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 #ifdef CONFIG_F2FS_FAULT_INJECTION 43 44 char *fault_name[FAULT_MAX] = { 45 [FAULT_KMALLOC] = "kmalloc", 46 [FAULT_PAGE_ALLOC] = "page alloc", 47 [FAULT_ALLOC_NID] = "alloc nid", 48 [FAULT_ORPHAN] = "orphan", 49 [FAULT_BLOCK] = "no more block", 50 [FAULT_DIR_DEPTH] = "too big dir depth", 51 [FAULT_EVICT_INODE] = "evict_inode fail", 52 [FAULT_TRUNCATE] = "truncate fail", 53 [FAULT_IO] = "IO error", 54 [FAULT_CHECKPOINT] = "checkpoint error", 55 }; 56 57 static void f2fs_build_fault_attr(struct f2fs_sb_info *sbi, 58 unsigned int rate) 59 { 60 struct f2fs_fault_info *ffi = &sbi->fault_info; 61 62 if (rate) { 63 atomic_set(&ffi->inject_ops, 0); 64 ffi->inject_rate = rate; 65 ffi->inject_type = (1 << FAULT_MAX) - 1; 66 } else { 67 memset(ffi, 0, sizeof(struct f2fs_fault_info)); 68 } 69 } 70 #endif 71 72 /* f2fs-wide shrinker description */ 73 static struct shrinker f2fs_shrinker_info = { 74 .scan_objects = f2fs_shrink_scan, 75 .count_objects = f2fs_shrink_count, 76 .seeks = DEFAULT_SEEKS, 77 }; 78 79 enum { 80 Opt_gc_background, 81 Opt_disable_roll_forward, 82 Opt_norecovery, 83 Opt_discard, 84 Opt_nodiscard, 85 Opt_noheap, 86 Opt_heap, 87 Opt_user_xattr, 88 Opt_nouser_xattr, 89 Opt_acl, 90 Opt_noacl, 91 Opt_active_logs, 92 Opt_disable_ext_identify, 93 Opt_inline_xattr, 94 Opt_noinline_xattr, 95 Opt_inline_data, 96 Opt_inline_dentry, 97 Opt_noinline_dentry, 98 Opt_flush_merge, 99 Opt_noflush_merge, 100 Opt_nobarrier, 101 Opt_fastboot, 102 Opt_extent_cache, 103 Opt_noextent_cache, 104 Opt_noinline_data, 105 Opt_data_flush, 106 Opt_mode, 107 Opt_io_size_bits, 108 Opt_fault_injection, 109 Opt_lazytime, 110 Opt_nolazytime, 111 Opt_err, 112 }; 113 114 static match_table_t f2fs_tokens = { 115 {Opt_gc_background, "background_gc=%s"}, 116 {Opt_disable_roll_forward, "disable_roll_forward"}, 117 {Opt_norecovery, "norecovery"}, 118 {Opt_discard, "discard"}, 119 {Opt_nodiscard, "nodiscard"}, 120 {Opt_noheap, "no_heap"}, 121 {Opt_heap, "heap"}, 122 {Opt_user_xattr, "user_xattr"}, 123 {Opt_nouser_xattr, "nouser_xattr"}, 124 {Opt_acl, "acl"}, 125 {Opt_noacl, "noacl"}, 126 {Opt_active_logs, "active_logs=%u"}, 127 {Opt_disable_ext_identify, "disable_ext_identify"}, 128 {Opt_inline_xattr, "inline_xattr"}, 129 {Opt_noinline_xattr, "noinline_xattr"}, 130 {Opt_inline_data, "inline_data"}, 131 {Opt_inline_dentry, "inline_dentry"}, 132 {Opt_noinline_dentry, "noinline_dentry"}, 133 {Opt_flush_merge, "flush_merge"}, 134 {Opt_noflush_merge, "noflush_merge"}, 135 {Opt_nobarrier, "nobarrier"}, 136 {Opt_fastboot, "fastboot"}, 137 {Opt_extent_cache, "extent_cache"}, 138 {Opt_noextent_cache, "noextent_cache"}, 139 {Opt_noinline_data, "noinline_data"}, 140 {Opt_data_flush, "data_flush"}, 141 {Opt_mode, "mode=%s"}, 142 {Opt_io_size_bits, "io_bits=%u"}, 143 {Opt_fault_injection, "fault_injection=%u"}, 144 {Opt_lazytime, "lazytime"}, 145 {Opt_nolazytime, "nolazytime"}, 146 {Opt_err, NULL}, 147 }; 148 149 /* Sysfs support for f2fs */ 150 enum { 151 GC_THREAD, /* struct f2fs_gc_thread */ 152 SM_INFO, /* struct f2fs_sm_info */ 153 DCC_INFO, /* struct discard_cmd_control */ 154 NM_INFO, /* struct f2fs_nm_info */ 155 F2FS_SBI, /* struct f2fs_sb_info */ 156 #ifdef CONFIG_F2FS_FAULT_INJECTION 157 FAULT_INFO_RATE, /* struct f2fs_fault_info */ 158 FAULT_INFO_TYPE, /* struct f2fs_fault_info */ 159 #endif 160 }; 161 162 struct f2fs_attr { 163 struct attribute attr; 164 ssize_t (*show)(struct f2fs_attr *, struct f2fs_sb_info *, char *); 165 ssize_t (*store)(struct f2fs_attr *, struct f2fs_sb_info *, 166 const char *, size_t); 167 int struct_type; 168 int offset; 169 }; 170 171 static unsigned char *__struct_ptr(struct f2fs_sb_info *sbi, int struct_type) 172 { 173 if (struct_type == GC_THREAD) 174 return (unsigned char *)sbi->gc_thread; 175 else if (struct_type == SM_INFO) 176 return (unsigned char *)SM_I(sbi); 177 else if (struct_type == DCC_INFO) 178 return (unsigned char *)SM_I(sbi)->dcc_info; 179 else if (struct_type == NM_INFO) 180 return (unsigned char *)NM_I(sbi); 181 else if (struct_type == F2FS_SBI) 182 return (unsigned char *)sbi; 183 #ifdef CONFIG_F2FS_FAULT_INJECTION 184 else if (struct_type == FAULT_INFO_RATE || 185 struct_type == FAULT_INFO_TYPE) 186 return (unsigned char *)&sbi->fault_info; 187 #endif 188 return NULL; 189 } 190 191 static ssize_t lifetime_write_kbytes_show(struct f2fs_attr *a, 192 struct f2fs_sb_info *sbi, char *buf) 193 { 194 struct super_block *sb = sbi->sb; 195 196 if (!sb->s_bdev->bd_part) 197 return snprintf(buf, PAGE_SIZE, "0\n"); 198 199 return snprintf(buf, PAGE_SIZE, "%llu\n", 200 (unsigned long long)(sbi->kbytes_written + 201 BD_PART_WRITTEN(sbi))); 202 } 203 204 static ssize_t f2fs_sbi_show(struct f2fs_attr *a, 205 struct f2fs_sb_info *sbi, char *buf) 206 { 207 unsigned char *ptr = NULL; 208 unsigned int *ui; 209 210 ptr = __struct_ptr(sbi, a->struct_type); 211 if (!ptr) 212 return -EINVAL; 213 214 ui = (unsigned int *)(ptr + a->offset); 215 216 return snprintf(buf, PAGE_SIZE, "%u\n", *ui); 217 } 218 219 static ssize_t f2fs_sbi_store(struct f2fs_attr *a, 220 struct f2fs_sb_info *sbi, 221 const char *buf, size_t count) 222 { 223 unsigned char *ptr; 224 unsigned long t; 225 unsigned int *ui; 226 ssize_t ret; 227 228 ptr = __struct_ptr(sbi, a->struct_type); 229 if (!ptr) 230 return -EINVAL; 231 232 ui = (unsigned int *)(ptr + a->offset); 233 234 ret = kstrtoul(skip_spaces(buf), 0, &t); 235 if (ret < 0) 236 return ret; 237 #ifdef CONFIG_F2FS_FAULT_INJECTION 238 if (a->struct_type == FAULT_INFO_TYPE && t >= (1 << FAULT_MAX)) 239 return -EINVAL; 240 #endif 241 *ui = t; 242 return count; 243 } 244 245 static ssize_t f2fs_attr_show(struct kobject *kobj, 246 struct attribute *attr, char *buf) 247 { 248 struct f2fs_sb_info *sbi = container_of(kobj, struct f2fs_sb_info, 249 s_kobj); 250 struct f2fs_attr *a = container_of(attr, struct f2fs_attr, attr); 251 252 return a->show ? a->show(a, sbi, buf) : 0; 253 } 254 255 static ssize_t f2fs_attr_store(struct kobject *kobj, struct attribute *attr, 256 const char *buf, size_t len) 257 { 258 struct f2fs_sb_info *sbi = container_of(kobj, struct f2fs_sb_info, 259 s_kobj); 260 struct f2fs_attr *a = container_of(attr, struct f2fs_attr, attr); 261 262 return a->store ? a->store(a, sbi, buf, len) : 0; 263 } 264 265 static void f2fs_sb_release(struct kobject *kobj) 266 { 267 struct f2fs_sb_info *sbi = container_of(kobj, struct f2fs_sb_info, 268 s_kobj); 269 complete(&sbi->s_kobj_unregister); 270 } 271 272 #define F2FS_ATTR_OFFSET(_struct_type, _name, _mode, _show, _store, _offset) \ 273 static struct f2fs_attr f2fs_attr_##_name = { \ 274 .attr = {.name = __stringify(_name), .mode = _mode }, \ 275 .show = _show, \ 276 .store = _store, \ 277 .struct_type = _struct_type, \ 278 .offset = _offset \ 279 } 280 281 #define F2FS_RW_ATTR(struct_type, struct_name, name, elname) \ 282 F2FS_ATTR_OFFSET(struct_type, name, 0644, \ 283 f2fs_sbi_show, f2fs_sbi_store, \ 284 offsetof(struct struct_name, elname)) 285 286 #define F2FS_GENERAL_RO_ATTR(name) \ 287 static struct f2fs_attr f2fs_attr_##name = __ATTR(name, 0444, name##_show, NULL) 288 289 F2FS_RW_ATTR(GC_THREAD, f2fs_gc_kthread, gc_min_sleep_time, min_sleep_time); 290 F2FS_RW_ATTR(GC_THREAD, f2fs_gc_kthread, gc_max_sleep_time, max_sleep_time); 291 F2FS_RW_ATTR(GC_THREAD, f2fs_gc_kthread, gc_no_gc_sleep_time, no_gc_sleep_time); 292 F2FS_RW_ATTR(GC_THREAD, f2fs_gc_kthread, gc_idle, gc_idle); 293 F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, reclaim_segments, rec_prefree_segments); 294 F2FS_RW_ATTR(DCC_INFO, discard_cmd_control, max_small_discards, max_discards); 295 F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, batched_trim_sections, trim_sections); 296 F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, ipu_policy, ipu_policy); 297 F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, min_ipu_util, min_ipu_util); 298 F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, min_fsync_blocks, min_fsync_blocks); 299 F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, min_hot_blocks, min_hot_blocks); 300 F2FS_RW_ATTR(NM_INFO, f2fs_nm_info, ram_thresh, ram_thresh); 301 F2FS_RW_ATTR(NM_INFO, f2fs_nm_info, ra_nid_pages, ra_nid_pages); 302 F2FS_RW_ATTR(NM_INFO, f2fs_nm_info, dirty_nats_ratio, dirty_nats_ratio); 303 F2FS_RW_ATTR(F2FS_SBI, f2fs_sb_info, max_victim_search, max_victim_search); 304 F2FS_RW_ATTR(F2FS_SBI, f2fs_sb_info, dir_level, dir_level); 