSIX_ACL)) 668 seq_puts(seq, ",acl"); 669 else 670 seq_puts(seq, ",noacl"); 671 #endif 672 if (test_opt(sbi, DISABLE_EXT_IDENTIFY)) 673 seq_puts(seq, ",disable_ext_identify"); 674 if (test_opt(sbi, INLINE_DATA)) 675 seq_puts(seq, ",inline_data"); 676 else 677 seq_puts(seq, ",noinline_data"); 678 if (test_opt(sbi, INLINE_DENTRY)) 679 seq_puts(seq, ",inline_dentry"); 680 if (!f2fs_readonly(sbi->sb) && test_opt(sbi, FLUSH_MERGE)) 681 seq_puts(seq, ",flush_merge"); 682 if (test_opt(sbi, NOBARRIER)) 683 seq_puts(seq, ",nobarrier"); 684 if (test_opt(sbi, FASTBOOT)) 685 seq_puts(seq, ",fastboot"); 686 if (test_opt(sbi, EXTENT_CACHE)) 687 seq_puts(seq, ",extent_cache"); 688 else 689 seq_puts(seq, ",noextent_cache"); 690 seq_printf(seq, ",active_logs=%u", sbi->active_logs); 691 692 return 0; 693 } 694 695 static int segment_info_seq_show(struct seq_file *seq, void *offset) 696 { 697 struct super_block *sb = seq->private; 698 struct f2fs_sb_info *sbi = F2FS_SB(sb); 699 unsigned int total_segs = 700 le32_to_cpu(sbi->raw_super->segment_count_main); 701 int i; 702 703 seq_puts(seq, "format: segment_type|valid_blocks\n" 704 "segment_type(0:HD, 1:WD, 2:CD, 3:HN, 4:WN, 5:CN)\n"); 705 706 for (i = 0; i < total_segs; i++) { 707 struct seg_entry *se = get_seg_entry(sbi, i); 708 709 if ((i % 10) == 0) 710 seq_printf(seq, "%-10d", i); 711 seq_printf(seq, "%d|%-3u", se->type, 712 get_valid_blocks(sbi, i, 1)); 713 if ((i % 10) == 9 || i == (total_segs - 1)) 714 seq_putc(seq, '\n'); 715 else 716 seq_putc(seq, ' '); 717 } 718 719 return 0; 720 } 721 722 static int segment_info_open_fs(struct inode *inode, struct file *file) 723 { 724 return single_open(file, segment_info_seq_show, PDE_DATA(inode)); 725 } 726 727 static const struct file_operations f2fs_seq_segment_info_fops = { 728 .owner = THIS_MODULE, 729 .open = segment_info_open_fs, 730 .read = seq_read, 731 .llseek = seq_lseek, 732 .release = single_release, 733 }; 734 735 static void default_options(struct f2fs_sb_info *sbi) 736 { 737 /* init some FS parameters */ 738 sbi->active_logs = NR_CURSEG_TYPE; 739 740 set_opt(sbi, BG_GC); 741 set_opt(sbi, INLINE_DATA); 742 set_opt(sbi, EXTENT_CACHE); 743 744 #ifdef CONFIG_F2FS_FS_XATTR 745 set_opt(sbi, XATTR_USER); 746 #endif 747 #ifdef CONFIG_F2FS_FS_POSIX_ACL 748 set_opt(sbi, POSIX_ACL); 749 #endif 750 } 751 752 static int f2fs_remount(struct super_block *sb, int *flags, char *data) 753 { 754 struct f2fs_sb_info *sbi = F2FS_SB(sb); 755 struct f2fs_mount_info org_mount_opt; 756 int err, active_logs; 757 bool need_restart_gc = false; 758 bool need_stop_gc = false; 759 bool no_extent_cache = !test_opt(sbi, EXTENT_CACHE); 760 761 sync_filesystem(sb); 762 763 /* 764 * Save the old mount options in case we 765 * need to restore them. 766 */ 767 org_mount_opt = sbi->mount_opt; 768 active_logs = sbi->active_logs; 769 770 sbi->mount_opt.opt = 0; 771 default_options(sbi); 772 773 /* parse mount options */ 774 err = parse_options(sb, data); 775 if (err) 776 goto restore_opts; 777 778 /* 779 * Previous and new state of filesystem is RO, 780 * so skip checking GC and FLUSH_MERGE conditions. 781 */ 782 if (f2fs_readonly(sb) && (*flags & MS_RDONLY)) 783 goto skip; 784 785 /* disallow enable/disable extent_cache dynamically */ 786 if (no_extent_cache == !!test_opt(sbi, EXTENT_CACHE)) { 787 err = -EINVAL; 788 f2fs_msg(sbi->sb, KERN_WARNING, 789 "switch extent_cache option is not allowed"); 790 goto restore_opts; 791 } 792 793 /* 794 * We stop the GC thread if FS is mounted as RO 795 * or if background_gc = off is passed in mount 796 * option. Also sync the filesystem. 