1 /* 2 * fs/f2fs/node.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/fs.h> 12 #include <linux/f2fs_fs.h> 13 #include <linux/mpage.h> 14 #include <linux/backing-dev.h> 15 #include <linux/blkdev.h> 16 #include <linux/pagevec.h> 17 #include <linux/swap.h> 18 19 #include "f2fs.h" 20 #include "node.h" 21 #include "segment.h" 22 #include <trace/events/f2fs.h> 23 24 static struct kmem_cache *nat_entry_slab; 25 static struct kmem_cache *free_nid_slab; 26 27 static void clear_node_page_dirty(struct page *page) 28 { 29 struct address_space *mapping = page->mapping; 30 struct f2fs_sb_info *sbi = F2FS_SB(mapping->host->i_sb); 31 unsigned int long flags; 32 33 if (PageDirty(page)) { 34 spin_lock_irqsave(&mapping->tree_lock, flags); 35 radix_tree_tag_clear(&mapping->page_tree, 36 page_index(page), 37 PAGECACHE_TAG_DIRTY); 38 spin_unlock_irqrestore(&mapping->tree_lock, flags); 39 40 clear_page_dirty_for_io(page); 41 dec_page_count(sbi, F2FS_DIRTY_NODES); 42 } 43 ClearPageUptodate(page); 44 } 45 46 static struct page *get_current_nat_page(struct f2fs_sb_info *sbi, nid_t nid) 47 { 48 pgoff_t index = current_nat_addr(sbi, nid); 49 return get_meta_page(sbi, index); 50 } 51 52 static struct page *get_next_nat_page(struct f2fs_sb_info *sbi, nid_t nid) 53 { 54 struct page *src_page; 55 struct page *dst_page; 56 pgoff_t src_off; 57 pgoff_t dst_off; 58 void *src_addr; 59 void *dst_addr; 60 struct f2fs_nm_info *nm_i = NM_I(sbi); 61 62 src_off = current_nat_addr(sbi, nid); 63 dst_off = next_nat_addr(sbi, src_off); 64 65 /* get current nat block page with lock */ 66 src_page = get_meta_page(sbi, src_off); 67 68 /* Dirty src_page means that it is already the new target NAT page. */ 69 if (PageDirty(src_page)) 70 return src_page; 71 72 dst_page = grab_meta_page(sbi, dst_off); 73 74 src_addr = page_address(src_page); 75 dst_addr = page_address(dst_page); 76 memcpy(dst_addr, src_addr, PAGE_CACHE_SIZE); 77 set_page_dirty(dst_page); 78 f2fs_put_page(src_page, 1); 79 80 set_to_next_nat(nm_i, nid); 81 82 return dst_page; 83 } 84 85 /* 86 * Readahead NAT pages 87 */ 88 static void ra_nat_pages(struct f2fs_sb_info *sbi, int nid) 89 { 90 struct address_space *mapping = sbi->meta_inode->i_mapping; 91 struct f2fs_nm_info *nm_i = NM_I(sbi); 92 struct blk_plug plug; 93 struct page *page; 94 pgoff_t index; 95 int i; 96 97 blk_start_plug(&plug); 98 99 for (i = 0; i < FREE_NID_PAGES; i++, nid += NAT_ENTRY_PER_BLOCK) { 100 if (nid >= nm_i->max_nid) 101 nid = 0; 102 index = current_nat_addr(sbi, nid); 103 104 page = grab_cache_page(mapping, index); 105 if (!page) 106 continue; 107 if (PageUptodate(page)) { 108 f2fs_put_page(page, 1); 109 continue; 110 } 111 if (f2fs_readpage(sbi, page, index, READ)) 112 continue; 113 114 f2fs_put_page(page, 0); 115 } 116 blk_finish_plug(&plug); 117 } 118 119 static struct nat_entry *__lookup_nat_cache(struct f2fs_nm_info *nm_i, nid_t n) 120 { 121 return radix_tree_lookup(&nm_i->nat_root, n); 122 } 123 124 static unsigned int __gang_lookup_nat_cache(struct f2fs_nm_info *nm_i, 125 nid_t start, unsigned int nr, struct nat_entry **ep) 126 { 127 return radix_tree_gang_lookup(&nm_i->nat_root, (void **)ep, start, nr); 128 } 129 130 static void __del_from_nat_cache(struct f2fs_nm_info *nm_i, struct nat_entry *e) 131 { 132 list_del(&e->list); 133 radix_tree_delete(&nm_i->nat_root, nat_get_nid(e)); 134 nm_i->nat_cnt--; 135 kmem_cache_free(nat_entry_slab, e); 136 } 137 138 int is_checkpointed_node(struct f2fs_sb_info *sbi, nid_t nid) 139 { 140 struct f2fs_nm_info *nm_i = NM_I(sbi); 141 struct nat_entry *e; 142 int is_cp = 1; 143 144 read_lock(&nm_i->nat_tree_lock); 145 e = __lookup_nat_cache(nm_i, nid); 146 if (e && !e->checkpointed) 147 is_cp = 0; 148 read_unlock(&nm_i->nat_tree_lock); 149 return is_cp; 150 } 151 152 static struct nat_entry *grab_nat_entry(struct f2fs_nm_info *nm_i, nid_t nid) 153 { 154 struct nat_entry *new; 155 156 new = kmem_cache_alloc(nat_entry_slab, GFP_ATOMIC); 157 if (!new) 158 return NULL; 159 if (radix_tree_insert(&nm_i->nat_root, nid, new)) { 160 kmem_cache_free(nat_entry_slab, new); 161 return NULL; 162 } 163 memset(new, 0, sizeof(struct nat_entry)); 164 nat_set_nid(new, nid); 165 list_add_tail(&new->list, &nm_i->nat_entries); 166 nm_i->nat_cnt++; 167 return new; 168 } 169 170 static void cache_nat_entry(struct f2fs_nm_info *nm_i, nid_t nid, 171 struct f2fs_nat_entry *ne) 172 { 173 struct nat_entry *e; 174 retry: 175 write_lock(&nm_i->nat_tree_lock); 176 e = __lookup_nat_cache(nm_i, nid); 177 if (!e) { 178 e = grab_nat_entry(nm_i, nid); 179 if (!e) { 180 write_unlock(&nm_i->nat_tree_lock); 181 goto retry; 182 } 183 nat_set_blkaddr(e, le32_to_cpu(ne->block_addr)); 184 nat_set_ino(e, le32_to_cpu(ne->ino)); 185 nat_set_version(e, ne->version); 186 e->checkpointed = true; 187 } 188 write_unlock(&nm_i->nat_tree_lock); 189 } 190 191 static void set_node_addr(struct f2fs_sb_info *sbi, struct node_info *ni, 192 block_t new_blkaddr) 193 { 194 struct f2fs_nm_info *nm_i = NM_I(sbi); 195 struct nat_entry *e; 196 retry: 197 write_lock(&nm_i->nat_tree_lock); 198 e = __lookup_nat_cache(nm_i, ni->nid); 199 if (!e) { 200 e = grab_nat_entry(nm_i, ni->nid); 201 if (!e) { 202 write_unlock(&nm_i->nat_tree_lock); 203 goto retry; 204 } 205 e->ni = *ni; 206 e->checkpointed = true; 207 BUG_ON(ni->blk_addr == NEW_ADDR); 208 } else if (new_blkaddr == NEW_ADDR) { 209 /* 210 * when nid is reallocated, 211 * previous nat entry can be remained in nat cache. 212 * So, reinitialize it with new information. 