305 F2FS_RW_ATTR(F2FS_SBI, f2fs_sb_info, cp_interval, interval_time[CP_TIME]); 306 F2FS_RW_ATTR(F2FS_SBI, f2fs_sb_info, idle_interval, interval_time[REQ_TIME]); 307 #ifdef CONFIG_F2FS_FAULT_INJECTION 308 F2FS_RW_ATTR(FAULT_INFO_RATE, f2fs_fault_info, inject_rate, inject_rate); 309 F2FS_RW_ATTR(FAULT_INFO_TYPE, f2fs_fault_info, inject_type, inject_type); 310 #endif 311 F2FS_GENERAL_RO_ATTR(lifetime_write_kbytes); 312 313 #define ATTR_LIST(name) (&f2fs_attr_##name.attr) 314 static struct attribute *f2fs_attrs[] = { 315 ATTR_LIST(gc_min_sleep_time), 316 ATTR_LIST(gc_max_sleep_time), 317 ATTR_LIST(gc_no_gc_sleep_time), 318 ATTR_LIST(gc_idle), 319 ATTR_LIST(reclaim_segments), 320 ATTR_LIST(max_small_discards), 321 ATTR_LIST(batched_trim_sections), 322 ATTR_LIST(ipu_policy), 323 ATTR_LIST(min_ipu_util), 324 ATTR_LIST(min_fsync_blocks), 325 ATTR_LIST(min_hot_blocks), 326 ATTR_LIST(max_victim_search), 327 ATTR_LIST(dir_level), 328 ATTR_LIST(ram_thresh), 329 ATTR_LIST(ra_nid_pages), 330 ATTR_LIST(dirty_nats_ratio), 331 ATTR_LIST(cp_interval), 332 ATTR_LIST(idle_interval), 333 #ifdef CONFIG_F2FS_FAULT_INJECTION 334 ATTR_LIST(inject_rate), 335 ATTR_LIST(inject_type), 336 #endif 337 ATTR_LIST(lifetime_write_kbytes), 338 NULL, 339 }; 340 341 static const struct sysfs_ops f2fs_attr_ops = { 342 .show = f2fs_attr_show, 343 .store = f2fs_attr_store, 344 }; 345 346 static struct kobj_type f2fs_ktype = { 347 .default_attrs = f2fs_attrs, 348 .sysfs_ops = &f2fs_attr_ops, 349 .release = f2fs_sb_release, 350 }; 351 352 void f2fs_msg(struct super_block *sb, const char *level, const char *fmt, ...) 353 { 354 struct va_format vaf; 355 va_list args; 356 357 va_start(args, fmt); 358 vaf.fmt = fmt; 359 vaf.va = &args; 360 printk("%sF2FS-fs (%s): %pV\n", level, sb->s_id, &vaf); 361 va_end(args); 362 } 363 364 static void init_once(void *foo) 365 { 366 struct f2fs_inode_info *fi = (struct f2fs_inode_info *) foo; 367 368 inode_init_once(&fi->vfs_inode); 369 } 370 371 static int parse_options(struct super_block *sb, char *options) 372 { 373 struct f2fs_sb_info *sbi = F2FS_SB(sb); 374 struct request_queue *q; 375 substring_t args[MAX_OPT_ARGS]; 376 char *p, *name; 377 int arg = 0; 378 379 if (!options) 380 return 0; 381 382 while ((p = strsep(&options, ",")) != NULL) { 383 int token; 384 if (!*p) 385 continue; 386 /* 387 * Initialize args struct so we know whether arg was 388 * found; some options take optional arguments. 389 */ 390 args[0].to = args[0].from = NULL; 391 token = match_token(p, f2fs_tokens, args); 392 393 switch (token) { 394 case Opt_gc_background: 395 name = match_strdup(&args[0]); 396 397 if (!name) 398 return -ENOMEM; 399 if (strlen(name) == 2 && !strncmp(name, "on", 2)) { 400 set_opt(sbi, BG_GC); 401 clear_opt(sbi, FORCE_FG_GC); 402 } else if (strlen(name) == 3 && !strncmp(name, "off", 3)) { 403 clear_opt(sbi, BG_GC); 404 clear_opt(sbi, FORCE_FG_GC); 405 } else if (strlen(name) == 4 && !strncmp(name, "sync", 4)) { 406 set_opt(sbi, BG_GC); 407 set_opt(sbi, FORCE_FG_GC); 408 } else { 409 kfree(name); 410 return -EINVAL; 411 } 412 kfree(name); 413 break; 414 case Opt_disable_roll_forward: 415 set_opt(sbi, DISABLE_ROLL_FORWARD); 416 break; 417 case Opt_norecovery: 418 /* this option mounts f2fs with ro */ 419 set_opt(sbi, DISABLE_ROLL_FORWARD); 420 if (!f2fs_readonly(sb)) 421 return -EINVAL; 422 break; 423 case Opt_discard: 424 q = bdev_get_queue(sb->s_bdev); 425 if (blk_queue_discard(q)) { 426 set_opt(sbi, DISCARD); 427 } else if (!f2fs_sb_mounted_blkzoned(sb)) { 428 f2fs_msg(sb, KERN_WARNING, 429 "mounting with \"discard\" option, but " 430 "the device does not support discard"); 431 } 432 break; 433 case Opt_nodiscard: 434 if (f2fs_sb_mounted_blkzoned(sb)) { 435 f2fs_msg(sb, KERN_WARNING, 436 "discard is required for zoned block devices"); 437 return -EINVAL; 438 } 439 clear_opt(sbi, DISCARD); 440 break; 441 case Opt_noheap: 442 set_opt(sbi, NOHEAP); 443 break; 444 case Opt_heap: 445 clear_opt(sbi, NOHEAP); 446 break; 447 #ifdef CONFIG_F2FS_FS_XATTR 448 case Opt_user_xattr: 449 set_opt(sbi, XATTR_USER); 450 break; 451 case Opt_nouser_xattr: 452 clear_opt(sbi, XATTR_USER); 453 break; 454 case Opt_inline_xattr: 455 set_opt(sbi, INLINE_XATTR); 456 break; 457 case Opt_noinline_xattr: 458 clear_opt(sbi, INLINE_XATTR); 459 break; 460 #else 461 case Opt_user_xattr: 462 f2fs_msg(sb, KERN_INFO, 463 "user_xattr options not supported"); 464 break; 465 case Opt_nouser_xattr: 466 f2fs_msg(sb, KERN_INFO, 467 "nouser_xattr options not supported"); 468 break; 469 case Opt_inline_xattr: 470 f2fs_msg(sb, KERN_INFO, 471 "inline_xattr options not supported"); 472 break; 473 case Opt_noinline_xattr: 474 f2fs_msg(sb, KERN_INFO, 475 "noinline_xattr options not supported"); 476 break; 477 #endif 478 #ifdef CONFIG_F2FS_FS_POSIX_ACL 479 case Opt_acl: 480 set_opt(sbi, POSIX_ACL); 481 break; 482 case Opt_noacl: 483 clear_opt(sbi, POSIX_ACL); 484 break; 485 #else 486 case Opt_acl: 487 f2fs_msg(sb, KERN_INFO, "acl options not supported"); 488 break; 489 case Opt_noacl: 490 f2fs_msg(sb, KERN_INFO, "noacl options not supported"); 491 break; 492 #endif 493 case Opt_active_logs: 494 if (args->from && match_int(args, &arg)) 495 return -EINVAL; 496 if (arg != 2 && arg != 4 && arg != NR_CURSEG_TYPE) 497 return -EINVAL; 498 sbi->active_logs = arg; 499 break; 500 case Opt_disable_ext_identify: 501 set_opt(sbi, DISABLE_EXT_IDENTIFY); 502 break; 503 case Opt_inline_data: 504 set_opt(sbi, INLINE_DATA); 505 break; 506 case Opt_inline_dentry: 507 set_opt(sbi, INLINE_DENTRY); 508 break; 509 case Opt_noinline_dentry: 510 clear_opt(sbi, INLINE_DENTRY); 511 break; 512 case Opt_flush_merge: 513 set_opt(sbi, FLUSH_MERGE); 514 break; 515 case Opt_noflush_merge: 516 clear_opt(sbi, FLUSH_MERGE); 517 break; 518 case Opt_nobarrier: 519 set_opt(sbi, NOBARRIER); 520 break; 521 case Opt_fastboot: 522 set_opt(sbi, FASTBOOT); 523 break; 524 case Opt_extent_cache: 525 set_opt(sbi, EXTENT_CACHE); 526 break; 527 case Opt_noextent_cache: 528 clear_opt(sbi, EXTENT_CACHE); 529 break; 530 case Opt_noinline_data: 531 clear_opt(sbi, INLINE_DATA); 532 break; 533 case Opt_data_flush: 534 set_opt(sbi, DATA_FLUSH); 535 break; 536 case Opt_mode: 537 name = match_strdup(&args[0]); 538 539 if (!name) 540 return -ENOMEM; 541 if (strlen(name) == 8 && 542 !strncmp(name, "adaptive", 8)) { 543 if (f2fs_sb_mounted_blkzoned(sb)) { 544 f2fs_msg(sb, KERN_WARNING, 545 "adaptive mode is not allowed with " 546 "zoned block device feature"); 547 kfree(name); 548 return -EINVAL; 549 } 550 set_opt_mode(sbi, F2FS_MOUNT_ADAPTIVE); 551 } else if (strlen(name) == 3 && 552 !strncmp(name, "lfs", 3)) { 553 set_opt_mode(sbi, F2FS_MOUNT_LFS); 554 } else { 555 kfree(name); 556 return -EINVAL; 557 } 558 kfree(name); 559 break; 560 case Opt_io_size_bits: 561 if (args->from && match_int(args, &arg)) 562 return -EINVAL; 563 if (arg > __ilog2_u32(BIO_MAX_PAGES)) { 564 f2fs_msg(sb, KERN_WARNING, 565 "Not support %d, larger than %d", 566 1 << arg, BIO_MAX_PAGES); 567 return -EINVAL; 568 } 569 sbi->write_io_size_bits = arg; 570 break; 571 case Opt_fault_injection: 572 if (args->from && match_int(args, &arg)) 573 return -EINVAL; 574 #ifdef CONFIG_F2FS_FAULT_INJECTION 575 f2fs_build_fault_attr(sbi, arg); 576 set_opt(sbi, FAULT_INJECTION); 577 #else 578 f2fs_msg(sb, KERN_INFO, 579 "FAULT_INJECTION was not selected"); 580 #endif 581 break; 582 case Opt_lazytime: 583 sb->s_flags |= MS_LAZYTIME; 584 break; 585 case Opt_nolazytime: 586 sb->s_flags &= ~MS_LAZYTIME; 587 break; 588 default: 589 f2fs_msg(sb, KERN_ERR, 590 "Unrecognized mount option \"%s\" or missing value", 591 p); 592 return -EINVAL; 593 } 594 } 595 596 if (F2FS_IO_SIZE_BITS(sbi) && !test_opt(sbi, LFS)) { 597 f2fs_msg(sb, KERN_ERR, 598 "Should set mode=lfs with %uKB-sized IO", 599 F2FS_IO_SIZE_KB(sbi)); 600 return -EINVAL; 601 } 602 return 0; 603 } 604 605 static struct inode *f2fs_alloc_inode(struct super_block *sb) 606 { 607 struct f2fs_inode_info *fi; 608 609 fi = kmem_cache_alloc(f2fs_inode_cachep, GFP_F2FS_ZERO); 610 if (!fi) 611 return NULL; 612 613 init_once((void *) fi); 614 615 /* Initialize f2fs-specific inode info */ 616 fi->vfs_inode.i_version = 1; 617 atomic_set(&fi->dirty_pages, 0); 618 fi->i_current_depth = 1; 619 fi->i_advise = 0; 620 init_rwsem(&fi->i_sem); 621 INIT_LIST_HEAD(&fi->dirty_list); 622 INIT_LIST_HEAD(&fi->gdirty_list); 623 INIT_LIST_HEAD(&fi->inmem_pages); 624 mutex_init(&fi->inmem_lock); 625 init_rwsem(&fi->dio_rwsem[READ]); 626 init_rwsem(&fi->dio_rwsem[WRITE]); 627 628 /* Will be used by directory only */ 629 fi->i_dir_level = F2FS_SB(sb)->dir_level; 630 return &fi->vfs_inode; 631 } 632 633 static int f2fs_drop_inode(struct inode *inode) 634 { 635 int ret; 636 /* 637 * This is to avoid a deadlock condition like below. 