797 */ 798 if ((*flags & MS_RDONLY) || !test_opt(sbi, BG_GC)) { 799 if (sbi->gc_thread) { 800 stop_gc_thread(sbi); 801 f2fs_sync_fs(sb, 1); 802 need_restart_gc = true; 803 } 804 } else if (!sbi->gc_thread) { 805 err = start_gc_thread(sbi); 806 if (err) 807 goto restore_opts; 808 need_stop_gc = true; 809 } 810 811 /* 812 * We stop issue flush thread if FS is mounted as RO 813 * or if flush_merge is not passed in mount option. 814 */ 815 if ((*flags & MS_RDONLY) || !test_opt(sbi, FLUSH_MERGE)) { 816 destroy_flush_cmd_control(sbi); 817 } else if (!SM_I(sbi)->cmd_control_info) { 818 err = create_flush_cmd_control(sbi); 819 if (err) 820 goto restore_gc; 821 } 822 skip: 823 /* Update the POSIXACL Flag */ 824 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) | 825 (test_opt(sbi, POSIX_ACL) ? MS_POSIXACL : 0); 826 return 0; 827 restore_gc: 828 if (need_restart_gc) { 829 if (start_gc_thread(sbi)) 830 f2fs_msg(sbi->sb, KERN_WARNING, 831 "background gc thread has stopped"); 832 } else if (need_stop_gc) { 833 stop_gc_thread(sbi); 834 } 835 restore_opts: 836 sbi->mount_opt = org_mount_opt; 837 sbi->active_logs = active_logs; 838 return err; 839 } 840 841 static struct super_operations f2fs_sops = { 842 .alloc_inode = f2fs_alloc_inode, 843 .drop_inode = f2fs_drop_inode, 844 .destroy_inode = f2fs_destroy_inode, 845 .write_inode = f2fs_write_inode, 846 .dirty_inode = f2fs_dirty_inode, 847 .show_options = f2fs_show_options, 848 .evict_inode = f2fs_evict_inode, 849 .put_super = f2fs_put_super, 850 .sync_fs = f2fs_sync_fs, 851 .freeze_fs = f2fs_freeze, 852 .unfreeze_fs = f2fs_unfreeze, 853 .statfs = f2fs_statfs, 854 .remount_fs = f2fs_remount, 855 }; 856 857 static struct inode *f2fs_nfs_get_inode(struct super_block *sb, 858 u64 ino, u32 generation) 859 { 860 struct f2fs_sb_info *sbi = F2FS_SB(sb); 861 struct inode *inode; 862 863 if (check_nid_range(sbi, ino)) 864 return ERR_PTR(-ESTALE); 865 866 /* 867 * f2fs_iget isn't quite right if the inode is currently unallocated! 868 * However f2fs_iget currently does appropriate checks to handle stale 869 * inodes so everything is OK. 870 */ 871 inode = f2fs_iget(sb, ino); 872 if (IS_ERR(inode)) 873 return ERR_CAST(inode); 874 if (unlikely(generation && inode->i_generation != generation)) { 875 /* we didn't find the right inode.. */ 876 iput(inode); 877 return ERR_PTR(-ESTALE); 878 } 879 return inode; 880 } 881 882 static struct dentry *f2fs_fh_to_dentry(struct super_block *sb, struct fid *fid, 883 int fh_len, int fh_type) 884 { 885 return generic_fh_to_dentry(sb, fid, fh_len, fh_type, 886 f2fs_nfs_get_inode); 887 } 888 889 static struct dentry *f2fs_fh_to_parent(struct super_block *sb, struct fid *fid, 890 int fh_len, int fh_type) 891 { 892 return generic_fh_to_parent(sb, fid, fh_len, fh_type, 893 f2fs_nfs_get_inode); 894 } 895 896 static const struct export_operations f2fs_export_ops = { 897 .fh_to_dentry = f2fs_fh_to_dentry, 898 .fh_to_parent = f2fs_fh_to_parent, 899 .get_parent = f2fs_get_parent, 900 }; 901 902 static loff_t max_file_size(unsigned bits) 903 { 904 loff_t result = (DEF_ADDRS_PER_INODE - F2FS_INLINE_XATTR_ADDRS); 905 loff_t leaf_count = ADDRS_PER_BLOCK; 906 907 /* two direct node blocks */ 908 result += (leaf_count * 2); 909 910 /* two indirect node blocks */ 911 leaf_count *= NIDS_PER_BLOCK; 912 result += (leaf_count * 2); 913 914 /* one double indirect node block */ 915 leaf_count *= NIDS_PER_BLOCK; 916 result += leaf_count; 917 918 result <<= bits; 