213 */ 214 e->ni = *ni; 215 BUG_ON(ni->blk_addr != NULL_ADDR); 216 } 217 218 if (new_blkaddr == NEW_ADDR) 219 e->checkpointed = false; 220 221 /* sanity check */ 222 BUG_ON(nat_get_blkaddr(e) != ni->blk_addr); 223 BUG_ON(nat_get_blkaddr(e) == NULL_ADDR && 224 new_blkaddr == NULL_ADDR); 225 BUG_ON(nat_get_blkaddr(e) == NEW_ADDR && 226 new_blkaddr == NEW_ADDR); 227 BUG_ON(nat_get_blkaddr(e) != NEW_ADDR && 228 nat_get_blkaddr(e) != NULL_ADDR && 229 new_blkaddr == NEW_ADDR); 230 231 /* increament version no as node is removed */ 232 if (nat_get_blkaddr(e) != NEW_ADDR && new_blkaddr == NULL_ADDR) { 233 unsigned char version = nat_get_version(e); 234 nat_set_version(e, inc_node_version(version)); 235 } 236 237 /* change address */ 238 nat_set_blkaddr(e, new_blkaddr); 239 __set_nat_cache_dirty(nm_i, e); 240 write_unlock(&nm_i->nat_tree_lock); 241 } 242 243 static int try_to_free_nats(struct f2fs_sb_info *sbi, int nr_shrink) 244 { 245 struct f2fs_nm_info *nm_i = NM_I(sbi); 246 247 if (nm_i->nat_cnt <= NM_WOUT_THRESHOLD) 248 return 0; 249 250 write_lock(&nm_i->nat_tree_lock); 251 while (nr_shrink && !list_empty(&nm_i->nat_entries)) { 252 struct nat_entry *ne; 253 ne = list_first_entry(&nm_i->nat_entries, 254 struct nat_entry, list); 255 __del_from_nat_cache(nm_i, ne); 256 nr_shrink--; 257 } 258 write_unlock(&nm_i->nat_tree_lock); 259 return nr_shrink; 260 } 261 262 /* 263 * This function returns always success 264 */ 265 void get_node_info(struct f2fs_sb_info *sbi, nid_t nid, struct node_info *ni) 266 { 267 struct f2fs_nm_info *nm_i = NM_I(sbi); 268 struct curseg_info *curseg = CURSEG_I(sbi, CURSEG_HOT_DATA); 269 struct f2fs_summary_block *sum = curseg->sum_blk; 270 nid_t start_nid = START_NID(nid); 271 struct f2fs_nat_block *nat_blk; 272 struct page *page = NULL; 273 struct f2fs_nat_entry ne; 274 struct nat_entry *e; 275 int i; 276 277 memset(&ne, 0, sizeof(struct f2fs_nat_entry)); 278 ni->nid = nid; 279 280 /* Check nat cache */ 281 read_lock(&nm_i->nat_tree_lock); 282 e = __lookup_nat_cache(nm_i, nid); 283 if (e) { 284 ni->ino = nat_get_ino(e); 285 ni->blk_addr = nat_get_blkaddr(e); 286 ni->version = nat_get_version(e); 287 } 288 read_unlock(&nm_i->nat_tree_lock); 289 if (e) 290 return; 291 292 /* Check current segment summary */ 293 mutex_lock(&curseg->curseg_mutex); 294 i = lookup_journal_in_cursum(sum, NAT_JOURNAL, nid, 0); 295 if (i >= 0) { 296 ne = nat_in_journal(sum, i); 297 node_info_from_raw_nat(ni, &ne); 298 } 299 mutex_unlock(&curseg->curseg_mutex); 300 if (i >= 0) 301 goto cache; 302 303 /* Fill node_info from nat page */ 304 page = get_current_nat_page(sbi, start_nid); 305 nat_blk = (struct f2fs_nat_block *)page_address(page); 306 ne = nat_blk->entries[nid - start_nid]; 307 node_info_from_raw_nat(ni, &ne); 308 f2fs_put_page(page, 1); 309 cache: 310 /* cache nat entry */ 311 cache_nat_entry(NM_I(sbi), nid, &ne); 312 } 313 314 /* 315 * The maximum depth is four. 316 * Offset[0] will have raw inode offset. 317 */ 318 static int get_node_path(long block, int offset[4], unsigned int noffset[4]) 319 { 320 const long direct_index = ADDRS_PER_INODE; 321 const long direct_blks = ADDRS_PER_BLOCK; 322 const long dptrs_per_blk = NIDS_PER_BLOCK; 323 const long indirect_blks = ADDRS_PER_BLOCK * NIDS_PER_BLOCK; 324 const long dindirect_blks = indirect_blks * NIDS_PER_BLOCK; 325 int n = 0; 326 int level = 0; 327 328 noffset[0] = 0; 329 330 if (block < direct_index) { 331 offset[n] = block; 332 goto got; 333 } 334 block -= direct_index; 335 if (block < direct_blks) { 336 offset[n++] = NODE_DIR1_BLOCK; 337 noffset[n] = 1; 338 offset[n] = block; 339 level = 1; 340 goto got; 341 } 342 block -= direct_blks; 343 if (block < direct_blks) { 344 offset[n++] = NODE_DIR2_BLOCK; 345 noffset[n] = 2; 346 offset[n] = block; 347 level = 1; 348 goto got; 349 } 350 block -= direct_blks; 351 if (block < indirect_blks) { 352 offset[n++] = NODE_IND1_BLOCK; 353 noffset[n] = 3; 354 offset[n++] = block / direct_blks; 355 noffset[n] = 4 + offset[n - 1]; 356 offset[n] = block % direct_blks; 357 level = 2; 358 goto got; 359 } 360 block -= indirect_blks; 361 if (block < indirect_blks) { 362 offset[n++] = NODE_IND2_BLOCK; 363 noffset[n] = 4 + dptrs_per_blk; 364 offset[n++] = block / direct_blks; 365 noffset[n] = 5 + dptrs_per_blk + offset[n - 1]; 366 offset[n] = block % direct_blks; 367 level = 2; 368 goto got; 369 } 370 block -= indirect_blks; 371 if (block < dindirect_blks) { 372 offset[n++] = NODE_DIND_BLOCK; 373 noffset[n] = 5 + (dptrs_per_blk * 2); 374 offset[n++] = block / indirect_blks; 375 noffset[n] = 6 + (dptrs_per_blk * 2) + 376 offset[n - 1] * (dptrs_per_blk + 1); 377 offset[n++] = (block / direct_blks) % dptrs_per_blk; 378 noffset[n] = 7 + (dptrs_per_blk * 2) + 379 offset[n - 2] * (dptrs_per_blk + 1) + 380 offset[n - 1]; 381 offset[n] = block % direct_blks; 382 level = 3; 383 goto got; 384 } else { 385 BUG(); 386 } 387 got: 388 return level; 389 } 390 391 /* 392 * Caller should call f2fs_put_dnode(dn). 393 * Also, it should grab and release a mutex by calling mutex_lock_op() and 394 * mutex_unlock_op() only if ro is not set RDONLY_NODE. 395 * In the case of RDONLY_NODE, we don't need to care about mutex. 