638 * writeback_single_inode(inode) 639 * - f2fs_write_data_page 640 * - f2fs_gc -> iput -> evict 641 * - inode_wait_for_writeback(inode) 642 */ 643 if ((!inode_unhashed(inode) && inode->i_state & I_SYNC)) { 644 if (!inode->i_nlink && !is_bad_inode(inode)) { 645 /* to avoid evict_inode call simultaneously */ 646 atomic_inc(&inode->i_count); 647 spin_unlock(&inode->i_lock); 648 649 /* some remained atomic pages should discarded */ 650 if (f2fs_is_atomic_file(inode)) 651 drop_inmem_pages(inode); 652 653 /* should remain fi->extent_tree for writepage */ 654 f2fs_destroy_extent_node(inode); 655 656 sb_start_intwrite(inode->i_sb); 657 f2fs_i_size_write(inode, 0); 658 659 if (F2FS_HAS_BLOCKS(inode)) 660 f2fs_truncate(inode); 661 662 sb_end_intwrite(inode->i_sb); 663 664 fscrypt_put_encryption_info(inode, NULL); 665 spin_lock(&inode->i_lock); 666 atomic_dec(&inode->i_count); 667 } 668 trace_f2fs_drop_inode(inode, 0); 669 return 0; 670 } 671 ret = generic_drop_inode(inode); 672 trace_f2fs_drop_inode(inode, ret); 673 return ret; 674 } 675 676 int f2fs_inode_dirtied(struct inode *inode, bool sync) 677 { 678 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 679 int ret = 0; 680 681 spin_lock(&sbi->inode_lock[DIRTY_META]); 682 if (is_inode_flag_set(inode, FI_DIRTY_INODE)) { 683 ret = 1; 684 } else { 685 set_inode_flag(inode, FI_DIRTY_INODE); 686 stat_inc_dirty_inode(sbi, DIRTY_META); 687 } 688 if (sync && list_empty(&F2FS_I(inode)->gdirty_list)) { 689 list_add_tail(&F2FS_I(inode)->gdirty_list, 690 &sbi->inode_list[DIRTY_META]); 691 inc_page_count(sbi, F2FS_DIRTY_IMETA); 692 } 693 spin_unlock(&sbi->inode_lock[DIRTY_META]); 694 return ret; 695 } 696 697 void f2fs_inode_synced(struct inode *inode) 698 { 699 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 700 701 spin_lock(&sbi->inode_lock[DIRTY_META]); 702 if (!is_inode_flag_set(inode, FI_DIRTY_INODE)) { 703 spin_unlock(&sbi->inode_lock[DIRTY_META]); 704 return; 705 } 706 if (!list_empty(&F2FS_I(inode)->gdirty_list)) { 707 list_del_init(&F2FS_I(inode)->gdirty_list); 708 dec_page_count(sbi, F2FS_DIRTY_IMETA); 709 } 710 clear_inode_flag(inode, FI_DIRTY_INODE); 711 clear_inode_flag(inode, FI_AUTO_RECOVER); 712 stat_dec_dirty_inode(F2FS_I_SB(inode), DIRTY_META); 713 spin_unlock(&sbi->inode_lock[DIRTY_META]); 714 } 715 716 /* 717 * f2fs_dirty_inode() is called from __mark_inode_dirty() 718 * 719 * We should call set_dirty_inode to write the dirty inode through write_inode. 720 */ 721 static void f2fs_dirty_inode(struct inode *inode, int flags) 722 { 723 struct f2fs_sb_info *sbi = F2FS_I_SB(inode); 724 725 if (inode->i_ino == F2FS_NODE_INO(sbi) || 726 inode->i_ino == F2FS_META_INO(sbi)) 727 return; 728 729 if (flags == I_DIRTY_TIME) 730 return; 731 732 if (is_inode_flag_set(inode, FI_AUTO_RECOVER)) 733 clear_inode_flag(inode, FI_AUTO_RECOVER); 734 735 f2fs_inode_dirtied(inode, false); 736 } 737 738 static void f2fs_i_callback(struct rcu_head *head) 739 { 740 struct inode *inode = container_of(head, struct inode, i_rcu); 741 kmem_cache_free(f2fs_inode_cachep, F2FS_I(inode)); 742 } 743 744 static void f2fs_destroy_inode(struct inode *inode) 745 { 746 call_rcu(&inode->i_rcu, f2fs_i_callback); 747 } 748 749 static void destroy_percpu_info(struct f2fs_sb_info *sbi) 750 { 751 percpu_counter_destroy(&sbi->alloc_valid_block_count); 752 percpu_counter_destroy(&sbi->total_valid_inode_count); 753 } 754 755 static void destroy_device_list(struct f2fs_sb_info *sbi) 756 { 757 int i; 758 759 for (i = 0; i < sbi->s_ndevs; i++) { 760 blkdev_put(FDEV(i).bdev, FMODE_EXCL); 761 #ifdef CONFIG_BLK_DEV_ZONED 762 kfree(FDEV(i).blkz_type); 763 #endif 764 } 765 kfree(sbi->devs); 766 } 767 768 static void f2fs_put_super(struct super_block *sb) 769 { 770 struct f2fs_sb_info *sbi = F2FS_SB(sb); 771 772 if (sbi->s_proc) { 773 remove_proc_entry("segment_info", sbi->s_proc); 774 remove_proc_entry("segment_bits", sbi->s_proc); 775 remove_proc_entry(sb->s_id, f2fs_proc_root); 776 } 777 kobject_del(&sbi->s_kobj); 778 779 stop_gc_thread(sbi); 780 781 /* prevent remaining shrinker jobs */ 782 mutex_lock(&sbi->umount_mutex); 783 784 /* 785 * We don't need to do checkpoint when superblock is clean. 786 * But, the previous checkpoint was not done by umount, it needs to do 787 * clean checkpoint again. 788 */ 789 if (is_sbi_flag_set(sbi, SBI_IS_DIRTY) || 790 !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG)) { 791 struct cp_control cpc = { 792 .reason = CP_UMOUNT, 793 }; 794 write_checkpoint(sbi, &cpc); 795 } 796 797 /* be sure to wait for any on-going discard commands */ 798 f2fs_wait_discard_bio(sbi, NULL_ADDR); 799 800 /* write_checkpoint can update stat informaion */ 801 f2fs_destroy_stats(sbi); 802 803 /* 804 * normally superblock is clean, so we need to release this. 805 * In addition, EIO will skip do checkpoint, we need this as well. 806 */ 807 release_ino_entry(sbi, true); 808 809 f2fs_leave_shrinker(sbi); 810 mutex_unlock(&sbi->umount_mutex); 811 812 /* our cp_error case, we can wait for any writeback page */ 813 f2fs_flush_merged_bios(sbi); 814 815 iput(sbi->node_inode); 816 iput(sbi->meta_inode); 817 818 /* destroy f2fs internal modules */ 819 destroy_node_manager(sbi); 820 destroy_segment_manager(sbi); 821 822 kfree(sbi->ckpt); 823 kobject_put(&sbi->s_kobj); 824 wait_for_completion(&sbi->s_kobj_unregister); 825 826 sb->s_fs_info = NULL; 827 if (sbi->s_chksum_driver) 828 crypto_free_shash(sbi->s_chksum_driver); 829 kfree(sbi->raw_super); 830 831 destroy_device_list(sbi); 832 mempool_destroy(sbi->write_io_dummy); 833 destroy_percpu_info(sbi); 834 kfree(sbi); 835 } 836 837 int f2fs_sync_fs(struct super_block *sb, int sync) 838 { 839 struct f2fs_sb_info *sbi = F2FS_SB(sb); 840 int err = 0; 841 842 trace_f2fs_sync_fs(sb, sync); 843 844 if (sync) { 845 struct cp_control cpc; 846 847 cpc.reason = __get_cp_reason(sbi); 848 849 mutex_lock(&sbi->gc_mutex); 850 err = write_checkpoint(sbi, &cpc); 851 mutex_unlock(&sbi->gc_mutex); 852 } 853 f2fs_trace_ios(NULL, 1); 854 855 return err; 856 } 857 858 static int f2fs_freeze(struct super_block *sb) 859 { 860 if (f2fs_readonly(sb)) 861 return 0; 862 863 /* IO error happened before */ 864 if (unlikely(f2fs_cp_error(F2FS_SB(sb)))) 865 return -EIO; 866 867 /* must be clean, since sync_filesystem() was already called */ 868 if (is_sbi_flag_set(F2FS_SB(sb), SBI_IS_DIRTY)) 869 return -EINVAL; 870 return 0; 871 } 872 873 static int f2fs_unfreeze(struct super_block *sb) 874 { 875 return 0; 876 } 877 878 static int f2fs_statfs(struct dentry *dentry, struct kstatfs *buf) 879 { 880 struct super_block *sb = dentry->d_sb; 881 struct f2fs_sb_info *sbi = F2FS_SB(sb); 882 u64 id = huge_encode_dev(sb->s_bdev->bd_dev); 883 block_t total_count, user_block_count, start_count, ovp_count; 884 885 total_count = le64_to_cpu(sbi->raw_super->block_count); 886 user_block_count = sbi->user_block_count; 887 start_count = le32_to_cpu(sbi->raw_super->segment0_blkaddr); 888 ovp_count = SM_I(sbi)->ovp_segments << sbi->log_blocks_per_seg; 889 buf->f_type = F2FS_SUPER_MAGIC; 890 buf->f_bsize = sbi->blocksize; 891 892 buf->f_blocks = total_count - start_count; 893 buf->f_bfree = user_block_count - valid_user_blocks(sbi) + ovp_count; 894 buf->f_bavail = user_block_count - valid_user_blocks(sbi); 895 896 buf->f_files = sbi->total_node_count - F2FS_RESERVED_NODE_NUM; 897 buf->f_ffree = min(buf->f_files - valid_node_count(sbi), 898 buf->f_bavail); 899 900 buf->f_namelen = F2FS_NAME_LEN; 901 buf->f_fsid.val[0] = (u32)id; 902 buf->f_fsid.val[1] = (u32)(id >> 32); 903 904 return 0; 905 } 906 907 static int f2fs_show_options(struct seq_file *seq, struct dentry *root) 908 { 909 struct f2fs_sb_info *sbi = F2FS_SB(root->d_sb); 910 911 if (!f2fs_readonly(sbi->sb) && test_opt(sbi, BG_GC)) { 912 if (test_opt(sbi, FORCE_FG_GC)) 913 seq_printf(seq, ",background_gc=%s", "sync"); 914 else 915 seq_printf(seq, ",background_gc=%s", "on"); 916 } else { 917 seq_printf(seq, ",background_gc=%s", "off"); 918 } 919 if (test_opt(sbi, DISABLE_ROLL_FORWARD)) 920 seq_puts(seq, ",disable_roll_forward"); 921 if (test_opt(sbi, DISCARD)) 922 seq_puts(seq, ",discard"); 923 