919 return result; 920 } 921 922 static inline bool sanity_check_area_boundary(struct super_block *sb, 923 struct f2fs_super_block *raw_super) 924 { 925 u32 segment0_blkaddr = le32_to_cpu(raw_super->segment0_blkaddr); 926 u32 cp_blkaddr = le32_to_cpu(raw_super->cp_blkaddr); 927 u32 sit_blkaddr = le32_to_cpu(raw_super->sit_blkaddr); 928 u32 nat_blkaddr = le32_to_cpu(raw_super->nat_blkaddr); 929 u32 ssa_blkaddr = le32_to_cpu(raw_super->ssa_blkaddr); 930 u32 main_blkaddr = le32_to_cpu(raw_super->main_blkaddr); 931 u32 segment_count_ckpt = le32_to_cpu(raw_super->segment_count_ckpt); 932 u32 segment_count_sit = le32_to_cpu(raw_super->segment_count_sit); 933 u32 segment_count_nat = le32_to_cpu(raw_super->segment_count_nat); 934 u32 segment_count_ssa = le32_to_cpu(raw_super->segment_count_ssa); 935 u32 segment_count_main = le32_to_cpu(raw_super->segment_count_main); 936 u32 segment_count = le32_to_cpu(raw_super->segment_count); 937 u32 log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg); 938 939 if (segment0_blkaddr != cp_blkaddr) { 940 f2fs_msg(sb, KERN_INFO, 941 "Mismatch start address, segment0(%u) cp_blkaddr(%u)", 942 segment0_blkaddr, cp_blkaddr); 943 return true; 944 } 945 946 if (cp_blkaddr + (segment_count_ckpt << log_blocks_per_seg) != 947 sit_blkaddr) { 948 f2fs_msg(sb, KERN_INFO, 949 "Wrong CP boundary, start(%u) end(%u) blocks(%u)", 950 cp_blkaddr, sit_blkaddr, 951 segment_count_ckpt << log_blocks_per_seg); 952 return true; 953 } 954 955 if (sit_blkaddr + (segment_count_sit << log_blocks_per_seg) != 956 nat_blkaddr) { 957 f2fs_msg(sb, KERN_INFO, 958 "Wrong SIT boundary, start(%u) end(%u) blocks(%u)", 959 sit_blkaddr, nat_blkaddr, 960 segment_count_sit << log_blocks_per_seg); 961 return true; 962 } 963 964 if (nat_blkaddr + (segment_count_nat << log_blocks_per_seg) != 965 ssa_blkaddr) { 966 f2fs_msg(sb, KERN_INFO, 967 "Wrong NAT boundary, start(%u) end(%u) blocks(%u)", 968 nat_blkaddr, ssa_blkaddr, 969 segment_count_nat << log_blocks_per_seg); 970 return true; 971 } 972 973 if (ssa_blkaddr + (segment_count_ssa << log_blocks_per_seg) != 974 main_blkaddr) { 975 f2fs_msg(sb, KERN_INFO, 976 "Wrong SSA boundary, start(%u) end(%u) blocks(%u)", 977 ssa_blkaddr, main_blkaddr, 978 segment_count_ssa << log_blocks_per_seg); 979 return true; 980 } 981 982 if (main_blkaddr + (segment_count_main << log_blocks_per_seg) != 983 segment0_blkaddr + (segment_count << log_blocks_per_seg)) { 984 f2fs_msg(sb, KERN_INFO, 985 "Wrong MAIN_AREA boundary, start(%u) end(%u) blocks(%u)", 986 main_blkaddr, 987 segment0_blkaddr + (segment_count << log_blocks_per_seg), 988 segment_count_main << log_blocks_per_seg); 989 return true; 990 } 991 992 return false; 993 } 994 995 static int sanity_check_raw_super(struct super_block *sb, 996 struct f2fs_super_block *raw_super) 997 { 998 unsigned int blocksize; 999 1000 if (F2FS_SUPER_MAGIC != le32_to_cpu(raw_super->magic)) { 1001 f2fs_msg(sb, KERN_INFO, 1002 "Magic Mismatch, valid(0x%x) - read(0x%x)", 1003 F2FS_SUPER_MAGIC, le32_to_cpu(raw_super->magic)); 1004 return 1; 1005 } 1006 1007 /* Currently, support only 4KB page cache size */ 1008 if (F2FS_BLKSIZE != PAGE_CACHE_SIZE) { 1009 f2fs_msg(sb, KERN_INFO, 1010 "Invalid page_cache_size (%lu), supports only 4KB\n", 1011 PAGE_CACHE_SIZE); 1012 return 1; 1013 } 1014 1015 /* Currently, support only 4KB block size */ 1016 blocksize = 1 << le32_to_cpu(raw_super->log_blocksize); 1017 if (blocksize != F2FS_BLKSIZE) { 1018 f2fs_msg(sb, KERN_INFO, 1019 "Invalid blocksize (%u), supports only 4KB\n", 