396 */ 397 int get_dnode_of_data(struct dnode_of_data *dn, pgoff_t index, int mode) 398 { 399 struct f2fs_sb_info *sbi = F2FS_SB(dn->inode->i_sb); 400 struct page *npage[4]; 401 struct page *parent; 402 int offset[4]; 403 unsigned int noffset[4]; 404 nid_t nids[4]; 405 int level, i; 406 int err = 0; 407 408 level = get_node_path(index, offset, noffset); 409 410 nids[0] = dn->inode->i_ino; 411 npage[0] = get_node_page(sbi, nids[0]); 412 if (IS_ERR(npage[0])) 413 return PTR_ERR(npage[0]); 414 415 parent = npage[0]; 416 if (level != 0) 417 nids[1] = get_nid(parent, offset[0], true); 418 dn->inode_page = npage[0]; 419 dn->inode_page_locked = true; 420 421 /* get indirect or direct nodes */ 422 for (i = 1; i <= level; i++) { 423 bool done = false; 424 425 if (!nids[i] && mode == ALLOC_NODE) { 426 /* alloc new node */ 427 if (!alloc_nid(sbi, &(nids[i]))) { 428 err = -ENOSPC; 429 goto release_pages; 430 } 431 432 dn->nid = nids[i]; 433 npage[i] = new_node_page(dn, noffset[i]); 434 if (IS_ERR(npage[i])) { 435 alloc_nid_failed(sbi, nids[i]); 436 err = PTR_ERR(npage[i]); 437 goto release_pages; 438 } 439 440 set_nid(parent, offset[i - 1], nids[i], i == 1); 441 alloc_nid_done(sbi, nids[i]); 442 done = true; 443 } else if (mode == LOOKUP_NODE_RA && i == level && level > 1) { 444 npage[i] = get_node_page_ra(parent, offset[i - 1]); 445 if (IS_ERR(npage[i])) { 446 err = PTR_ERR(npage[i]); 447 goto release_pages; 448 } 449 done = true; 450 } 451 if (i == 1) { 452 dn->inode_page_locked = false; 453 unlock_page(parent); 454 } else { 455 f2fs_put_page(parent, 1); 456 } 457 458 if (!done) { 459 npage[i] = get_node_page(sbi, nids[i]); 460 if (IS_ERR(npage[i])) { 461 err = PTR_ERR(npage[i]); 462 f2fs_put_page(npage[0], 0); 463 goto release_out; 464 } 465 } 466 if (i < level) { 467 parent = npage[i]; 468 nids[i + 1] = get_nid(parent, offset[i], false); 469 } 470 } 471 dn->nid = nids[level]; 472 dn->ofs_in_node = offset[level]; 473 dn->node_page = npage[level]; 474 dn->data_blkaddr = datablock_addr(dn->node_page, dn->ofs_in_node); 475 return 0; 476 477 release_pages: 478 f2fs_put_page(parent, 1); 479 if (i > 1) 480 f2fs_put_page(npage[0], 0); 481 release_out: 482 dn->inode_page = NULL; 483 dn->node_page = NULL; 484 return err; 485 } 486 487 static void truncate_node(struct dnode_of_data *dn) 488 { 489 struct f2fs_sb_info *sbi = F2FS_SB(dn->inode->i_sb); 490 struct node_info ni; 491 492 get_node_info(sbi, dn->nid, &ni); 493 if (dn->inode->i_blocks == 0) { 494 BUG_ON(ni.blk_addr != NULL_ADDR); 495 goto invalidate; 496 } 497 BUG_ON(ni.blk_addr == NULL_ADDR); 498 499 /* Deallocate node address */ 500 invalidate_blocks(sbi, ni.blk_addr); 501 dec_valid_node_count(sbi, dn->inode, 1); 502 set_node_addr(sbi, &ni, NULL_ADDR); 503 504 if (dn->nid == dn->inode->i_ino) { 505 remove_orphan_inode(sbi, dn->nid); 506 dec_valid_inode_count(sbi); 507 } else { 508 sync_inode_page(dn); 509 } 510 invalidate: 511 clear_node_page_dirty(dn->node_page); 512 F2FS_SET_SB_DIRT(sbi); 513 514 f2fs_put_page(dn->node_page, 1); 515 dn->node_page = NULL; 516 trace_f2fs_truncate_node(dn->inode, dn->nid, ni.blk_addr); 517 } 518 519 static int truncate_dnode(struct dnode_of_data *dn) 520 { 521 struct f2fs_sb_info *sbi = F2FS_SB(dn->inode->i_sb); 522 struct page *page; 523 524 if (dn->nid == 0) 525 return 1; 526 527 /* get direct node */ 528 page = get_node_page(sbi, dn->nid); 529 if (IS_ERR(page) && PTR_ERR(page) == -ENOENT) 530 return 1; 531 else if (IS_ERR(page)) 532 return PTR_ERR(page); 533 534 /* Make dnode_of_data for parameter */ 535 dn->node_page = page; 536 dn->ofs_in_node = 0; 537 truncate_data_blocks(dn); 538 truncate_node(dn); 539 return 1; 540 } 541 542 static int truncate_nodes(struct dnode_of_data *dn, unsigned int nofs, 543 int ofs, int depth) 544 { 545 struct f2fs_sb_info *sbi = F2FS_SB(dn->inode->i_sb); 546 struct dnode_of_data rdn = *dn; 547 struct page *page; 548 struct f2fs_node *rn; 549 nid_t child_nid; 550 unsigned int child_nofs; 551 int freed = 0; 552 int i, ret; 553 554 if (dn->nid == 0) 555 return NIDS_PER_BLOCK + 1; 556 557 trace_f2fs_truncate_nodes_enter(dn->inode, dn->nid, dn->data_blkaddr); 558 559 page = get_node_page(sbi, dn->nid); 560 if (IS_ERR(page)) { 561 trace_f2fs_truncate_nodes_exit(dn->inode, PTR_ERR(page)); 562 return PTR_ERR(page); 563 } 564 565 rn = (struct f2fs_node *)page_address(page); 566 if (depth < 3) { 567 for (i = ofs; i < NIDS_PER_BLOCK; i++, freed++) { 568 child_nid = le32_to_cpu(rn->in.nid[i]); 569 if (child_nid == 0) 570 continue; 571 rdn.nid = child_nid; 572 ret = truncate_dnode(&rdn); 573 if (ret < 0) 574 goto out_err; 575 set_nid(page, i, 0, false); 576 } 577 } else { 578 child_nofs = nofs + ofs * (NIDS_PER_BLOCK + 1) + 1; 579 for (i = ofs; i < NIDS_PER_BLOCK; i++) { 580 child_nid = le32_to_cpu(rn->in.nid[i]); 581 if (child_nid == 0) { 582 child_nofs += NIDS_PER_BLOCK + 1; 583 continue; 584 } 585 rdn.nid = child_nid; 586 ret = truncate_nodes(&rdn, child_nofs, 0, depth - 1); 587 if (ret == (NIDS_PER_BLOCK + 1)) { 588 set_nid(page, i, 0, false); 589 child_nofs += ret; 590 } else if (ret < 0 && ret != -ENOENT) { 591 goto out_err; 592 } 593 } 594 freed = child_nofs; 595 } 596 597 if (!ofs) { 598 /* remove current indirect node */ 599 dn->node_page = page; 600 truncate_node(dn); 601 freed++; 602 } else { 603 f2fs_put_page(page, 1); 604 } 605 trace_f2fs_truncate_nodes_exit(dn->inode, freed); 606 return freed; 607 608 out_err: 609 f2fs_put_page(page, 1); 610 trace_f2fs_truncate_nodes_exit(dn->inode, ret); 611 return ret; 612 } 613 614 static int truncate_partial_nodes(struct dnode_of_data *dn, 615 struct f2fs_inode *ri, int *offset, int depth) 616 { 617 struct f2fs_sb_info *sbi = F2FS_SB(dn->inode->i_sb); 618 struct page *pages[2]; 619 nid_t nid[3]; 620 nid_t child_nid; 621 int err = 0; 622 int i; 623 int idx = depth - 2; 624 625 nid[0] = le32_to_cpu(ri->i_nid[offset[0] - NODE_DIR1_BLOCK]); 626 if (!nid[0]) 627 return 0; 628 629 /* get indirect nodes in the path */ 630 for (i = 0; i < depth - 1; i++) { 631 /* refernece count'll be increased */ 632 pages[i] = get_node_page(sbi, nid[i]); 633 if (IS_ERR(pages[i])) { 634 depth = i + 1; 635 err = PTR_ERR(pages[i]); 636 goto fail; 637 } 638 nid[i + 1] = get_nid(pages[i], offset[i + 1], false); 639 } 640 641 /* free direct nodes linked to a partial indirect node */ 642 for (i = offset[depth - 1]; i < NIDS_PER_BLOCK; i++) { 643 child_nid = get_nid(pages[idx], i, false); 644 if (!child_nid) 645 continue; 646 dn->nid = child_nid; 647 err = truncate_dnode(dn); 648 if (err < 0) 649 goto fail; 650 set_nid(pages[idx], i, 0, false); 651 } 652 653 if (offset[depth - 1] == 0) { 654 dn->node_page = pages[idx]; 655 dn->nid = nid[idx]; 656 truncate_node(dn); 657 } else { 658 f2fs_put_page(pages[idx], 1); 659 } 660 offset[idx]++; 661 offset[depth - 1] = 0; 662 fail: 663 for (i = depth - 3; i >= 0; i--) 664 f2fs_put_page(pages[i], 1); 665 666 trace_f2fs_truncate_partial_nodes(dn->inode, nid, depth, err); 667 668 return err; 669 } 670 671 /* 672 * All the block addresses of data and nodes should be nullified. 