if (test_opt(sbi, NOHEAP)) 924 seq_puts(seq, ",no_heap"); 925 else 926 seq_puts(seq, ",heap"); 927 #ifdef CONFIG_F2FS_FS_XATTR 928 if (test_opt(sbi, XATTR_USER)) 929 seq_puts(seq, ",user_xattr"); 930 else 931 seq_puts(seq, ",nouser_xattr"); 932 if (test_opt(sbi, INLINE_XATTR)) 933 seq_puts(seq, ",inline_xattr"); 934 else 935 seq_puts(seq, ",noinline_xattr"); 936 #endif 937 #ifdef CONFIG_F2FS_FS_POSIX_ACL 938 if (test_opt(sbi, POSIX_ACL)) 939 seq_puts(seq, ",acl"); 940 else 941 seq_puts(seq, ",noacl"); 942 #endif 943 if (test_opt(sbi, DISABLE_EXT_IDENTIFY)) 944 seq_puts(seq, ",disable_ext_identify"); 945 if (test_opt(sbi, INLINE_DATA)) 946 seq_puts(seq, ",inline_data"); 947 else 948 seq_puts(seq, ",noinline_data"); 949 if (test_opt(sbi, INLINE_DENTRY)) 950 seq_puts(seq, ",inline_dentry"); 951 else 952 seq_puts(seq, ",noinline_dentry"); 953 if (!f2fs_readonly(sbi->sb) && test_opt(sbi, FLUSH_MERGE)) 954 seq_puts(seq, ",flush_merge"); 955 if (test_opt(sbi, NOBARRIER)) 956 seq_puts(seq, ",nobarrier"); 957 if (test_opt(sbi, FASTBOOT)) 958 seq_puts(seq, ",fastboot"); 959 if (test_opt(sbi, EXTENT_CACHE)) 960 seq_puts(seq, ",extent_cache"); 961 else 962 seq_puts(seq, ",noextent_cache"); 963 if (test_opt(sbi, DATA_FLUSH)) 964 seq_puts(seq, ",data_flush"); 965 966 seq_puts(seq, ",mode="); 967 if (test_opt(sbi, ADAPTIVE)) 968 seq_puts(seq, "adaptive"); 969 else if (test_opt(sbi, LFS)) 970 seq_puts(seq, "lfs"); 971 seq_printf(seq, ",active_logs=%u", sbi->active_logs); 972 if (F2FS_IO_SIZE_BITS(sbi)) 973 seq_printf(seq, ",io_size=%uKB", F2FS_IO_SIZE_KB(sbi)); 974 #ifdef CONFIG_F2FS_FAULT_INJECTION 975 if (test_opt(sbi, FAULT_INJECTION)) 976 seq_puts(seq, ",fault_injection"); 977 #endif 978 979 return 0; 980 } 981 982 static int segment_info_seq_show(struct seq_file *seq, void *offset) 983 { 984 struct super_block *sb = seq->private; 985 struct f2fs_sb_info *sbi = F2FS_SB(sb); 986 unsigned int total_segs = 987 le32_to_cpu(sbi->raw_super->segment_count_main); 988 int i; 989 990 seq_puts(seq, "format: segment_type|valid_blocks\n" 991 "segment_type(0:HD, 1:WD, 2:CD, 3:HN, 4:WN, 5:CN)\n"); 992 993 for (i = 0; i < total_segs; i++) { 994 struct seg_entry *se = get_seg_entry(sbi, i); 995 996 if ((i % 10) == 0) 997 seq_printf(seq, "%-10d", i); 998 seq_printf(seq, "%d|%-3u", se->type, 999 get_valid_blocks(sbi, i, 1)); 1000 if ((i % 10) == 9 || i == (total_segs - 1)) 1001 seq_putc(seq, '\n'); 1002 else 1003 seq_putc(seq, ' '); 1004 } 1005 1006 return 0; 1007 } 1008 1009 static int segment_bits_seq_show(struct seq_file *seq, void *offset) 1010 { 1011 struct super_block *sb = seq->private; 1012 struct f2fs_sb_info *sbi = F2FS_SB(sb); 1013 unsigned int total_segs = 1014 le32_to_cpu(sbi->raw_super->segment_count_main); 1015 int i, j; 1016 1017 seq_puts(seq, "format: segment_type|valid_blocks|bitmaps\n" 1018 "segment_type(0:HD, 1:WD, 2:CD, 3:HN, 4:WN, 5:CN)\n"); 1019 1020 for (i = 0; i < total_segs; i++) { 1021 struct seg_entry *se = get_seg_entry(sbi, i); 1022 1023 seq_printf(seq, "%-10d", i); 1024 seq_printf(seq, "%d|%-3u|", se->type, 1025 get_valid_blocks(sbi, i, 1)); 1026 for (j = 0; j < SIT_VBLOCK_MAP_SIZE; j++) 1027 seq_printf(seq, " %.2x", se->cur_valid_map[j]); 1028 seq_putc(seq, '\n'); 1029 } 1030 return 0; 1031 } 1032 1033 #define F2FS_PROC_FILE_DEF(_name) \ 1034 static int _name##_open_fs(struct inode *inode, struct file *file) \ 1035 { \ 1036 return single_open(file, _name##_seq_show, PDE_DATA(inode)); \ 1037 } \ 1038 \ 1039 static const struct file_operations f2fs_seq_##_name##_fops = { \ 1040 .open = _name##_open_fs, \ 1041 .read = seq_read, \ 1042 .llseek = seq_lseek, \ 1043 .release = single_release, \ 1044 }; 1045 1046 F2FS_PROC_FILE_DEF(segment_info); 1047 F2FS_PROC_FILE_DEF(segment_bits); 1048 1049 static void default_options(struct f2fs_sb_info *sbi) 1050 { 1051 /* init some FS parameters */ 1052 sbi->active_logs = NR_CURSEG_TYPE; 1053 1054 set_opt(sbi, BG_GC); 1055 set_opt(sbi, INLINE_XATTR); 1056 set_opt(sbi, INLINE_DATA); 1057 set_opt(sbi, INLINE_DENTRY); 1058 set_opt(sbi, EXTENT_CACHE); 1059 set_opt(sbi, NOHEAP); 1060 sbi->sb->s_flags |= MS_LAZYTIME; 1061 set_opt(sbi, FLUSH_MERGE); 1062 if (f2fs_sb_mounted_blkzoned(sbi->sb)) { 1063 set_opt_mode(sbi, F2FS_MOUNT_LFS); 1064 set_opt(sbi, DISCARD); 1065 } else { 1066 set_opt_mode(sbi, F2FS_MOUNT_ADAPTIVE); 1067 } 1068 1069 #ifdef CONFIG_F2FS_FS_XATTR 1070 set_opt(sbi, XATTR_USER); 1071 #endif 1072 #ifdef CONFIG_F2FS_FS_POSIX_ACL 1073 set_opt(sbi, POSIX_ACL); 1074 #endif 1075 1076 #ifdef CONFIG_F2FS_FAULT_INJECTION 1077 f2fs_build_fault_attr(sbi, 0); 1078 #endif 1079 } 1080 1081 static int f2fs_remount(struct super_block *sb, int *flags, char *data) 1082 { 1083 struct f2fs_sb_info *sbi = F2FS_SB(sb); 1084 struct f2fs_mount_info org_mount_opt; 1085 int err, active_logs; 1086 bool need_restart_gc = false; 1087 bool need_stop_gc = false; 1088 bool no_extent_cache = !test_opt(sbi, EXTENT_CACHE); 1089 #ifdef CONFIG_F2FS_FAULT_INJECTION 1090 struct f2fs_fault_info ffi = sbi->fault_info; 1091 #endif 1092 1093 /* 1094 * Save the old mount options in case we 1095 * need to restore them. 1096 */ 1097 org_mount_opt = sbi->mount_opt; 1098 active_logs = sbi->active_logs; 1099 1100 /* recover superblocks we couldn't write due to previous RO mount */ 1101 if (!(*flags & MS_RDONLY) && is_sbi_flag_set(sbi, SBI_NEED_SB_WRITE)) { 1102 err = f2fs_commit_super(sbi, false); 1103 f2fs_msg(sb, KERN_INFO, 1104 "Try to recover all the superblocks, ret: %d", err); 1105 if (!err) 1106 clear_sbi_flag(sbi, SBI_NEED_SB_WRITE); 1107 } 1108 1109 sbi->mount_opt.opt = 0; 1110 default_options(sbi); 1111 1112 /* parse mount options */ 1113 err = parse_options(sb, data); 1114 if (err) 1115 goto restore_opts; 1116 1117 /* 1118 * Previous and new state of filesystem is RO, 1119 * so skip checking GC and FLUSH_MERGE conditions. 1120 */ 1121 if (f2fs_readonly(sb) && (*flags & MS_RDONLY)) 1122 goto skip; 1123 1124 /* disallow enable/disable extent_cache dynamically */ 1125 if (no_extent_cache == !!test_opt(sbi, EXTENT_CACHE)) { 1126 err = -EINVAL; 1127 f2fs_msg(sbi->sb, KERN_WARNING, 1128 "switch extent_cache option is not allowed"); 1129 goto restore_opts; 1130 } 1131 1132 /* 1133 * We stop the GC thread if FS is mounted as RO 1134 * or if background_gc = off is passed in mount 1135 * option. Also sync the filesystem. 1136 */ 1137 if ((*flags & MS_RDONLY) || !test_opt(sbi, BG_GC)) { 1138 if (sbi->gc_thread) { 1139 stop_gc_thread(sbi); 1140 need_restart_gc = true; 1141 } 1142 } else if (!sbi->gc_thread) { 1143 err = start_gc_thread(sbi); 1144 if (err) 1145 goto restore_opts; 1146 need_stop_gc = true; 1147 } 1148 1149 if (*flags & MS_RDONLY) { 1150 writeback_inodes_sb(sb, WB_REASON_SYNC); 1151 sync_inodes_sb(sb); 1152 1153 set_sbi_flag(sbi, SBI_IS_DIRTY); 1154 set_sbi_flag(sbi, SBI_IS_CLOSE); 1155 f2fs_sync_fs(sb, 1); 1156 clear_sbi_flag(sbi, SBI_IS_CLOSE); 1157 } 1158 1159 /* 1160 * We stop issue flush thread if FS is mounted as RO 1161 * or if flush_merge is not passed in mount option. 1162 */ 1163 if ((*flags & MS_RDONLY) || !test_opt(sbi, FLUSH_MERGE)) { 1164 clear_opt(sbi, FLUSH_MERGE); 1165 destroy_flush_cmd_control(sbi, false); 1166 } else { 1167 err = create_flush_cmd_control(sbi); 1168 if (err) 1169 goto restore_gc; 1170 } 1171 skip: 1172 /* Update the POSIXACL Flag */ 1173 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) | 1174 (test_opt(sbi, POSIX_ACL) ? MS_POSIXACL : 0); 1175 1176 return 0; 1177 restore_gc: 1178 if (need_restart_gc) { 1179 if (start_gc_thread(sbi)) 1180 f2fs_msg(sbi->sb, KERN_WARNING, 1181 "background gc thread has stopped"); 1182 } else if (need_stop_gc) { 1183 stop_gc_thread(sbi); 1184 } 1185 restore_opts: 1186 sbi->mount_opt = org_mount_opt; 1187 sbi->active_logs = active_logs; 1188 #ifdef CONFIG_F2FS_FAULT_INJECTION 1189 sbi->fault_info = ffi; 1190 #endif 1191 return err; 1192 } 1193 1194 static struct super_operations f2fs_sops = { 1195 .alloc_inode = f2fs_alloc_inode, 1196 .drop_inode = f2fs_drop_inode, 1197 .destroy_inode = f2fs_destroy_inode, 1198 .write_inode = f2fs_write_inode, 1199 .dirty_inode = f2fs_dirty_inode, 1200 .show_options = f2fs_show_options, 1201 .evict_inode = f2fs_evict_inode, 1202 .put_super = f2fs_put_super, 1203 .sync_fs = f2fs_sync_fs, 1204 .freeze_fs = f2fs_freeze, 1205 .unfreeze_fs = f2fs_unfreeze, 1206 .statfs = f2fs_statfs, 1207 .remount_fs = f2fs_remount, 1208 }; 1209 1210 #ifdef CONFIG_F2FS_FS_ENCRYPTION 1211 static int f2fs_get_context(struct inode *inode, void *ctx, size_t len) 1212 { 1213 return f2fs_getxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION, 1214 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT, 1215 ctx, len, NULL); 1216 } 1217 1218 static int f2fs_set_context(struct inode *inode, const void *ctx, size_t len, 1219 void *fs_data) 1220 { 1221 return f2fs_setxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION, 1222 F2FS_XATTR_NAME_ENCRYPTION_CONTEXT, 1223 ctx, len, fs_data, XATTR_CREATE); 1224 } 1225 1226 static unsigned f2fs_max_namelen(struct inode *inode) 1227 { 1228 return S_ISLNK(inode->i_mode) ? 