1020 blocksize); 1021 return 1; 1022 } 1023 1024 /* check log blocks per segment */ 1025 if (le32_to_cpu(raw_super->log_blocks_per_seg) != 9) { 1026 f2fs_msg(sb, KERN_INFO, 1027 "Invalid log blocks per segment (%u)\n", 1028 le32_to_cpu(raw_super->log_blocks_per_seg)); 1029 return 1; 1030 } 1031 1032 /* Currently, support 512/1024/2048/4096 bytes sector size */ 1033 if (le32_to_cpu(raw_super->log_sectorsize) > 1034 F2FS_MAX_LOG_SECTOR_SIZE || 1035 le32_to_cpu(raw_super->log_sectorsize) < 1036 F2FS_MIN_LOG_SECTOR_SIZE) { 1037 f2fs_msg(sb, KERN_INFO, "Invalid log sectorsize (%u)", 1038 le32_to_cpu(raw_super->log_sectorsize)); 1039 return 1; 1040 } 1041 if (le32_to_cpu(raw_super->log_sectors_per_block) + 1042 le32_to_cpu(raw_super->log_sectorsize) != 1043 F2FS_MAX_LOG_SECTOR_SIZE) { 1044 f2fs_msg(sb, KERN_INFO, 1045 "Invalid log sectors per block(%u) log sectorsize(%u)", 1046 le32_to_cpu(raw_super->log_sectors_per_block), 1047 le32_to_cpu(raw_super->log_sectorsize)); 1048 return 1; 1049 } 1050 1051 /* check reserved ino info */ 1052 if (le32_to_cpu(raw_super->node_ino) != 1 || 1053 le32_to_cpu(raw_super->meta_ino) != 2 || 1054 le32_to_cpu(raw_super->root_ino) != 3) { 1055 f2fs_msg(sb, KERN_INFO, 1056 "Invalid Fs Meta Ino: node(%u) meta(%u) root(%u)", 1057 le32_to_cpu(raw_super->node_ino), 1058 le32_to_cpu(raw_super->meta_ino), 1059 le32_to_cpu(raw_super->root_ino)); 1060 return 1; 1061 } 1062 1063 /* check CP/SIT/NAT/SSA/MAIN_AREA area boundary */ 1064 if (sanity_check_area_boundary(sb, raw_super)) 1065 return 1; 1066 1067 return 0; 1068 } 1069 1070 static int sanity_check_ckpt(struct f2fs_sb_info *sbi) 1071 { 1072 unsigned int total, fsmeta; 1073 struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi); 1074 struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi); 1075 1076 total = le32_to_cpu(raw_super->segment_count); 1077 fsmeta = le32_to_cpu(raw_super->segment_count_ckpt); 1078 fsmeta += le32_to_cpu(raw_super->segment_count_sit); 1079 fsmeta += le32_to_cpu(raw_super->segment_count_nat); 1080 fsmeta += le32_to_cpu(ckpt->rsvd_segment_count); 1081 fsmeta += le32_to_cpu(raw_super->segment_count_ssa); 1082 1083 if (unlikely(fsmeta >= total)) 1084 return 1; 1085 1086 if (unlikely(f2fs_cp_error(sbi))) { 1087 f2fs_msg(sbi->sb, KERN_ERR, "A bug case: need to run fsck"); 1088 return 1; 1089 } 1090 return 0; 1091 } 1092 1093 static void init_sb_info(struct f2fs_sb_info *sbi) 1094 { 1095 struct f2fs_super_block *raw_super = sbi->raw_super; 1096 int i; 1097 1098 sbi->log_sectors_per_block = 1099 le32_to_cpu(raw_super->log_sectors_per_block); 1100 sbi->log_blocksize = le32_to_cpu(raw_super->log_blocksize); 1101 sbi->blocksize = 1 << sbi->log_blocksize; 1102 sbi->log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg); 1103 sbi->blocks_per_seg = 1 << sbi->log_blocks_per_seg; 1104 sbi->segs_per_sec = le32_to_cpu(raw_super->segs_per_sec); 1105 sbi->secs_per_zone = le32_to_cpu(raw_super->secs_per_zone); 1106 sbi->total_sections = le32_to_cpu(raw_super->section_count); 1107 sbi->total_node_count = 1108 (le32_to_cpu(raw_super->segment_count_nat) / 2) 1109 * sbi->blocks_per_seg * NAT_ENTRY_PER_BLOCK; 1110 sbi->root_ino_num = le32_to_cpu(raw_super->root_ino); 1111 sbi->node_ino_num = le32_to_cpu(raw_super->node_ino); 1112 sbi->meta_ino_num = le32_to_cpu(raw_super->meta_ino); 1113 sbi->cur_victim_sec = NULL_SECNO; 1114 sbi->max_victim_search = DEF_MAX_VICTIM_SEARCH; 1115 1116 for (i = 0; i < NR_COUNT_TYPE; i++) 1117 atomic_set(&sbi->nr_pages[i], 0); 1118 1119 sbi->dir_level = DEF_DIR_LEVEL; 1120 sbi->cp_interval = DEF_CP_INTERVAL; 1121 clear_sbi_flag(sbi, SBI_NEED_FSCK); 1122 1123 INIT_LIST_HEAD(&sbi->s_list); 1124 mutex_init(&sbi->umount_mutex); 1125 } 1126 1127 /* 1128 * Read f2fs raw super block. 