673 */ 674 int truncate_inode_blocks(struct inode *inode, pgoff_t from) 675 { 676 struct f2fs_sb_info *sbi = F2FS_SB(inode->i_sb); 677 struct address_space *node_mapping = sbi->node_inode->i_mapping; 678 int err = 0, cont = 1; 679 int level, offset[4], noffset[4]; 680 unsigned int nofs = 0; 681 struct f2fs_node *rn; 682 struct dnode_of_data dn; 683 struct page *page; 684 685 trace_f2fs_truncate_inode_blocks_enter(inode, from); 686 687 level = get_node_path(from, offset, noffset); 688 restart: 689 page = get_node_page(sbi, inode->i_ino); 690 if (IS_ERR(page)) { 691 trace_f2fs_truncate_inode_blocks_exit(inode, PTR_ERR(page)); 692 return PTR_ERR(page); 693 } 694 695 set_new_dnode(&dn, inode, page, NULL, 0); 696 unlock_page(page); 697 698 rn = page_address(page); 699 switch (level) { 700 case 0: 701 case 1: 702 nofs = noffset[1]; 703 break; 704 case 2: 705 nofs = noffset[1]; 706 if (!offset[level - 1]) 707 goto skip_partial; 708 err = truncate_partial_nodes(&dn, &rn->i, offset, level); 709 if (err < 0 && err != -ENOENT) 710 goto fail; 711 nofs += 1 + NIDS_PER_BLOCK; 712 break; 713 case 3: 714 nofs = 5 + 2 * NIDS_PER_BLOCK; 715 if (!offset[level - 1]) 716 goto skip_partial; 717 err = truncate_partial_nodes(&dn, &rn->i, offset, level); 718 if (err < 0 && err != -ENOENT) 719 goto fail; 720 break; 721 default: 722 BUG(); 723 } 724 725 skip_partial: 726 while (cont) { 727 dn.nid = le32_to_cpu(rn->i.i_nid[offset[0] - NODE_DIR1_BLOCK]); 728 switch (offset[0]) { 729 case NODE_DIR1_BLOCK: 730 case NODE_DIR2_BLOCK: 731 err = truncate_dnode(&dn); 732 break; 733 734 case NODE_IND1_BLOCK: 735 case NODE_IND2_BLOCK: 736 err = truncate_nodes(&dn, nofs, offset[1], 2); 737 break; 738 739 case NODE_DIND_BLOCK: 740 err = truncate_nodes(&dn, nofs, offset[1], 3); 741 cont = 0; 742 break; 743 744 default: 745 BUG(); 746 } 747 if (err < 0 && err != -ENOENT) 748 goto fail; 749 if (offset[1] == 0 && 750 rn->i.i_nid[offset[0] - NODE_DIR1_BLOCK]) { 751 lock_page(page); 752 if (page->mapping != node_mapping) { 753 f2fs_put_page(page, 1); 754 goto restart; 755 } 756 wait_on_page_writeback(page); 757 rn->i.i_nid[offset[0] - NODE_DIR1_BLOCK] = 0; 758 set_page_dirty(page); 759 unlock_page(page); 760 } 761 offset[1] = 0; 762 offset[0]++; 763 nofs += err; 764 } 765 fail: 766 f2fs_put_page(page, 0); 767 trace_f2fs_truncate_inode_blocks_exit(inode, err); 768 return err > 0 ? 0 : err; 769 } 770 771 /* 772 * Caller should grab and release a mutex by calling mutex_lock_op() and 773 * mutex_unlock_op(). 774 */ 775 int remove_inode_page(struct inode *inode) 776 { 777 struct f2fs_sb_info *sbi = F2FS_SB(inode->i_sb); 778 struct page *page; 779 nid_t ino = inode->i_ino; 780 struct dnode_of_data dn; 781 782 page = get_node_page(sbi, ino); 783 if (IS_ERR(page)) 784 return PTR_ERR(page); 785 786 if (F2FS_I(inode)->i_xattr_nid) { 787 nid_t nid = F2FS_I(inode)->i_xattr_nid; 788 struct page *npage = get_node_page(sbi, nid); 789 790 if (IS_ERR(npage)) 791 return PTR_ERR(npage); 792 793 F2FS_I(inode)->i_xattr_nid = 0; 794 set_new_dnode(&dn, inode, page, npage, nid); 795 dn.inode_page_locked = 1; 796 truncate_node(&dn); 797 } 798 799 /* 0 is possible, after f2fs_new_inode() is failed */ 800 BUG_ON(inode->i_blocks != 0 && inode->i_blocks != 1); 801 set_new_dnode(&dn, inode, page, page, ino); 802 truncate_node(&dn); 803 return 0; 804 } 805 806 int new_inode_page(struct inode *inode, const struct qstr *name) 807 { 808 struct page *page; 809 struct dnode_of_data dn; 810 811 /* allocate inode page for new inode */ 812 set_new_dnode(&dn, inode, NULL, NULL, inode->i_ino); 813 page = new_node_page(&dn, 0); 814 init_dent_inode(name, page); 815 if (IS_ERR(page)) 816 return PTR_ERR(page); 817 f2fs_put_page(page, 1); 818 return 0; 819 } 820 821 struct page *new_node_page(struct dnode_of_data *dn, unsigned int ofs) 822 { 823 struct f2fs_sb_info *sbi = F2FS_SB(dn->inode->i_sb); 824 struct address_space *mapping = sbi->node_inode->i_mapping; 825 struct node_info old_ni, new_ni; 826 struct page *page; 827 int err; 828 829 if (is_inode_flag_set(F2FS_I(dn->inode), FI_NO_ALLOC)) 830 return ERR_PTR(-EPERM); 831 832 page = grab_cache_page(mapping, dn->nid); 833 if (!page) 834 return ERR_PTR(-ENOMEM); 835 836 get_node_info(sbi, dn->nid, &old_ni); 837 838 SetPageUptodate(page); 839 fill_node_footer(page, dn->nid, dn->inode->i_ino, ofs, true); 840 841 /* Reinitialize old_ni with new node page */ 842 BUG_ON(old_ni.blk_addr != NULL_ADDR); 843 new_ni = old_ni; 844 new_ni.ino = dn->inode->i_ino; 845 846 if (!inc_valid_node_count(sbi, dn->inode, 1)) { 847 err = -ENOSPC; 848 goto fail; 849 } 850 set_node_addr(sbi, &new_ni, NEW_ADDR); 851 set_cold_node(dn->inode, page); 852 853 dn->node_page = page; 854 sync_inode_page(dn); 855 set_page_dirty(page); 856 if (ofs == 0) 857 inc_valid_inode_count(sbi); 858 859 return page; 860 861 fail: 862 clear_node_page_dirty(page); 863 f2fs_put_page(page, 1); 864 return ERR_PTR(err); 865 } 866 867 /* 868 * Caller should do after getting the following values. 