1229 inode->i_sb->s_blocksize : F2FS_NAME_LEN; 1230 } 1231 1232 static const struct fscrypt_operations f2fs_cryptops = { 1233 .key_prefix = "f2fs:", 1234 .get_context = f2fs_get_context, 1235 .set_context = f2fs_set_context, 1236 .is_encrypted = f2fs_encrypted_inode, 1237 .empty_dir = f2fs_empty_dir, 1238 .max_namelen = f2fs_max_namelen, 1239 }; 1240 #else 1241 static const struct fscrypt_operations f2fs_cryptops = { 1242 .is_encrypted = f2fs_encrypted_inode, 1243 }; 1244 #endif 1245 1246 static struct inode *f2fs_nfs_get_inode(struct super_block *sb, 1247 u64 ino, u32 generation) 1248 { 1249 struct f2fs_sb_info *sbi = F2FS_SB(sb); 1250 struct inode *inode; 1251 1252 if (check_nid_range(sbi, ino)) 1253 return ERR_PTR(-ESTALE); 1254 1255 /* 1256 * f2fs_iget isn't quite right if the inode is currently unallocated! 1257 * However f2fs_iget currently does appropriate checks to handle stale 1258 * inodes so everything is OK. 1259 */ 1260 inode = f2fs_iget(sb, ino); 1261 if (IS_ERR(inode)) 1262 return ERR_CAST(inode); 1263 if (unlikely(generation && inode->i_generation != generation)) { 1264 /* we didn't find the right inode.. */ 1265 iput(inode); 1266 return ERR_PTR(-ESTALE); 1267 } 1268 return inode; 1269 } 1270 1271 static struct dentry *f2fs_fh_to_dentry(struct super_block *sb, struct fid *fid, 1272 int fh_len, int fh_type) 1273 { 1274 return generic_fh_to_dentry(sb, fid, fh_len, fh_type, 1275 f2fs_nfs_get_inode); 1276 } 1277 1278 static struct dentry *f2fs_fh_to_parent(struct super_block *sb, struct fid *fid, 1279 int fh_len, int fh_type) 1280 { 1281 return generic_fh_to_parent(sb, fid, fh_len, fh_type, 1282 f2fs_nfs_get_inode); 1283 } 1284 1285 static const struct export_operations f2fs_export_ops = { 1286 .fh_to_dentry = f2fs_fh_to_dentry, 1287 .fh_to_parent = f2fs_fh_to_parent, 1288 .get_parent = f2fs_get_parent, 1289 }; 1290 1291 static loff_t max_file_blocks(void) 1292 { 1293 loff_t result = (DEF_ADDRS_PER_INODE - F2FS_INLINE_XATTR_ADDRS); 1294 loff_t leaf_count = ADDRS_PER_BLOCK; 1295 1296 /* two direct node blocks */ 1297 result += (leaf_count * 2); 1298 1299 /* two indirect node blocks */ 1300 leaf_count *= NIDS_PER_BLOCK; 1301 result += (leaf_count * 2); 1302 1303 /* one double indirect node block */ 1304 leaf_count *= NIDS_PER_BLOCK; 1305 result += leaf_count; 1306 1307 return result; 1308 } 1309 1310 static int __f2fs_commit_super(struct buffer_head *bh, 1311 struct f2fs_super_block *super) 1312 { 1313 lock_buffer(bh); 1314 if (super) 1315 memcpy(bh->b_data + F2FS_SUPER_OFFSET, super, sizeof(*super)); 1316 set_buffer_uptodate(bh); 1317 set_buffer_dirty(bh); 1318 unlock_buffer(bh); 1319 1320 /* it's rare case, we can do fua all the time */ 1321 return __sync_dirty_buffer(bh, REQ_PREFLUSH | REQ_FUA); 1322 } 1323 1324 static inline bool sanity_check_area_boundary(struct f2fs_sb_info *sbi, 1325 struct buffer_head *bh) 1326 { 1327 struct f2fs_super_block *raw_super = (struct f2fs_super_block *) 1328 (bh->b_data + F2FS_SUPER_OFFSET); 1329 struct super_block *sb = sbi->sb; 1330 u32 segment0_blkaddr = le32_to_cpu(raw_super->segment0_blkaddr); 1331 u32 cp_blkaddr = le32_to_cpu(raw_super->cp_blkaddr); 1332 u32 sit_blkaddr = le32_to_cpu(raw_super->sit_blkaddr); 1333 u32 nat_blkaddr = le32_to_cpu(raw_super->nat_blkaddr); 1334 u32 ssa_blkaddr = le32_to_cpu(raw_super->ssa_blkaddr); 1335 u32 main_blkaddr = le32_to_cpu(raw_super->main_blkaddr); 1336 u32 segment_count_ckpt = le32_to_cpu(raw_super->segment_count_ckpt); 1337 u32 segment_count_sit = le32_to_cpu(raw_super->segment_count_sit); 1338 u32 segment_count_nat = le32_to_cpu(raw_super->segment_count_nat); 1339 u32 segment_count_ssa = le32_to_cpu(raw_super->segment_count_ssa); 1340 u32 segment_count_main = le32_to_cpu(raw_super->segment_count_main); 1341 u32 segment_count = le32_to_cpu(raw_super->segment_count); 1342 u32 log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg); 1343 u64 main_end_blkaddr = main_blkaddr + 1344 (segment_count_main << log_blocks_per_seg); 1345 u64 seg_end_blkaddr = segment0_blkaddr + 1346 (segment_count << log_blocks_per_seg); 1347 1348 if (segment0_blkaddr != cp_blkaddr) { 1349 f2fs_msg(sb, KERN_INFO, 1350 "Mismatch start address, segment0(%u) cp_blkaddr(%u)", 1351 segment0_blkaddr, cp_blkaddr); 1352 return true; 1353 } 1354 1355 if (cp_blkaddr + (segment_count_ckpt << log_blocks_per_seg) != 1356 sit_blkaddr) { 1357 f2fs_msg(sb, KERN_INFO, 1358 "Wrong CP boundary, start(%u) end(%u) blocks(%u)", 1359 cp_blkaddr, sit_blkaddr, 1360 segment_count_ckpt << log_blocks_per_seg); 1361 return true; 1362 } 1363 1364 if (sit_blkaddr + (segment_count_sit << log_blocks_per_seg) != 1365 nat_blkaddr) { 1366 f2fs_msg(sb, KERN_INFO, 1367 "Wrong SIT boundary, start(%u) end(%u) blocks(%u)", 1368 sit_blkaddr, nat_blkaddr, 1369 segment_count_sit << log_blocks_per_seg); 1370 return true; 1371 } 1372 1373 if (nat_blkaddr + (segment_count_nat << log_blocks_per_seg) != 1374 ssa_blkaddr) { 1375 f2fs_msg(sb, KERN_INFO, 1376 "Wrong NAT boundary, start(%u) end(%u) blocks(%u)", 1377 nat_blkaddr, ssa_blkaddr, 1378 segment_count_nat << log_blocks_per_seg); 1379 return true; 1380 } 1381 1382 if (ssa_blkaddr + (segment_count_ssa << log_blocks_per_seg) != 1383 main_blkaddr) { 1384 f2fs_msg(sb, KERN_INFO, 1385 "Wrong SSA boundary, start(%u) end(%u) blocks(%u)", 1386 ssa_blkaddr, main_blkaddr, 1387 segment_count_ssa << log_blocks_per_seg); 1388 return true; 1389 } 1390 1391 if (main_end_blkaddr > seg_end_blkaddr) { 1392 f2fs_msg(sb, KERN_INFO, 1393 "Wrong MAIN_AREA boundary, start(%u) end(%u) block(%u)", 1394 main_blkaddr, 1395 segment0_blkaddr + 1396 (segment_count << log_blocks_per_seg), 1397 segment_count_main << log_blocks_per_seg); 1398 return true; 1399 } else if (main_end_blkaddr < seg_end_blkaddr) { 1400 int err = 0; 1401 char *res; 1402 1403 /* fix in-memory information all the time */ 1404 raw_super->segment_count = cpu_to_le32((main_end_blkaddr - 1405 segment0_blkaddr) >> log_blocks_per_seg); 1406 1407 if (f2fs_readonly(sb) || bdev_read_only(sb->s_bdev)) { 1408 set_sbi_flag(sbi, SBI_NEED_SB_WRITE); 1409 res = "internally"; 1410 } else { 1411 err = __f2fs_commit_super(bh, NULL); 1412 res = err ? "failed" : "done"; 1413 } 1414 f2fs_msg(sb, KERN_INFO, 1415 "Fix alignment : %s, start(%u) end(%u) block(%u)", 1416 res, main_blkaddr, 1417 segment0_blkaddr + 1418 (segment_count << log_blocks_per_seg), 1419 segment_count_main << log_blocks_per_seg); 1420 if (err) 1421 return true; 1422 } 1423 return false; 1424 } 1425 1426 static int sanity_check_raw_super(struct f2fs_sb_info *sbi, 1427 struct buffer_head *bh) 1428 { 1429 struct f2fs_super_block *raw_super = (struct f2fs_super_block *) 1430 (bh->b_data + F2FS_SUPER_OFFSET); 1431 struct super_block *sb = sbi->sb; 1432 unsigned int blocksize; 1433 1434 if (F2FS_SUPER_MAGIC != le32_to_cpu(raw_super->magic)) { 1435 f2fs_msg(sb, KERN_INFO, 1436 "Magic Mismatch, valid(0x%x) - read(0x%x)", 1437 F2FS_SUPER_MAGIC, le32_to_cpu(raw_super->magic)); 1438 return 1; 1439 } 1440 1441 /* Currently, support only 4KB page cache size */ 1442 if (F2FS_BLKSIZE != PAGE_SIZE) { 1443 f2fs_msg(sb, KERN_INFO, 1444 "Invalid page_cache_size (%lu), supports only 4KB\n", 1445 PAGE_SIZE); 1446 return 1; 1447 } 1448 1449 /* Currently, support only 4KB block size */ 1450 blocksize = 1 << le32_to_cpu(raw_super->log_blocksize); 1451 if (blocksize != F2FS_BLKSIZE) { 1452 f2fs_msg(sb, KERN_INFO, 1453 "Invalid blocksize (%u), supports only 4KB\n", 1454 blocksize); 1455 return 1; 1456 } 1457 1458 /* check log blocks per segment */ 1459 if (le32_to_cpu(raw_super->log_blocks_per_seg) != 9) { 1460 f2fs_msg(sb, KERN_INFO, 1461 "Invalid log blocks per segment (%u)\n", 1462 le32_to_cpu(raw_super->log_blocks_per_seg)); 1463 return 1; 1464 } 1465 1466 /* Currently, support 512/1024/2048/4096 bytes