1129 * Because we have two copies of super block, so read the first one at first, 1130 * if the first one is invalid, move to read the second one. 1131 */ 1132 static int read_raw_super_block(struct super_block *sb, 1133 struct f2fs_super_block **raw_super, 1134 int *valid_super_block, int *recovery) 1135 { 1136 int block = 0; 1137 struct buffer_head *bh; 1138 struct f2fs_super_block *super, *buf; 1139 int err = 0; 1140 1141 super = kzalloc(sizeof(struct f2fs_super_block), GFP_KERNEL); 1142 if (!super) 1143 return -ENOMEM; 1144 retry: 1145 bh = sb_bread(sb, block); 1146 if (!bh) { 1147 *recovery = 1; 1148 f2fs_msg(sb, KERN_ERR, "Unable to read %dth superblock", 1149 block + 1); 1150 err = -EIO; 1151 goto next; 1152 } 1153 1154 buf = (struct f2fs_super_block *)(bh->b_data + F2FS_SUPER_OFFSET); 1155 1156 /* sanity checking of raw super */ 1157 if (sanity_check_raw_super(sb, buf)) { 1158 brelse(bh); 1159 *recovery = 1; 1160 f2fs_msg(sb, KERN_ERR, 1161 "Can't find valid F2FS filesystem in %dth superblock", 1162 block + 1); 1163 err = -EINVAL; 1164 goto next; 1165 } 1166 1167 if (!*raw_super) { 1168 memcpy(super, buf, sizeof(*super)); 1169 *valid_super_block = block; 1170 *raw_super = super; 1171 } 1172 brelse(bh); 1173 1174 next: 1175 /* check the validity of the second superblock */ 1176 if (block == 0) { 1177 block++; 1178 goto retry; 1179 } 1180 1181 /* No valid superblock */ 1182 if (!*raw_super) { 1183 kfree(super); 1184 return err; 1185 } 1186 1187 return 0; 1188 } 1189 1190 int __f2fs_commit_super(struct f2fs_sb_info *sbi, int block) 1191 { 1192 struct f2fs_super_block *super = F2FS_RAW_SUPER(sbi); 1193 struct buffer_head *bh; 1194 int err; 1195 1196 bh = sb_getblk(sbi->sb, block); 1197 if (!bh) 1198 return -EIO; 1199 1200 lock_buffer(bh); 1201 memcpy(bh->b_data + F2FS_SUPER_OFFSET, super, sizeof(*super)); 1202 set_buffer_uptodate(bh); 1203 set_buffer_dirty(bh); 1204 unlock_buffer(bh); 1205 1206 /* it's rare case, we can do fua all the time */ 1207 err = __sync_dirty_buffer(bh, WRITE_FLUSH_FUA); 1208 brelse(bh); 1209 1210 return err; 1211 } 1212 1213 int f2fs_commit_super(struct f2fs_sb_info *sbi, bool recover) 1214 { 1215 int err; 1216 1217 /* write back-up superblock first */ 1218 err = __f2fs_commit_super(sbi, sbi->valid_super_block ? 0 : 1); 1219 1220 /* if we are in recovery path, skip writing valid superblock */ 1221 if (recover || err) 1222 return err; 1223 1224 /* write current valid superblock */ 1225 return __f2fs_commit_super(sbi, sbi->valid_super_block); 1226 } 1227 1228 static int f2fs_fill_super(struct super_block *sb, void *data, int silent) 1229 { 1230 struct f2fs_sb_info *sbi; 1231 struct f2fs_super_block *raw_super; 1232 struct inode *root; 1233 long err; 1234 bool retry = true, need_fsck = false; 1235 char *options = NULL; 1236 int recovery, i, valid_super_block; 1237 1238 try_onemore: 1239 err = -EINVAL; 1240 raw_super = NULL; 1241 valid_super_block = -1; 1242 recovery = 0; 1243 1244 /* allocate memory for f2fs-specific super block info */ 1245 sbi = kzalloc(sizeof(struct f2fs_sb_info), GFP_KERNEL); 1246 if (!sbi) 1247 return -ENOMEM; 1248 1249 /* set a block size */ 1250 if (unlikely(!sb_set_blocksize(sb, F2FS_BLKSIZE))) { 1251 f2fs_msg(sb, KERN_ERR, "unable to set blocksize"); 1252 goto free_sbi; 1253 } 1254 