869 * 0: f2fs_put_page(page, 0) 870 * LOCKED_PAGE: f2fs_put_page(page, 1) 871 * error: nothing 872 */ 873 static int read_node_page(struct page *page, int type) 874 { 875 struct f2fs_sb_info *sbi = F2FS_SB(page->mapping->host->i_sb); 876 struct node_info ni; 877 878 get_node_info(sbi, page->index, &ni); 879 880 if (ni.blk_addr == NULL_ADDR) { 881 f2fs_put_page(page, 1); 882 return -ENOENT; 883 } 884 885 if (PageUptodate(page)) 886 return LOCKED_PAGE; 887 888 return f2fs_readpage(sbi, page, ni.blk_addr, type); 889 } 890 891 /* 892 * Readahead a node page 893 */ 894 void ra_node_page(struct f2fs_sb_info *sbi, nid_t nid) 895 { 896 struct address_space *mapping = sbi->node_inode->i_mapping; 897 struct page *apage; 898 int err; 899 900 apage = find_get_page(mapping, nid); 901 if (apage && PageUptodate(apage)) { 902 f2fs_put_page(apage, 0); 903 return; 904 } 905 f2fs_put_page(apage, 0); 906 907 apage = grab_cache_page(mapping, nid); 908 if (!apage) 909 return; 910 911 err = read_node_page(apage, READA); 912 if (err == 0) 913 f2fs_put_page(apage, 0); 914 else if (err == LOCKED_PAGE) 915 f2fs_put_page(apage, 1); 916 return; 917 } 918 919 struct page *get_node_page(struct f2fs_sb_info *sbi, pgoff_t nid) 920 { 921 struct address_space *mapping = sbi->node_inode->i_mapping; 922 struct page *page; 923 int err; 924 repeat: 925 page = grab_cache_page(mapping, nid); 926 if (!page) 927 return ERR_PTR(-ENOMEM); 928 929 err = read_node_page(page, READ_SYNC); 930 if (err < 0) 931 return ERR_PTR(err); 932 else if (err == LOCKED_PAGE) 933 goto got_it; 934 935 lock_page(page); 936 if (!PageUptodate(page)) { 937 f2fs_put_page(page, 1); 938 return ERR_PTR(-EIO); 939 } 940 if (page->mapping != mapping) { 941 f2fs_put_page(page, 1); 942 goto repeat; 943 } 944 got_it: 945 BUG_ON(nid != nid_of_node(page)); 946 mark_page_accessed(page); 947 return page; 948 } 949 950 /* 951 * Return a locked page for the desired node page. 952 * And, readahead MAX_RA_NODE number of node pages. 953 */ 954 struct page *get_node_page_ra(struct page *parent, int start) 955 { 956 struct f2fs_sb_info *sbi = F2FS_SB(parent->mapping->host->i_sb); 957 struct address_space *mapping = sbi->node_inode->i_mapping; 958 struct blk_plug plug; 959 struct page *page; 960 int err, i, end; 961 nid_t nid; 962 963 /* First, try getting the desired direct node. */ 964 nid = get_nid(parent, start, false); 965 if (!nid) 966 return ERR_PTR(-ENOENT); 967 repeat: 968 page = grab_cache_page(mapping, nid); 969 if (!page) 970 return ERR_PTR(-ENOMEM); 971 972 err = read_node_page(page, READ_SYNC); 973 if (err < 0) 974 return ERR_PTR(err); 975 else if (err == LOCKED_PAGE) 976 goto page_hit; 977 978 blk_start_plug(&plug); 979 980 /* Then, try readahead for siblings of the desired node */ 981 end = start + MAX_RA_NODE; 982 end = min(end, NIDS_PER_BLOCK); 983 for (i = start + 1; i < end; i++) { 984 nid = get_nid(parent, i, false); 985 if (!nid) 986 continue; 987 ra_node_page(sbi, nid); 988 } 989 990 blk_finish_plug(&plug); 991 992 lock_page(page); 993 if (page->mapping != mapping) { 994 f2fs_put_page(page, 1); 995 goto repeat; 996 } 997 page_hit: 998 if (!PageUptodate(page)) { 999 f2fs_put_page(page, 1); 1000 return ERR_PTR(-EIO); 1001 } 1002 mark_page_accessed(page); 1003 return page; 1004 } 1005 1006 void sync_inode_page(struct dnode_of_data *dn) 1007 { 1008 if (IS_INODE(dn->node_page) || dn->inode_page == dn->node_page) { 1009 update_inode(dn->inode, dn->node_page); 1010 } else if (dn->inode_page) { 1011 if (!dn->inode_page_locked) 1012 lock_page(dn->inode_page); 1013 update_inode(dn->inode, dn->inode_page); 1014 if (!dn->inode_page_locked) 1015 unlock_page(dn->inode_page); 1016 } else { 1017 update_inode_page(dn->inode); 1018 } 1019 } 1020 1021 int sync_node_pages(struct f2fs_sb_info *sbi, nid_t ino, 1022 struct writeback_control *wbc) 1023 { 1024 struct address_space *mapping = sbi->node_inode->i_mapping; 1025 pgoff_t index, end; 1026 struct pagevec pvec; 1027 int step = ino ? 2 : 0; 1028 int nwritten = 0, wrote = 0; 1029 1030 pagevec_init(&pvec, 0); 1031 1032 next_step: 1033 index = 0; 1034 end = LONG_MAX; 1035 1036 while (index <= end) { 1037 int i, nr_pages; 1038 nr_pages = pagevec_lookup_tag(&pvec, mapping, &index, 1039 PAGECACHE_TAG_DIRTY, 1040 min(end - index, (pgoff_t)PAGEVEC_SIZE-1) + 1); 1041 if (nr_pages == 0) 1042 break; 1043 1044 for (i = 0; i < nr_pages; i++) { 1045 struct page *page = pvec.pages[i]; 1046 1047 /* 1048 * flushing sequence with step: 1049 * 0. indirect nodes 1050 * 1. dentry dnodes 1051 * 2. file dnodes 1052 */ 1053 if (step == 0 && IS_DNODE(page)) 1054 continue; 1055 if (step == 1 && (!IS_DNODE(page) || 1056 is_cold_node(page))) 1057 continue; 1058 if (step == 2 && (!IS_DNODE(page) || 1059 !is_cold_node(page))) 1060 continue; 1061 1062 /* 1063 * If an fsync mode, 1064 * we should not skip writing node pages. 