sector size */ 1467 if (le32_to_cpu(raw_super->log_sectorsize) > 1468 F2FS_MAX_LOG_SECTOR_SIZE || 1469 le32_to_cpu(raw_super->log_sectorsize) < 1470 F2FS_MIN_LOG_SECTOR_SIZE) { 1471 f2fs_msg(sb, KERN_INFO, "Invalid log sectorsize (%u)", 1472 le32_to_cpu(raw_super->log_sectorsize)); 1473 return 1; 1474 } 1475 if (le32_to_cpu(raw_super->log_sectors_per_block) + 1476 le32_to_cpu(raw_super->log_sectorsize) != 1477 F2FS_MAX_LOG_SECTOR_SIZE) { 1478 f2fs_msg(sb, KERN_INFO, 1479 "Invalid log sectors per block(%u) log sectorsize(%u)", 1480 le32_to_cpu(raw_super->log_sectors_per_block), 1481 le32_to_cpu(raw_super->log_sectorsize)); 1482 return 1; 1483 } 1484 1485 /* check reserved ino info */ 1486 if (le32_to_cpu(raw_super->node_ino) != 1 || 1487 le32_to_cpu(raw_super->meta_ino) != 2 || 1488 le32_to_cpu(raw_super->root_ino) != 3) { 1489 f2fs_msg(sb, KERN_INFO, 1490 "Invalid Fs Meta Ino: node(%u) meta(%u) root(%u)", 1491 le32_to_cpu(raw_super->node_ino), 1492 le32_to_cpu(raw_super->meta_ino), 1493 le32_to_cpu(raw_super->root_ino)); 1494 return 1; 1495 } 1496 1497 /* check CP/SIT/NAT/SSA/MAIN_AREA area boundary */ 1498 if (sanity_check_area_boundary(sbi, bh)) 1499 return 1; 1500 1501 return 0; 1502 } 1503 1504 int sanity_check_ckpt(struct f2fs_sb_info *sbi) 1505 { 1506 unsigned int total, fsmeta; 1507 struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi); 1508 struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi); 1509 unsigned int ovp_segments, reserved_segments; 1510 1511 total = le32_to_cpu(raw_super->segment_count); 1512 fsmeta = le32_to_cpu(raw_super->segment_count_ckpt); 1513 fsmeta += le32_to_cpu(raw_super->segment_count_sit); 1514 fsmeta += le32_to_cpu(raw_super->segment_count_nat); 1515 fsmeta += le32_to_cpu(ckpt->rsvd_segment_count); 1516 fsmeta += le32_to_cpu(raw_super->segment_count_ssa); 1517 1518 if (unlikely(fsmeta >= total)) 1519 return 1; 1520 1521 ovp_segments = le32_to_cpu(ckpt->overprov_segment_count); 1522 reserved_segments = le32_to_cpu(ckpt->rsvd_segment_count); 1523 1524 if (unlikely(fsmeta < F2FS_MIN_SEGMENTS || 1525 ovp_segments == 0 || reserved_segments == 0)) { 1526 f2fs_msg(sbi->sb, KERN_ERR, 1527 "Wrong layout: check mkfs.f2fs version"); 1528 return 1; 1529 } 1530 1531 if (unlikely(f2fs_cp_error(sbi))) { 1532 f2fs_msg(sbi->sb, KERN_ERR, "A bug case: need to run fsck"); 1533 return 1; 1534 } 1535 return 0; 1536 } 1537 1538 static void init_sb_info(struct f2fs_sb_info *sbi) 1539 { 1540 struct f2fs_super_block *raw_super = sbi->raw_super; 1541 int i; 1542 1543 sbi->log_sectors_per_block = 1544 le32_to_cpu(raw_super->log_sectors_per_block); 1545 sbi->log_blocksize = le32_to_cpu(raw_super->log_blocksize); 1546 sbi->blocksize = 1 << sbi->log_blocksize; 1547 sbi->log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg); 1548 sbi->blocks_per_seg = 1 << sbi->log_blocks_per_seg; 1549 sbi->segs_per_sec = le32_to_cpu(raw_super->segs_per_sec); 1550 sbi->secs_per_zone = le32_to_cpu(raw_super->secs_per_zone); 1551 sbi->total_sections = le32_to_cpu(raw_super->section_count); 1552 sbi->total_node_count = 1553 (le32_to_cpu(raw_super->segment_count_nat) / 2) 1554 * sbi->blocks_per_seg * NAT_ENTRY_PER_BLOCK; 1555 sbi->root_ino_num = le32_to_cpu(raw_super->root_ino); 1556 sbi->node_ino_num = le32_to_cpu(raw_super->node_ino); 1557 sbi->meta_ino_num = le32_to_cpu(raw_super->meta_ino); 1558 sbi->cur_victim_sec = NULL_SECNO; 1559 sbi->max_victim_search = DEF_MAX_VICTIM_SEARCH; 1560 1561 sbi->dir_level = DEF_DIR_LEVEL; 1562 sbi->interval_time[CP_TIME] = DEF_CP_INTERVAL; 1563 sbi->interval_time[REQ_TIME] = DEF_IDLE_INTERVAL; 1564 clear_sbi_flag(sbi, SBI_NEED_FSCK); 1565 1566 for (i = 0; i < NR_COUNT_TYPE; i++) 1567 atomic_set(&sbi->nr_pages[i], 0); 1568 1569 atomic_set(&sbi->wb_sync_req, 0); 1570 1571 INIT_LIST_HEAD(&sbi->s_list); 1572 mutex_init(&sbi->umount_mutex); 1573 mutex_init(&sbi->wio_mutex[NODE]); 1574 mutex_init(&sbi->wio_mutex[DATA]); 1575 spin_lock_init(&sbi->cp_lock); 1576 } 1577 1578 static int init_percpu_info(struct f2fs_sb_info *sbi) 1579 { 1580 int err; 1581 1582 err = percpu_counter_init(&sbi->alloc_valid_block_count, 0, GFP_KERNEL); 1583 if (err) 1584 return err; 1585 1586 return percpu_counter_init(&sbi->total_valid_inode_count, 0, 1587 GFP_KERNEL); 1588 } 1589 1590 #ifdef CONFIG_BLK_DEV_ZONED 1591 static int init_blkz_info(struct f2fs_sb_info *sbi, int devi) 1592 { 1593 struct block_device *bdev = FDEV(devi).bdev; 1594 sector_t nr_sectors = bdev->bd_part->nr_sects; 1595 sector_t sector = 0; 1596 struct blk_zone *zones; 1597 unsigned int i, nr_zones; 1598 unsigned int n = 0; 1599 int err = -EIO; 1600 1601 if (!f2fs_sb_mounted_blkzoned(sbi->sb)) 1602 return 0; 1603 1604 if (sbi->blocks_per_blkz && sbi->blocks_per_blkz != 1605 SECTOR_TO_BLOCK(bdev_zone_sectors(bdev))) 1606 return -EINVAL; 1607 sbi->blocks_per_blkz = SECTOR_TO_BLOCK(bdev_zone_sectors(bdev)); 1608 if (sbi->log_blocks_per_blkz && sbi->log_blocks_per_blkz != 1609 __ilog2_u32(sbi->blocks_per_blkz)) 1610 return -EINVAL; 1611 sbi->log_blocks_per_blkz = __ilog2_u32(sbi->blocks_per_blkz); 1612 FDEV(devi).nr_blkz = SECTOR_TO_BLOCK(nr_sectors) >> 1613 sbi->log_blocks_per_blkz; 1614 if (nr_sectors & (bdev_zone_sectors(bdev) - 1)) 1615 FDEV(devi).nr_blkz++; 1616 1617 FDEV(devi).blkz_type = kmalloc(FDEV(devi).nr_blkz, GFP_KERNEL); 1618 if (!FDEV(devi).blkz_type) 1619 return -ENOMEM; 1620 1621 #define F2FS_REPORT_NR_ZONES 4096 1622 1623 zones = kcalloc(F2FS_REPORT_NR_ZONES, sizeof(struct blk_zone), 1624 GFP_KERNEL); 1625 if (!zones) 1626 return -ENOMEM; 1627 1628 /* Get block zones type */ 1629 while (zones && sector < nr_sectors) { 1630 1631 nr_zones = F2FS_REPORT_NR_ZONES; 1632 err = blkdev_report_zones(bdev, sector, 1633 zones, &nr_zones, 1634 GFP_KERNEL); 1635 if (err) 1636 break; 1637 if (!nr_zones) { 1638 err = -EIO; 1639 break; 1640 } 1641 1642 for (i = 0; i < nr_zones; i++) { 1643 FDEV(devi).blkz_type[n] = zones[i].type; 1644 sector += zones[i].len; 1645 n++; 1646 } 1647 } 1648 1649 kfree(zones); 1650 1651 return err; 1652 } 1653 #endif 1654 1655 /* 1656 * Read f2fs raw super block. 1657 * Because we have two copies of super block, so read both of them 1658 * to get the first valid one. If any one of them is broken, we pass 1659 * them recovery flag back to the caller. 1660 */ 1661 static int read_raw_super_block(struct f2fs_sb_info *sbi, 1662 struct f2fs_super_block **raw_super, 1663 int *valid_super_block, int *recovery) 1664 { 1665 struct super_block *sb = sbi->sb; 1666 int block; 1667 struct buffer_head *bh; 1668 struct f2fs_super_block *super; 1669 int err = 0; 1670 1671 super = kzalloc(sizeof(struct f2fs_super_block), GFP_KERNEL); 1672 if (!super) 1673 return -ENOMEM; 1674 1675 for (block = 0; block < 2; block++) { 1676 bh = sb_bread(sb, block); 1677 if (!bh) { 1678 f2fs_msg(sb, KERN_ERR, "Unable to read %dth superblock", 1679 block + 1); 1680 err = -EIO; 1681 continue; 1682 } 1683 1684 /* sanity checking of raw super */ 1685 if (sanity_check_raw_super(sbi, bh)) { 1686 f2fs_msg(sb, KERN_ERR, 1687 "Can't find valid F2FS filesystem in %dth superblock", 1688 block + 1); 1689 err = -EINVAL; 1690 brelse(bh); 1691 continue; 1692 } 1693 1694 if (!*raw_super) { 1695 memcpy(super, bh->b_data + F2FS_SUPER_OFFSET, 1696 sizeof(*super)); 1697 *valid_super_block = block; 1698 *raw_super = super; 1699 } 1700 brelse(bh); 1701 } 1702 1703 /* Fail to read any one of the superblocks*/ 1704 if (err < 0) 1705 *recovery = 1; 1706 1707 /* No valid superblock */ 1708 if (!*raw_super) 1709 kfree(super); 1710 else 1711 err = 0; 1712 1713 return err; 1714 } 1715 1716 int f2fs_commit_super(struct f2fs_sb_info *sbi, bool recover) 1717 { 1718 struct buffer_head *bh; 1719 int err; 1720 1721 if ((recover && f2fs_readonly(sbi->sb)) || 1722 bdev_read_only(sbi->sb->s_bdev)) { 1723 set_sbi_flag(sbi, SBI_NEED_SB_WRITE); 1724 return -EROFS; 1725 } 1726 1727 /* write back-up superblock first */ 1728 bh = sb_getblk(sbi->sb, sbi->valid_super_block ? 0: 1); 1729 if (!bh) 1730 return -EIO; 1731 err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi)); 1732 brelse(bh); 1733 1734 /* if we are in recovery path, skip writing valid superblock */ 1735 if (recover || err) 1736 return err; 1737 1738 /* write current valid superblock */ 1739 bh = sb_getblk(sbi->sb, sbi->valid_super_block); 1740 if (!bh) 1741 return -EIO; 1742 err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi)); 1743 brelse(bh); 1744 return err; 1745 } 1746 1747 static int f2fs_scan_devices(struct f2fs_sb_info *sbi) 1748 { 1749 struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi); 1750 unsigned int max_devices = MAX_DEVICES; 1751 int i; 1752 1753 /* Initialize single device information */ 1754 if (!RDEV(0).path[0]) { 1755 if (!bdev_is_zoned(sbi->sb->s_bdev)) 1756 return 0; 1757 max_devices = 1; 1758 } 1759 1760 /* 1761 * Initialize multiple devices information, or single 1762 * zoned block device information. 