1255 err = read_raw_super_block(sb, &raw_super, &valid_super_block, 1256 &recovery); 1257 if (err) 1258 goto free_sbi; 1259 1260 sb->s_fs_info = sbi; 1261 default_options(sbi); 1262 /* parse mount options */ 1263 options = kstrdup((const char *)data, GFP_KERNEL); 1264 if (data && !options) { 1265 err = -ENOMEM; 1266 goto free_sb_buf; 1267 } 1268 1269 err = parse_options(sb, options); 1270 if (err) 1271 goto free_options; 1272 1273 sb->s_maxbytes = max_file_size(le32_to_cpu(raw_super->log_blocksize)); 1274 sb->s_max_links = F2FS_LINK_MAX; 1275 get_random_bytes(&sbi->s_next_generation, sizeof(u32)); 1276 1277 sb->s_op = &f2fs_sops; 1278 sb->s_xattr = f2fs_xattr_handlers; 1279 sb->s_export_op = &f2fs_export_ops; 1280 sb->s_magic = F2FS_SUPER_MAGIC; 1281 sb->s_time_gran = 1; 1282 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) | 1283 (test_opt(sbi, POSIX_ACL) ? MS_POSIXACL : 0); 1284 memcpy(sb->s_uuid, raw_super->uuid, sizeof(raw_super->uuid)); 1285 1286 /* init f2fs-specific super block info */ 1287 sbi->sb = sb; 1288 sbi->raw_super = raw_super; 1289 sbi->valid_super_block = valid_super_block; 1290 mutex_init(&sbi->gc_mutex); 1291 mutex_init(&sbi->writepages); 1292 mutex_init(&sbi->cp_mutex); 1293 init_rwsem(&sbi->node_write); 1294 1295 /* disallow all the data/node/meta page writes */ 1296 set_sbi_flag(sbi, SBI_POR_DOING); 1297 spin_lock_init(&sbi->stat_lock); 1298 1299 init_rwsem(&sbi->read_io.io_rwsem); 1300 sbi->read_io.sbi = sbi; 1301 sbi->read_io.bio = NULL; 1302 for (i = 0; i < NR_PAGE_TYPE; i++) { 1303 init_rwsem(&sbi->write_io[i].io_rwsem); 1304 sbi->write_io[i].sbi = sbi; 1305 sbi->write_io[i].bio = NULL; 1306 } 1307 1308 init_rwsem(&sbi->cp_rwsem); 1309 init_waitqueue_head(&sbi->cp_wait); 1310 init_sb_info(sbi); 1311 1312 /* get an inode for meta space */ 1313 sbi->meta_inode = f2fs_iget(sb, F2FS_META_INO(sbi)); 1314 if (IS_ERR(sbi->meta_inode)) { 1315 f2fs_msg(sb, KERN_ERR, "Failed to read F2FS meta data inode"); 1316 err = PTR_ERR(sbi->meta_inode); 1317 goto free_options; 1318 } 1319 1320 err = get_valid_checkpoint(sbi); 1321 if (err) { 1322 f2fs_msg(sb, KERN_ERR, "Failed to get valid F2FS checkpoint"); 1323 goto free_meta_inode; 1324 } 1325 1326 /* sanity checking of checkpoint */ 1327 err = -EINVAL; 1328 if (sanity_check_ckpt(sbi)) { 1329 f2fs_msg(sb, KERN_ERR, "Invalid F2FS checkpoint"); 1330 goto free_cp; 1331 } 1332 1333 sbi->total_valid_node_count = 1334 le32_to_cpu(sbi->ckpt->valid_node_count); 1335 sbi->total_valid_inode_count = 1336 le32_to_cpu(sbi->ckpt->valid_inode_count); 1337 sbi->user_block_count = le64_to_cpu(sbi->ckpt->user_block_count); 1338 sbi->total_valid_block_count = 1339 le64_to_cpu(sbi->ckpt->valid_block_count); 1340 sbi->last_valid_block_count = sbi->total_valid_block_count; 1341 sbi->alloc_valid_block_count = 0; 1342 INIT_LIST_HEAD(&sbi->dir_inode_list); 1343 spin_lock_init(&sbi->dir_inode_lock); 1344 1345 init_extent_cache_info(sbi); 1346 1347 init_ino_entry_info(sbi); 1348 1349 /* setup f2fs internal modules */ 1350 err = build_segment_manager(sbi); 1351 if (err) { 1352 f2fs_msg(sb, KERN_ERR, 1353 "Failed to initialize F2FS segment manager"); 1354 goto free_sm; 1355 } 1356 err = build_node_manager(sbi); 1357 if (err) { 1358 f2fs_msg(sb, KERN_ERR, 1359 "Failed to initialize F2FS node manager"); 1360 goto free_nm; 1361 } 1362 1363 build_gc_manager(sbi); 1364 1365 /* get an inode for node space */ 1366 sbi->node_inode = f2fs_iget(sb, F2FS_NODE_INO(sbi)); 1367 if (IS_ERR(sbi->node_inode)) { 1368 f2fs_msg(sb, KERN_ERR, "Failed