1065 */ 1066 if (ino && ino_of_node(page) == ino) 1067 lock_page(page); 1068 else if (!trylock_page(page)) 1069 continue; 1070 1071 if (unlikely(page->mapping != mapping)) { 1072 continue_unlock: 1073 unlock_page(page); 1074 continue; 1075 } 1076 if (ino && ino_of_node(page) != ino) 1077 goto continue_unlock; 1078 1079 if (!PageDirty(page)) { 1080 /* someone wrote it for us */ 1081 goto continue_unlock; 1082 } 1083 1084 if (!clear_page_dirty_for_io(page)) 1085 goto continue_unlock; 1086 1087 /* called by fsync() */ 1088 if (ino && IS_DNODE(page)) { 1089 int mark = !is_checkpointed_node(sbi, ino); 1090 set_fsync_mark(page, 1); 1091 if (IS_INODE(page)) 1092 set_dentry_mark(page, mark); 1093 nwritten++; 1094 } else { 1095 set_fsync_mark(page, 0); 1096 set_dentry_mark(page, 0); 1097 } 1098 mapping->a_ops->writepage(page, wbc); 1099 wrote++; 1100 1101 if (--wbc->nr_to_write == 0) 1102 break; 1103 } 1104 pagevec_release(&pvec); 1105 cond_resched(); 1106 1107 if (wbc->nr_to_write == 0) { 1108 step = 2; 1109 break; 1110 } 1111 } 1112 1113 if (step < 2) { 1114 step++; 1115 goto next_step; 1116 } 1117 1118 if (wrote) 1119 f2fs_submit_bio(sbi, NODE, wbc->sync_mode == WB_SYNC_ALL); 1120 1121 return nwritten; 1122 } 1123 1124 static int f2fs_write_node_page(struct page *page, 1125 struct writeback_control *wbc) 1126 { 1127 struct f2fs_sb_info *sbi = F2FS_SB(page->mapping->host->i_sb); 1128 nid_t nid; 1129 block_t new_addr; 1130 struct node_info ni; 1131 1132 wait_on_page_writeback(page); 1133 1134 /* get old block addr of this node page */ 1135 nid = nid_of_node(page); 1136 BUG_ON(page->index != nid); 1137 1138 get_node_info(sbi, nid, &ni); 1139 1140 /* This page is already truncated */ 1141 if (ni.blk_addr == NULL_ADDR) { 1142 dec_page_count(sbi, F2FS_DIRTY_NODES); 1143 unlock_page(page); 1144 return 0; 1145 } 1146 1147 if (wbc->for_reclaim) { 1148 dec_page_count(sbi, F2FS_DIRTY_NODES); 1149 wbc->pages_skipped++; 1150 set_page_dirty(page); 1151 return AOP_WRITEPAGE_ACTIVATE; 1152 } 1153 1154 mutex_lock(&sbi->node_write); 1155 set_page_writeback(page); 1156 write_node_page(sbi, page, nid, ni.blk_addr, &new_addr); 1157 set_node_addr(sbi, &ni, new_addr); 1158 dec_page_count(sbi, F2FS_DIRTY_NODES); 1159 mutex_unlock(&sbi->node_write); 1160 unlock_page(page); 1161 return 0; 1162 } 1163 1164 /* 1165 * It is very important to gather dirty pages and write at once, so that we can 1166 * submit a big bio without interfering other data writes. 1167 * Be default, 512 pages (2MB), a segment size, is quite reasonable. 1168 */ 1169 #define COLLECT_DIRTY_NODES 512 1170 static int f2fs_write_node_pages(struct address_space *mapping, 1171 struct writeback_control *wbc) 1172 { 1173 struct f2fs_sb_info *sbi = F2FS_SB(mapping->host->i_sb); 1174 long nr_to_write = wbc->nr_to_write; 1175 1176 /* First check balancing cached NAT entries */ 1177 if (try_to_free_nats(sbi, NAT_ENTRY_PER_BLOCK)) { 1178 f2fs_sync_fs(sbi->sb, true); 1179 return 0; 1180 } 1181 1182 /* collect a number of dirty node pages and write together */ 1183 if (get_pages(sbi, F2FS_DIRTY_NODES) < COLLECT_DIRTY_NODES) 1184 return 0; 1185 1186 /* if mounting is failed, skip writing node pages */ 1187 wbc->nr_to_write = max_hw_blocks(sbi); 1188 sync_node_pages(sbi, 0, wbc); 1189 wbc->nr_to_write = nr_to_write - (max_hw_blocks(sbi) - wbc->nr_to_write); 1190 return 0; 1191 } 1192 1193 static int f2fs_set_node_page_dirty(struct page *page) 1194 { 1195 struct address_space *mapping = page->mapping; 1196 struct f2fs_sb_info *sbi = F2FS_SB(mapping->host->i_sb); 1197 1198 SetPageUptodate(page); 1199 if (!PageDirty(page)) { 1200 __set_page_dirty_nobuffers(page); 1201 inc_page_count(sbi, F2FS_DIRTY_NODES); 1202 SetPagePrivate(page); 1203 return 1; 1204 } 1205 return 0; 1206 } 1207 1208 static void f2fs_invalidate_node_page(struct page *page, unsigned long offset) 1209 { 1210 struct inode *inode = page->mapping->host; 1211 struct f2fs_sb_info *sbi = F2FS_SB(inode->i_sb); 1212 if (PageDirty(page)) 1213 dec_page_count(sbi, F2FS_DIRTY_NODES); 1214 ClearPagePrivate(page); 1215 } 1216 1217 static int f2fs_release_node_page(struct page *page, gfp_t wait) 1218 { 1219 ClearPagePrivate(page); 1220 return 1; 1221 } 1222 1223 /* 1224 * Structure of the f2fs node operations 1225 */ 1226 const struct address_space_operations f2fs_node_aops = { 1227 .writepage = f2fs_write_node_page, 1228 .writepages = f2fs_write_node_pages, 1229 .set_page_dirty = f2fs_set_node_page_dirty, 1230 .invalidatepage = f2fs_invalidate_node_page, 1231 .releasepage = f2fs_release_node_page, 1232 }; 1233 1234 static struct free_nid *__lookup_free_nid_list(nid_t n, struct list_head *head) 1235 { 1236 struct list_head *this; 1237 struct free_nid *i; 1238 list_for_each(this, head) { 1239 i = list_entry(this, struct free_nid, list); 1240 if (i->nid == n) 1241 return i; 1242 } 1243 return NULL; 1244 } 1245 1246 static void __del_from_free_nid_list(struct free_nid *i) 1247 { 1248 list_del(&i->list); 1249 kmem_cache_free(free_nid_slab, i); 1250 } 1251 1252 static int add_free_nid(struct f2fs_nm_info *nm_i, nid_t nid, bool build) 1253 { 1254 struct free_nid *i; 1255 struct nat_entry *ne; 1256 bool allocated = false; 1257 1258 if (nm_i->fcnt > 2 * MAX_FREE_NIDS) 1259 return -1; 1260 1261 /* 0 nid should not be used */ 1262 if (nid == 0) 1263 return 0; 1264 1265 if (!build) 1266 goto retry; 1267 1268 /* do not add allocated nids */ 1269 read_lock(&nm_i->nat_tree_lock); 1270 ne = __lookup_nat_cache(nm_i, nid); 1271 if (ne && nat_get_blkaddr(ne) != NULL_ADDR) 1272 allocated = true; 1273 read_unlock(&nm_i->nat_tree_lock); 1274 if (allocated) 1275 return 0; 1276 retry: 1277 i = kmem_cache_alloc(free_nid_slab, GFP_NOFS); 1278 if (!i) { 1279 cond_resched(); 1280 goto retry; 1281 } 1282 i->nid = nid; 1283 i->state = NID_NEW; 1284 1285 spin_lock(&nm_i->free_nid_list_lock); 1286 if (__lookup_free_nid_list(nid, &nm_i->free_nid_list)) { 1287 spin_unlock(&nm_i->free_nid_list_lock); 1288 kmem_cache_free(free_nid_slab, i); 1289 return 0; 1290 } 1291 list_add_tail(&i->list, &nm_i->free_nid_list); 1292 nm_i->fcnt++; 1293 spin_unlock(&nm_i->free_nid_list_lock); 1294 return 1; 1295 } 1296 1297 static void remove_free_nid(struct f2fs_nm_info *nm_i, nid_t nid) 1298 { 1299 struct free_nid *i; 1300 spin_lock(&nm_i->free_nid_list_lock); 