1763 */ 1764 sbi->devs = kcalloc(max_devices, sizeof(struct f2fs_dev_info), 1765 GFP_KERNEL); 1766 if (!sbi->devs) 1767 return -ENOMEM; 1768 1769 for (i = 0; i < max_devices; i++) { 1770 1771 if (i > 0 && !RDEV(i).path[0]) 1772 break; 1773 1774 if (max_devices == 1) { 1775 /* Single zoned block device mount */ 1776 FDEV(0).bdev = 1777 blkdev_get_by_dev(sbi->sb->s_bdev->bd_dev, 1778 sbi->sb->s_mode, sbi->sb->s_type); 1779 } else { 1780 /* Multi-device mount */ 1781 memcpy(FDEV(i).path, RDEV(i).path, MAX_PATH_LEN); 1782 FDEV(i).total_segments = 1783 le32_to_cpu(RDEV(i).total_segments); 1784 if (i == 0) { 1785 FDEV(i).start_blk = 0; 1786 FDEV(i).end_blk = FDEV(i).start_blk + 1787 (FDEV(i).total_segments << 1788 sbi->log_blocks_per_seg) - 1 + 1789 le32_to_cpu(raw_super->segment0_blkaddr); 1790 } else { 1791 FDEV(i).start_blk = FDEV(i - 1).end_blk + 1; 1792 FDEV(i).end_blk = FDEV(i).start_blk + 1793 (FDEV(i).total_segments << 1794 sbi->log_blocks_per_seg) - 1; 1795 } 1796 FDEV(i).bdev = blkdev_get_by_path(FDEV(i).path, 1797 sbi->sb->s_mode, sbi->sb->s_type); 1798 } 1799 if (IS_ERR(FDEV(i).bdev)) 1800 return PTR_ERR(FDEV(i).bdev); 1801 1802 /* to release errored devices */ 1803 sbi->s_ndevs = i + 1; 1804 1805 #ifdef CONFIG_BLK_DEV_ZONED 1806 if (bdev_zoned_model(FDEV(i).bdev) == BLK_ZONED_HM && 1807 !f2fs_sb_mounted_blkzoned(sbi->sb)) { 1808 f2fs_msg(sbi->sb, KERN_ERR, 1809 "Zoned block device feature not enabled\n"); 1810 return -EINVAL; 1811 } 1812 if (bdev_zoned_model(FDEV(i).bdev) != BLK_ZONED_NONE) { 1813 if (init_blkz_info(sbi, i)) { 1814 f2fs_msg(sbi->sb, KERN_ERR, 1815 "Failed to initialize F2FS blkzone information"); 1816 return -EINVAL; 1817 } 1818 if (max_devices == 1) 1819 break; 1820 f2fs_msg(sbi->sb, KERN_INFO, 1821 "Mount Device [%2d]: %20s, %8u, %8x - %8x (zone: %s)", 1822 i, FDEV(i).path, 1823 FDEV(i).total_segments, 1824 FDEV(i).start_blk, FDEV(i).end_blk, 1825 bdev_zoned_model(FDEV(i).bdev) == BLK_ZONED_HA ? 1826 "Host-aware" : "Host-managed"); 1827 continue; 1828 } 1829 #endif 1830 f2fs_msg(sbi->sb, KERN_INFO, 1831 "Mount Device [%2d]: %20s, %8u, %8x - %8x", 1832 i, FDEV(i).path, 1833 FDEV(i).total_segments, 1834 FDEV(i).start_blk, FDEV(i).end_blk); 1835 } 1836 f2fs_msg(sbi->sb, KERN_INFO, 1837 "IO Block Size: %8d KB", F2FS_IO_SIZE_KB(sbi)); 1838 return 0; 1839 } 1840 1841 static int f2fs_fill_super(struct super_block *sb, void *data, int silent) 1842 { 1843 struct f2fs_sb_info *sbi; 1844 struct f2fs_super_block *raw_super; 1845 struct inode *root; 1846 int err; 1847 bool retry = true, need_fsck = false; 1848 char *options = NULL; 1849 int recovery, i, valid_super_block; 1850 struct curseg_info *seg_i; 1851 1852 try_onemore: 1853 err = -EINVAL; 1854 raw_super = NULL; 1855 valid_super_block = -1; 1856 recovery = 0; 1857 1858 /* allocate memory for f2fs-specific super block info */ 1859 sbi = kzalloc(sizeof(struct f2fs_sb_info), GFP_KERNEL); 1860 if (!sbi) 1861 return -ENOMEM; 1862 1863 sbi->sb = sb; 1864 1865 /* Load the checksum driver */ 1866 sbi->s_chksum_driver = crypto_alloc_shash("crc32", 0, 0); 1867 if (IS_ERR(sbi->s_chksum_driver)) { 1868 f2fs_msg(sb, KERN_ERR, "Cannot load crc32 driver."); 1869 err = PTR_ERR(sbi->s_chksum_driver); 1870 sbi->s_chksum_driver = NULL; 1871 goto free_sbi; 1872 } 1873 1874 /* set a block size */ 1875 if (unlikely(!sb_set_blocksize(sb, F2FS_BLKSIZE))) { 1876 f2fs_msg(sb, KERN_ERR, "unable to set blocksize"); 1877 goto free_sbi; 1878 } 1879 1880 err = read_raw_super_block(sbi, &raw_super, &valid_super_block, 1881 &recovery); 1882 if (err) 1883 goto free_sbi; 1884 1885 sb->s_fs_info = sbi; 1886 sbi->raw_super = raw_super; 1887 1888 /* 1889 * The BLKZONED feature indicates that the drive was formatted with 1890 * zone alignment optimization. This is optional for host-aware 1891 * devices, but mandatory for host-managed zoned block devices. 1892 */ 1893 #ifndef CONFIG_BLK_DEV_ZONED 1894 if (f2fs_sb_mounted_blkzoned(sb)) { 1895 f2fs_msg(sb, KERN_ERR, 1896 "Zoned block device support is not enabled\n"); 1897 goto free_sb_buf; 1898 } 1899 #endif 1900 default_options(sbi); 1901 /* parse mount options */ 1902 options = kstrdup((const char *)data, GFP_KERNEL); 1903 if (data && !options) { 1904 err = -ENOMEM; 1905 goto free_sb_buf; 1906 } 1907 1908 err = parse_options(sb, options); 1909 if (err) 1910 goto free_options; 1911 1912 sbi->max_file_blocks = max_file_blocks(); 1913 sb->s_maxbytes = sbi->max_file_blocks << 1914 le32_to_cpu(raw_super->log_blocksize); 1915 sb->s_max_links = F2FS_LINK_MAX; 1916 get_random_bytes(&sbi->s_next_generation, sizeof(u32)); 1917 1918 sb->s_op = &f2fs_sops; 1919 sb->s_cop = &f2fs_cryptops; 1920 sb->s_xattr = f2fs_xattr_handlers; 1921 sb->s_export_op = &f2fs_export_ops; 1922 sb->s_magic = F2FS_SUPER_MAGIC; 1923 sb->s_time_gran = 1; 1924 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) | 1925 (test_opt(sbi, POSIX_ACL) ? MS_POSIXACL : 0); 1926 memcpy(sb->s_uuid, raw_super->uuid, sizeof(raw_super->uuid)); 1927 1928 /* init f2fs-specific super block info */ 1929 sbi->valid_super_block = valid_super_block; 1930 mutex_init(&sbi->gc_mutex); 1931 mutex_init(&sbi->cp_mutex); 1932 init_rwsem(&sbi->node_write); 1933 init_rwsem(&sbi->node_change); 1934 1935 /* disallow all the data/node/meta page writes */ 1936 set_sbi_flag(sbi, SBI_POR_DOING); 1937 spin_lock_init(&sbi->stat_lock); 1938 1939 init_rwsem(&sbi->read_io.io_rwsem); 1940 sbi->read_io.sbi = sbi; 1941 sbi->read_io.bio = NULL; 1942 for (i = 0; i < NR_PAGE_TYPE; i++) { 1943 init_rwsem(&sbi->write_io[i].io_rwsem); 1944 sbi->write_io[i].sbi = sbi; 1945 sbi->write_io[i].bio = NULL; 1946 } 1947 1948 init_rwsem(&sbi->cp_rwsem); 1949 init_waitqueue_head(&sbi->cp_wait); 1950 init_sb_info(sbi); 1951 1952 err = init_percpu_info(sbi); 1953 if (err) 1954 goto free_options; 1955 1956 if (F2FS_IO_SIZE(sbi) > 1) { 1957 sbi->write_io_dummy = 1958 mempool_create_page_pool(2 * (F2FS_IO_SIZE(sbi) - 1), 0); 1959 if (!sbi->write_io_dummy) 1960 goto free_options; 1961 } 1962 1963 /* get an inode for meta space */ 1964 sbi->meta_inode = f2fs_iget(sb, F2FS_META_INO(sbi)); 1965 if (IS_ERR(sbi->meta_inode)) { 1966 f2fs_msg(sb, KERN_ERR, "Failed to read F2FS meta data inode"); 1967 err = PTR_ERR(sbi->meta_inode); 1968 goto free_io_dummy; 1969 } 1970 1971 err = get_valid_checkpoint(sbi); 1972 if (err) { 1973 f2fs_msg(sb, KERN_ERR, "Failed to get valid F2FS checkpoint"); 1974 goto free_meta_inode; 1975 } 1976 1977 /* Initialize device list */ 1978 err = f2fs_scan_devices(sbi); 1979 if (err) { 1980 f2fs_msg(sb, KERN_ERR, "Failed to find devices"); 1981 goto free_devices; 1982 } 1983 1984 sbi->total_valid_node_count = 1985 le32_to_cpu(sbi->ckpt->valid_node_count); 1986 percpu_counter_set(&sbi->total_valid_inode_count, 1987 le32_to_cpu(sbi->ckpt->valid_inode_count)); 1988 sbi->user_block_count = le64_to_cpu(sbi->ckpt->user_block_count); 1989 sbi->total_valid_block_count = 1990 le64_to_cpu(sbi->ckpt->valid_block_count); 1991 sbi->last_valid_block_count = sbi->total_valid_block_count; 1992 1993 for (i = 0; i < NR_INODE_TYPE; i++) { 1994 INIT_LIST_HEAD(&sbi->inode_list[i]); 1995 spin_lock_init(&sbi->inode_lock[i]); 1996 } 1997 1998 init_extent_cache_info(sbi); 1999 2000 init_ino_entry_info(sbi); 2001 2002 /* setup f2fs internal modules */ 2003 err = build_segment_manager(sbi); 2004 if (err) { 2005 f2fs_msg(sb, KERN_ERR, 2006 "Failed to initialize F2FS segment manager"); 2007 goto free_sm; 2008 } 2009 err = build_node_manager(sbi); 2010 if (err) { 2011 f2fs_msg(sb, KERN_ERR, 2012 "Failed to initialize F2FS node manager"); 2013 goto free_nm; 2014 } 2015 2016 /* For write statistics */ 2017 if (sb->s_bdev->bd_part) 2018 sbi->sectors_written_start = 2019 (u64)part_stat_read(sb->s_bdev->bd_part, sectors[1]); 2020 2021 /* Read accumulated write IO statistics if exists */ 2022 seg_i = CURSEG_I(sbi, CURSEG_HOT_NODE); 2023 if (__exist_node_summaries(sbi)) 2024 sbi->kbytes_written = 2025 le64_to_cpu(seg_i->journal->info.kbytes_written); 2026 