to read node inode"); 1369 err = PTR_ERR(sbi->node_inode); 1370 goto free_nm; 1371 } 1372 1373 f2fs_join_shrinker(sbi); 1374 1375 /* if there are nt orphan nodes free them */ 1376 err = recover_orphan_inodes(sbi); 1377 if (err) 1378 goto free_node_inode; 1379 1380 /* read root inode and dentry */ 1381 root = f2fs_iget(sb, F2FS_ROOT_INO(sbi)); 1382 if (IS_ERR(root)) { 1383 f2fs_msg(sb, KERN_ERR, "Failed to read root inode"); 1384 err = PTR_ERR(root); 1385 goto free_node_inode; 1386 } 1387 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) { 1388 iput(root); 1389 err = -EINVAL; 1390 goto free_node_inode; 1391 } 1392 1393 sb->s_root = d_make_root(root); /* allocate root dentry */ 1394 if (!sb->s_root) { 1395 err = -ENOMEM; 1396 goto free_root_inode; 1397 } 1398 1399 err = f2fs_build_stats(sbi); 1400 if (err) 1401 goto free_root_inode; 1402 1403 if (f2fs_proc_root) 1404 sbi->s_proc = proc_mkdir(sb->s_id, f2fs_proc_root); 1405 1406 if (sbi->s_proc) 1407 proc_create_data("segment_info", S_IRUGO, sbi->s_proc, 1408 &f2fs_seq_segment_info_fops, sb); 1409 1410 sbi->s_kobj.kset = f2fs_kset; 1411 init_completion(&sbi->s_kobj_unregister); 1412 err = kobject_init_and_add(&sbi->s_kobj, &f2fs_ktype, NULL, 1413 "%s", sb->s_id); 1414 if (err) 1415 goto free_proc; 1416 1417 /* recover fsynced data */ 1418 if (!test_opt(sbi, DISABLE_ROLL_FORWARD)) { 1419 /* 1420 * mount should be failed, when device has readonly mode, and 1421 * previous checkpoint was not done by clean system shutdown. 1422 */ 1423 if (bdev_read_only(sb->s_bdev) && 1424 !is_set_ckpt_flags(sbi->ckpt, CP_UMOUNT_FLAG)) { 1425 err = -EROFS; 1426 goto free_kobj; 1427 } 1428 1429 if (need_fsck) 1430 set_sbi_flag(sbi, SBI_NEED_FSCK); 1431 1432 err = recover_fsync_data(sbi); 1433 if (err) { 1434 need_fsck = true; 1435 f2fs_msg(sb, KERN_ERR, 1436 "Cannot recover all fsync data errno=%ld", err); 1437 goto free_kobj; 1438 } 1439 } 1440 /* recover_fsync_data() cleared this already */ 1441 clear_sbi_flag(sbi, SBI_POR_DOING); 1442 1443 /* 1444 * If filesystem is not mounted as read-only then 1445 * do start the gc_thread. 1446 */ 1447 if (test_opt(sbi, BG_GC) && !f2fs_readonly(sb)) { 1448 /* After POR, we can run background GC thread.*/ 1449 err = start_gc_thread(sbi); 1450 if (err) 1451 goto free_kobj; 1452 } 1453 kfree(options); 1454 1455 /* recover broken superblock */ 1456 if (recovery && !f2fs_readonly(sb) && !bdev_read_only(sb->s_bdev)) { 1457 f2fs_msg(sb, KERN_INFO, "Recover invalid superblock"); 1458 f2fs_commit_super(sbi, true); 1459 } 1460 1461 sbi->cp_expires = round_jiffies_up(jiffies); 1462 1463 return 0; 1464 1465 free_kobj: 1466 kobject_del(&sbi->s_kobj); 1467 kobject_put(&sbi->s_kobj); 1468 wait_for_completion(&sbi->s_kobj_unregister); 1469 free_proc: 1470 if (sbi->s_proc) { 1471 remove_proc_entry("segment_info", sbi->s_proc); 1472 remove_proc_entry(sb->s_id, f2fs_proc_root); 1473 } 1474 f2fs_destroy_stats(sbi); 1475 free_root_inode: 1476 dput(sb->s_root); 1477 sb->s_root = NULL; 1478 free_node_inode: 1479 mutex_lock(&sbi->umount_mutex); 1480 f2fs_leave_shrinker(sbi); 1481 iput(sbi->node_inode); 1482 mutex_unlock(&sbi->umount_mutex); 1483 free_nm: 1484 destroy_node_manager(sbi); 1485 free_sm: 1486 destroy_segment_manager(sbi); 1487 free_cp: 1488 kfree(sbi->ckpt); 1489 free_meta_inode: 1490 make_bad_inode(sbi->meta_inode); 1491 iput(sbi->meta_inode); 1492 free_options: 1493 kfree(options); 1494 free_sb_buf: 1495 kfree(raw_super); 1496 free_sbi: 1497 kfree(sbi); 1498 1499 /* give only one another chance */ 1500 if (retry) { 1501 retry = false; 1502 shrink_dcache_sb(sb); 1503 goto try_onemore; 1504 } 1505 return err; 1506 } 1507 1508 static struct dentry *f2fs_mount(struct file_system_type *fs_type, int flags, 1509 const char *dev_name, void *data) 1510 { 1511 return mount_bdev(fs_type, flags, dev_name, data, f2fs_fill_super); 1512 } 1513 1514 static void kill_f2fs_super(struct super_block *sb) 1515 { 1516 if (sb->s_root) 1517 set_sbi_flag(F2FS_SB(sb), SBI_IS_CLOSE); 1518 kill_block_super(sb); 1519 } 1520 1521 static struct file_system_type f2fs_fs_type = { 1522 .owner = THIS_MODULE, 1523 .name = "f2fs", 1524 .mount = f2fs_mount, 1525 .kill_sb = kill_f2fs_super, 1526 .fs_flags = FS_REQUIRES_DEV, 1527 }; 1528 MODULE_ALIAS_FS("f2fs"); 1529 1530 static int __init init_inodecache(void) 1531 { 1532 f2fs_inode_cachep = f2fs_kmem_cache_create("f2fs_inode_cache", 1533 sizeof(struct f2fs_inode_info)); 1534 if (!f2fs_inode_cachep) 1535 return -ENOMEM; 1536 return 0; 1537 } 1538 1539 static void destroy_inodecache(void) 1540 { 1541 /* 1542 * Make sure all delayed rcu free inodes are flushed before we 1543 * destroy cache. 1544 */ 1545 rcu_barrier(); 1546 kmem_cache_destroy(f2fs_inode_cachep); 1547 } 1548 1549 static int __init init_f2fs_fs(void) 1550 { 1551 int err; 1552 1553 f2fs_build_trace_ios(); 1554 1555 err = init_inodecache(); 1556 if (err) 1557 goto fail; 1558 err = create_node_manager_caches(); 1559 if (err) 1560 goto free_inodecache; 1561 err = create_segment_manager_caches(); 1562 if (err) 1563 goto free_node_manager_caches; 1564 err = create_checkpoint_caches(); 1565 if (err) 1566 goto free_segment_manager_caches; 1567 err = create_extent_cache(); 1568 if (err) 1569 goto free_checkpoint_caches; 1570 f2fs_kset = kset_create_and_add("f2fs", NULL, fs_kobj); 1571 if (!f2fs_kset) { 1572 err = -ENOMEM; 1573 goto free_extent_cache; 1574 } 1575 err = f2fs_init_crypto(); 1576 if (err) 1577 goto free_kset; 1578 1579 err = register_shrinker(&f2fs_shrinker_info); 1580 if (err) 1581 goto free_crypto; 1582 1583 err = register_filesystem(&f2fs_fs_type); 1584 if (err) 1585 goto free_shrinker; 1586 err = f2fs_create_root_stats(); 1587 if (err) 1588 goto free_filesystem; 1589 f2fs_proc_root = proc_mkdir("fs/f2fs", NULL); 1590 return 0; 1591 1592 free_filesystem: 1593 unregister_filesystem(&f2fs_fs_type); 1594 free_shrinker: 1595 unregister_shrinker(&f2fs_shrinker_info); 1596 free_crypto: 1597 f2fs_exit_crypto(); 1598 free_kset: 1599 kset_unregister(f2fs_kset); 1600 free_extent_cache: 1601 destroy_extent_cache(); 1602 free_checkpoint_caches: 1603 destroy_checkpoint_caches(); 1604 free_segment_manager_caches: 1605 destroy_segment_manager_caches(); 1606 free_node_manager_caches: 1607 destroy_node_manager_caches(); 1608 free_inodecache: 1609 destroy_inodecache(); 1610 fail: 1611 return err; 1612 } 1613 1614 static void __exit exit_f2fs_fs(void) 1615 { 1616 remove_proc_entry("fs/f2fs", NULL); 1617 f2fs_destroy_root_stats(); 1618 unregister_shrinker(&f2fs_shrinker_info); 1619 unregister_filesystem(&f2fs_fs_type); 1620 f2fs_exit_crypto(); 1621 destroy_extent_cache(); 1622 destroy_checkpoint_caches(); 1623 destroy_segment_manager_caches(); 1624 destroy_node_manager_caches(); 1625 destroy_inodecache(); 1626 kset_unregister(f2fs_kset); 1627 f2fs_destroy_trace_ios(); 1628 } 1629 1630 module_init(init_f2fs_fs) 1631 module_exit(exit_f2fs_fs) 1632 1633 MODULE_AUTHOR("Samsung Electronics's Praesto Team"); 1634 MODULE_DESCRIPTION("Flash Friendly File System"); 1635 MODULE_LICENSE("GPL"); 1636