2027 build_gc_manager(sbi); 2028 2029 /* get an inode for node space */ 2030 sbi->node_inode = f2fs_iget(sb, F2FS_NODE_INO(sbi)); 2031 if (IS_ERR(sbi->node_inode)) { 2032 f2fs_msg(sb, KERN_ERR, "Failed to read node inode"); 2033 err = PTR_ERR(sbi->node_inode); 2034 goto free_nm; 2035 } 2036 2037 f2fs_join_shrinker(sbi); 2038 2039 err = f2fs_build_stats(sbi); 2040 if (err) 2041 goto free_nm; 2042 2043 /* if there are nt orphan nodes free them */ 2044 err = recover_orphan_inodes(sbi); 2045 if (err) 2046 goto free_node_inode; 2047 2048 /* read root inode and dentry */ 2049 root = f2fs_iget(sb, F2FS_ROOT_INO(sbi)); 2050 if (IS_ERR(root)) { 2051 f2fs_msg(sb, KERN_ERR, "Failed to read root inode"); 2052 err = PTR_ERR(root); 2053 goto free_node_inode; 2054 } 2055 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) { 2056 iput(root); 2057 err = -EINVAL; 2058 goto free_node_inode; 2059 } 2060 2061 sb->s_root = d_make_root(root); /* allocate root dentry */ 2062 if (!sb->s_root) { 2063 err = -ENOMEM; 2064 goto free_root_inode; 2065 } 2066 2067 if (f2fs_proc_root) 2068 sbi->s_proc = proc_mkdir(sb->s_id, f2fs_proc_root); 2069 2070 if (sbi->s_proc) { 2071 proc_create_data("segment_info", S_IRUGO, sbi->s_proc, 2072 &f2fs_seq_segment_info_fops, sb); 2073 proc_create_data("segment_bits", S_IRUGO, sbi->s_proc, 2074 &f2fs_seq_segment_bits_fops, sb); 2075 } 2076 2077 sbi->s_kobj.kset = f2fs_kset; 2078 init_completion(&sbi->s_kobj_unregister); 2079 err = kobject_init_and_add(&sbi->s_kobj, &f2fs_ktype, NULL, 2080 "%s", sb->s_id); 2081 if (err) 2082 goto free_proc; 2083 2084 /* recover fsynced data */ 2085 if (!test_opt(sbi, DISABLE_ROLL_FORWARD)) { 2086 /* 2087 * mount should be failed, when device has readonly mode, and 2088 * previous checkpoint was not done by clean system shutdown. 2089 */ 2090 if (bdev_read_only(sb->s_bdev) && 2091 !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG)) { 2092 err = -EROFS; 2093 goto free_kobj; 2094 } 2095 2096 if (need_fsck) 2097 set_sbi_flag(sbi, SBI_NEED_FSCK); 2098 2099 if (!retry) 2100 goto skip_recovery; 2101 2102 err = recover_fsync_data(sbi, false); 2103 if (err < 0) { 2104 need_fsck = true; 2105 f2fs_msg(sb, KERN_ERR, 2106 "Cannot recover all fsync data errno=%d", err); 2107 goto free_kobj; 2108 } 2109 } else { 2110 err = recover_fsync_data(sbi, true); 2111 2112 if (!f2fs_readonly(sb) && err > 0) { 2113 err = -EINVAL; 2114 f2fs_msg(sb, KERN_ERR, 2115 "Need to recover fsync data"); 2116 goto free_kobj; 2117 } 2118 } 2119 skip_recovery: 2120 /* recover_fsync_data() cleared this already */ 2121 clear_sbi_flag(sbi, SBI_POR_DOING); 2122 2123 /* 2124 * If filesystem is not mounted as read-only then 2125 * do start the gc_thread. 2126 */ 2127 if (test_opt(sbi, BG_GC) && !f2fs_readonly(sb)) { 2128 /* After POR, we can run background GC thread.*/ 2129 err = start_gc_thread(sbi); 2130 if (err) 2131 goto free_kobj; 2132 } 2133 kfree(options); 2134 2135 /* recover broken superblock */ 2136 if (recovery) { 2137 err = f2fs_commit_super(sbi, true); 2138 f2fs_msg(sb, KERN_INFO, 2139 "Try to recover %dth superblock, ret: %d", 2140 sbi->valid_super_block ? 1 : 2, err); 2141 } 2142 2143 f2fs_msg(sbi->sb, KERN_NOTICE, "Mounted with checkpoint version = %llx", 2144 cur_cp_version(F2FS_CKPT(sbi))); 2145 f2fs_update_time(sbi, CP_TIME); 2146 f2fs_update_time(sbi, REQ_TIME); 2147 return 0; 2148 2149 free_kobj: 2150 f2fs_sync_inode_meta(sbi); 2151 kobject_del(&sbi->s_kobj); 2152 kobject_put(&sbi->s_kobj); 2153 wait_for_completion(&sbi->s_kobj_unregister); 2154 free_proc: 2155 if (sbi->s_proc) { 2156 remove_proc_entry("segment_info", sbi->s_proc); 2157 remove_proc_entry("segment_bits", sbi->s_proc); 2158 remove_proc_entry(sb->s_id, f2fs_proc_root); 2159 } 2160 free_root_inode: 2161 dput(sb->s_root); 2162 sb->s_root = NULL; 2163 free_node_inode: 2164 truncate_inode_pages_final(NODE_MAPPING(sbi)); 2165 mutex_lock(&sbi->umount_mutex); 2166 release_ino_entry(sbi, true); 2167 f2fs_leave_shrinker(sbi); 2168 /* 2169 * Some dirty meta pages can be produced by recover_orphan_inodes() 2170 * failed by EIO. Then, iput(node_inode) can trigger balance_fs_bg() 2171 * followed by write_checkpoint() through f2fs_write_node_pages(), which 2172 * falls into an infinite loop in sync_meta_pages(). 2173 */ 2174 truncate_inode_pages_final(META_MAPPING(sbi)); 2175 iput(sbi->node_inode); 2176 mutex_unlock(&sbi->umount_mutex); 2177 f2fs_destroy_stats(sbi); 2178 free_nm: 2179 destroy_node_manager(sbi); 2180 free_sm: 2181 destroy_segment_manager(sbi); 2182 free_devices: 2183 destroy_device_list(sbi); 2184 kfree(sbi->ckpt); 2185 free_meta_inode: 2186 make_bad_inode(sbi->meta_inode); 2187 iput(sbi->meta_inode); 2188 free_io_dummy: 2189 mempool_destroy(sbi->write_io_dummy); 2190 free_options: 2191 destroy_percpu_info(sbi); 2192 kfree(options); 2193 free_sb_buf: 2194 kfree(raw_super); 2195 free_sbi: 2196 if (sbi->s_chksum_driver) 2197 crypto_free_shash(sbi->s_chksum_driver); 2198 kfree(sbi); 2199 2200 /* give only one another chance */ 2201 if (retry) { 2202 retry = false; 2203 shrink_dcache_sb(sb); 2204 goto try_onemore; 2205 } 2206 return err; 2207 } 2208 2209 static struct dentry *f2fs_mount(struct file_system_type *fs_type, int flags, 2210 const char *dev_name, void *data) 2211 { 2212 return mount_bdev(fs_type, flags, dev_name, data, f2fs_fill_super); 2213 } 2214 2215 static void kill_f2fs_super(struct super_block *sb) 2216 { 2217 if (sb->s_root) 2218 set_sbi_flag(F2FS_SB(sb), SBI_IS_CLOSE); 2219 kill_block_super(sb); 2220 } 2221 2222 static struct file_system_type f2fs_fs_type = { 2223 .owner = THIS_MODULE, 2224 .name = "f2fs", 2225 .mount = f2fs_mount, 2226 .kill_sb = kill_f2fs_super, 2227 .fs_flags = FS_REQUIRES_DEV, 2228 }; 2229 MODULE_ALIAS_FS("f2fs"); 2230 2231 static int __init init_inodecache(void) 2232 { 2233 f2fs_inode_cachep = kmem_cache_create("f2fs_inode_cache", 2234 sizeof(struct f2fs_inode_info), 0, 2235 SLAB_RECLAIM_ACCOUNT|SLAB_ACCOUNT, NULL); 2236 if (!f2fs_inode_cachep) 2237 return -ENOMEM; 2238 return 0; 2239 } 2240 2241 static void destroy_inodecache(void) 2242 { 2243 /* 2244 * Make sure all delayed rcu free inodes are flushed before we 2245 * destroy cache. 2246 */ 2247 rcu_barrier(); 2248 kmem_cache_destroy(f2fs_inode_cachep); 2249 } 2250 2251 static int __init init_f2fs_fs(void) 2252 { 2253 int err; 2254 2255 f2fs_build_trace_ios(); 2256 2257 err = init_inodecache(); 2258 if (err) 2259 goto fail; 2260 err = create_node_manager_caches(); 2261 if (err) 2262 goto free_inodecache; 2263 err = create_segment_manager_caches(); 2264 if (err) 2265 goto free_node_manager_caches; 2266 err = create_checkpoint_caches(); 2267 if (err) 2268 goto free_segment_manager_caches; 2269 err = create_extent_cache(); 2270 if (err) 2271 goto free_checkpoint_caches; 2272 f2fs_kset = kset_create_and_add("f2fs", NULL, fs_kobj); 2273 if (!f2fs_kset) { 2274 err = -ENOMEM; 2275 goto free_extent_cache; 2276 } 2277 err = register_shrinker(&f2fs_shrinker_info); 2278 if (err) 2279 goto free_kset; 2280 2281 err = register_filesystem(&f2fs_fs_type); 2282 if (err) 2283 goto free_shrinker; 2284 err = f2fs_create_root_stats(); 2285 if (err) 2286 goto free_filesystem; 2287 f2fs_proc_root = proc_mkdir("fs/f2fs", NULL); 2288 return 0; 2289 2290 free_filesystem: 2291 unregister_filesystem(&f2fs_fs_type); 2292 free_shrinker: 2293 unregister_shrinker(&f2fs_shrinker_info); 2294 free_kset: 2295 kset_unregister(f2fs_kset); 2296 free_extent_cache: 2297 destroy_extent_cache(); 2298 free_checkpoint_caches: 2299 destroy_checkpoint_caches(); 2300 free_segment_manager_caches: 2301 destroy_segment_manager_caches(); 2302 free_node_manager_caches: 2303 destroy_node_manager_caches(); 2304 free_inodecache: 2305 destroy_inodecache(); 2306 fail: 2307 return err; 2308 } 2309 2310 static void __exit exit_f2fs_fs(void) 2311 { 2312 remove_proc_entry("fs/f2fs", NULL); 2313 f2fs_destroy_root_stats(); 2314 unregister_filesystem(&f2fs_fs_type); 2315 unregister_shrinker(&f2fs_shrinker_info); 2316 kset_unregister(f2fs_kset); 2317 destroy_extent_cache(); 2318 destroy_checkpoint_caches(); 2319 destroy_segment_manager_caches(); 2320 destroy_node_manager_caches(); 2321 destroy_inodecache(); 2322 f2fs_destroy_trace_ios(); 2323 } 2324 2325 module_init(init_f2fs_fs) 2326 module_exit(exit_f2fs_fs) 2327 2328 MODULE_AUTHOR("Samsung Electronics's Praesto Team"); 2329 MODULE_DESCRIPTION("Flash Friendly File System"); 2330 MODULE_LICENSE("GPL"); 2331 2332