1301 i = __lookup_free_nid_list(nid, &nm_i->free_nid_list); 1302 if (i && i->state == NID_NEW) { 1303 __del_from_free_nid_list(i); 1304 nm_i->fcnt--; 1305 } 1306 spin_unlock(&nm_i->free_nid_list_lock); 1307 } 1308 1309 static void scan_nat_page(struct f2fs_nm_info *nm_i, 1310 struct page *nat_page, nid_t start_nid) 1311 { 1312 struct f2fs_nat_block *nat_blk = page_address(nat_page); 1313 block_t blk_addr; 1314 int i; 1315 1316 i = start_nid % NAT_ENTRY_PER_BLOCK; 1317 1318 for (; i < NAT_ENTRY_PER_BLOCK; i++, start_nid++) { 1319 1320 if (start_nid >= nm_i->max_nid) 1321 break; 1322 1323 blk_addr = le32_to_cpu(nat_blk->entries[i].block_addr); 1324 BUG_ON(blk_addr == NEW_ADDR); 1325 if (blk_addr == NULL_ADDR) { 1326 if (add_free_nid(nm_i, start_nid, true) < 0) 1327 break; 1328 } 1329 } 1330 } 1331 1332 static void build_free_nids(struct f2fs_sb_info *sbi) 1333 { 1334 struct f2fs_nm_info *nm_i = NM_I(sbi); 1335 struct curseg_info *curseg = CURSEG_I(sbi, CURSEG_HOT_DATA); 1336 struct f2fs_summary_block *sum = curseg->sum_blk; 1337 int i = 0; 1338 nid_t nid = nm_i->next_scan_nid; 1339 1340 /* Enough entries */ 1341 if (nm_i->fcnt > NAT_ENTRY_PER_BLOCK) 1342 return; 1343 1344 /* readahead nat pages to be scanned */ 1345 ra_nat_pages(sbi, nid); 1346 1347 while (1) { 1348 struct page *page = get_current_nat_page(sbi, nid); 1349 1350 scan_nat_page(nm_i, page, nid); 1351 f2fs_put_page(page, 1); 1352 1353 nid += (NAT_ENTRY_PER_BLOCK - (nid % NAT_ENTRY_PER_BLOCK)); 1354 if (nid >= nm_i->max_nid) 1355 nid = 0; 1356 1357 if (i++ == FREE_NID_PAGES) 1358 break; 1359 } 1360 1361 /* go to the next free nat pages to find free nids abundantly */ 1362 nm_i->next_scan_nid = nid; 1363 1364 /* find free nids from current sum_pages */ 1365 mutex_lock(&curseg->curseg_mutex); 1366 for (i = 0; i < nats_in_cursum(sum); i++) { 1367 block_t addr = le32_to_cpu(nat_in_journal(sum, i).block_addr); 1368 nid = le32_to_cpu(nid_in_journal(sum, i)); 1369 if (addr == NULL_ADDR) 1370 add_free_nid(nm_i, nid, true); 1371 else 1372 remove_free_nid(nm_i, nid); 1373 } 1374 mutex_unlock(&curseg->curseg_mutex); 1375 } 1376 1377 /* 1378 * If this function returns success, caller can obtain a new nid 1379 * from second parameter of this function. 1380 * The returned nid could be used ino as well as nid when inode is created. 1381 */ 1382 bool alloc_nid(struct f2fs_sb_info *sbi, nid_t *nid) 1383 { 1384 struct f2fs_nm_info *nm_i = NM_I(sbi); 1385 struct free_nid *i = NULL; 1386 struct list_head *this; 1387 retry: 1388 if (sbi->total_valid_node_count + 1 >= nm_i->max_nid) 1389 return false; 1390 1391 spin_lock(&nm_i->free_nid_list_lock); 1392 1393 /* We should not use stale free nids created by build_free_nids */ 1394 if (nm_i->fcnt && !sbi->on_build_free_nids) { 1395 BUG_ON(list_empty(&nm_i->free_nid_list)); 1396 list_for_each(this, &nm_i->free_nid_list) { 1397 i = list_entry(this, struct free_nid, list); 1398 if (i->state == NID_NEW) 1399 break; 1400 } 1401 1402 BUG_ON(i->state != NID_NEW); 1403 *nid = i->nid; 1404 i->state = NID_ALLOC; 1405 nm_i->fcnt--; 1406 spin_unlock(&nm_i->free_nid_list_lock); 1407 return true; 1408 } 1409 spin_unlock(&nm_i->free_nid_list_lock); 1410 1411 /* Let's scan nat pages and its caches to get free nids */ 1412 mutex_lock(&nm_i->build_lock); 1413 sbi->on_build_free_nids = 1; 1414 build_free_nids(sbi); 1415 sbi->on_build_free_nids = 0; 1416 mutex_unlock(&nm_i->build_lock); 1417 goto retry; 1418 } 1419 1420 /* 1421 * alloc_nid() should be called prior to this function. 1422 */ 1423 void alloc_nid_done(struct f2fs_sb_info *sbi, nid_t nid) 1424 { 1425 struct f2fs_nm_info *nm_i = NM_I(sbi); 1426 struct free_nid *i; 1427 1428 spin_lock(&nm_i->free_nid_list_lock); 1429 i = __lookup_free_nid_list(nid, &nm_i->free_nid_list); 1430 BUG_ON(!i || i->state != NID_ALLOC); 1431 __del_from_free_nid_list(i); 1432 spin_unlock(&nm_i->free_nid_list_lock); 1433 } 1434 1435 /* 1436 * alloc_nid() should be called prior to this function. 1437 */ 1438 void alloc_nid_failed(struct f2fs_sb_info *sbi, nid_t nid) 1439 { 1440 struct f2fs_nm_info *nm_i = NM_I(sbi); 1441 struct free_nid *i; 1442 1443 spin_lock(&nm_i->free_nid_list_lock); 1444 i = __lookup_free_nid_list(nid, &nm_i->free_nid_list); 1445 BUG_ON(!i || i->state != NID_ALLOC); 1446 if (nm_i->fcnt > 2 * MAX_FREE_NIDS) { 1447 __del_from_free_nid_list(i); 1448 } else { 1449 i->state = NID_NEW; 1450 nm_i->fcnt++; 1451 } 1452 spin_unlock(&nm_i->free_nid_list_lock); 1453 } 1454 1455 void recover_node_page(struct f2fs_sb_info *sbi, struct page *page, 1456 struct f2fs_summary *sum, struct node_info *ni, 1457 block_t new_blkaddr) 1458 { 1459 rewrite_node_page(sbi, page, sum, ni->blk_addr, new_blkaddr); 1460 set_node_addr(sbi, ni, new_blkaddr); 1461 clear_node_page_dirty(page); 1462 } 1463 1464 int recover_inode_page(struct f2fs_sb_info *sbi, struct page *page) 1465 { 1466 struct address_space *mapping = sbi->node_inode->i_mapping; 1467 struct f2fs_node *src, *dst; 1468 nid_t ino = ino_of_node(page); 1469 struct node_info old_ni, new_ni; 1470 struct page *ipage; 1471 1472 ipage = grab_cache_page(mapping, ino); 1473 if (!ipage) 1474 return -ENOMEM; 1475 1476 /* Should not use this inode from free nid list */ 1477 remove_free_nid(NM_I(sbi), ino); 1478 1479 get_node_info(sbi, ino, &old_ni); 1480 SetPageUptodate(ipage); 1481 fill_node_footer(ipage, ino, ino, 0, true); 1482 1483 src = (struct f2fs_node *)page_address(page); 1484 dst = (struct f2fs_node *)page_address(ipage); 1485 1486 memcpy(dst, src, (unsigned long)&src->i.i_ext - (unsigned long)&src->i); 1487 dst->i.i_size = 0; 1488 dst->i.i_blocks = cpu_to_le64(1); 1489 dst->i.i_links = cpu_to_le32(1); 1490 dst->i.i_xattr_nid = 0; 1491 1492 new_ni = old_ni; 1493 new_ni.ino = ino; 1494 1495 set_node_addr(sbi, &new_ni, NEW_ADDR); 1496 inc_valid_inode_count(sbi); 1497 1498 f2fs_put_page(ipage, 1); 1499 return 0; 1500 } 1501 1502 int restore_node_summary(struct f2fs_sb_info *sbi, 1503 unsigned int segno, struct f2fs_summary_block *sum) 1504 { 1505 struct f2fs_node *rn; 1506 struct f2fs_summary *sum_entry; 1507 struct page *page; 1508 block_t addr; 1509 int i, last_offset; 1510 1511 /* alloc temporal page for read node */ 1512 page = alloc_page(GFP_NOFS | __GFP_ZERO); 1513 if (IS_ERR(page)) 1514 return PTR_ERR(page); 1515 lock_page(page); 1516 1517 /* scan the node segment */ 1518 last_offset = sbi->blocks_per_seg; 1519 addr = START_BLOCK(sbi, segno); 1520 sum_entry = &sum->entries[0]; 1521 1522 for (i = 0; i < last_offset; i++, sum_entry++) { 1523 /* 1524 * In order to read next node page, 1525 * we must clear PageUptodate flag. 1526 */ 1527 ClearPageUptodate(page); 1528 1529 if (f2fs_readpage(sbi, page, addr, READ_SYNC)) 1530 goto out; 1531 1532 lock_page(page); 1533 rn = (struct f2fs_node *)page_address(page); 1534 sum_entry->nid = rn->footer.nid; 1535 sum_entry->version = 0; 1536 sum_entry->ofs_in_node = 0; 1537 addr++; 1538 } 1539 unlock_page(page); 1540 out: 1541 __free_pages(page, 0); 1542 return 0; 1543 } 1544 1545 static bool flush_nats_in_journal(struct f2fs_sb_info *sbi) 1546 { 1547 struct f2fs_nm_info *nm_i = NM_I(sbi); 1548 struct curseg_info *curseg = CURSEG_I(sbi, CURSEG_HOT_DATA); 1549 struct f2fs_summary_block *sum = curseg->sum_blk; 1550 int i; 1551 1552 mutex_lock(&curseg->curseg_mutex); 1553 1554 if (nats_in_cursum(sum) < NAT_JOURNAL_ENTRIES) { 1555 mutex_unlock(&curseg->curseg_mutex); 1556 return false; 1557 } 1558 1559 for (i = 0; i < nats_in_cursum(sum); i++) { 1560 struct nat_entry *ne; 1561 struct f2fs_nat_entry raw_ne; 1562 nid_t nid = le32_to_cpu(nid_in_journal(sum, i)); 1563 1564 raw_ne = nat_in_journal(sum, i); 1565 retry: 1566 write_lock(&nm_i->nat_tree_lock); 1567 ne = __lookup_nat_cache(nm_i, nid); 1568 if (ne) { 1569 __set_nat_cache_dirty(nm_i, ne); 1570 write_unlock(&nm_i->nat_tree_lock); 1571 continue; 1572 } 1573 ne = grab_nat_entry(nm_i, nid); 1574 if (!ne) { 1575 write_unlock(&nm_i->nat_tree_lock); 1576 goto retry; 1577 } 1578 nat_set_blkaddr(ne, le32_to_cpu(raw_ne.block_addr)); 1579 nat_set_ino(ne, le32_to_cpu(raw_ne.ino)); 1580 nat_set_version(ne, raw_ne.version); 1581 __set_nat_cache_dirty(nm_i, ne); 1582 write_unlock(&nm_i->nat_tree_lock); 1583 } 1584 update_nats_in_cursum(sum, -i); 1585 mutex_unlock(&curseg->curseg_mutex); 1586 return true; 1587 } 1588 1589 /* 1590 * This function is called during the checkpointing process. 1591 */ 1592 void flush_nat_entries(struct f2fs_sb_info *sbi) 1593 { 1594 struct f2fs_nm_info *nm_i = NM_I(sbi); 1595 struct curseg_info *curseg = CURSEG_I(sbi, CURSEG_HOT_DATA); 1596 struct f2fs_summary_block *sum = curseg->sum_blk; 1597 struct list_head *cur, *n; 1598 struct page *page = NULL; 1599 struct f2fs_nat_block *nat_blk = NULL; 1600 nid_t start_nid = 0, end_nid = 0; 1601 bool flushed; 1602 1603 flushed = flush_nats_in_journal(sbi); 1604 1605 if (!flushed) 1606 mutex_lock(&curseg->curseg_mutex); 1607 1608 /* 1) flush dirty nat caches */ 1609 list_for_each_safe(cur, n, &nm_i->dirty_nat_entries) { 1610 struct nat_entry *ne; 1611 nid_t nid; 1612 struct f2fs_nat_entry raw_ne; 1613 int offset = -1; 1614 block_t new_blkaddr; 1615 1616 ne = list_entry(cur, struct nat_entry, list); 1617 nid = nat_get_nid(ne); 1618 1619 if (nat_get_blkaddr(ne) == NEW_ADDR) 1620 continue; 1621 if (flushed) 1622 goto to_nat_page; 1623 1624 /* if there is room for nat enries in curseg->sumpage */ 1625 offset = lookup_journal_in_cursum(sum, NAT_JOURNAL, nid, 1); 1626 if (offset >= 0) { 1627 raw_ne = nat_in_journal(sum, offset); 1628 goto flush_now; 1629 } 1630 to_nat_page: 1631 if (!page || (start_nid > nid || nid > end_nid)) { 1632 if (page) { 1633 f2fs_put_page(page, 1); 1634 page = NULL; 1635 } 1636 start_nid = START_NID(nid); 1637 end_nid = start_nid + NAT_ENTRY_PER_BLOCK - 1; 1638 1639 /* 1640 * get nat block with dirty flag, increased reference 1641 * count, mapped and lock 1642 */ 1643 page = get_next_nat_page(sbi, start_nid); 1644 nat_blk = page_address(page); 1645 } 1646 1647 BUG_ON(!nat_blk); 1648 raw_ne = nat_blk->entries[nid - start_nid]; 1649 flush_now: 1650 new_blkaddr = nat_get_blkaddr(ne); 1651 1652 raw_ne.ino = cpu_to_le32(nat_get_ino(ne)); 1653 raw_ne.block_addr = cpu_to_le32(new_blkaddr); 1654 raw_ne.version = nat_get_version(ne); 1655 1656 if (offset < 0) { 1657 nat_blk->entries[nid - start_nid] = raw_ne; 1658 } else { 1659 nat_in_journal(sum, offset) = raw_ne; 1660 nid_in_journal(sum, offset) = cpu_to_le32(nid); 1661 } 1662 1663 if (nat_get_blkaddr(ne) == NULL_ADDR && 1664 add_free_nid(NM_I(sbi), nid, false) <= 0) { 1665 write_lock(&nm_i->nat_tree_lock); 1666 __del_from_nat_cache(nm_i, ne); 1667 write_unlock(&nm_i->nat_tree_lock); 1668 } else { 1669 write_lock(&nm_i->nat_tree_lock); 1670 __clear_nat_cache_dirty(nm_i, ne); 1671 ne->checkpointed = true; 1672 write_unlock(&nm_i->nat_tree_lock); 1673 } 1674 } 1675 if (!flushed) 1676 mutex_unlock(&curseg->curseg_mutex); 1677 f2fs_put_page(page, 1); 1678 1679 /* 2) shrink nat caches if necessary */ 1680 try_to_free_nats(sbi, nm_i->nat_cnt - NM_WOUT_THRESHOLD); 1681 } 1682 1683 static int init_node_manager(struct f2fs_sb_info *sbi) 1684 { 1685 struct f2fs_super_block *sb_raw = F2FS_RAW_SUPER(sbi); 1686 struct f2fs_nm_info *nm_i = NM_I(sbi); 1687 unsigned char *version_bitmap; 1688 unsigned int nat_segs, nat_blocks; 1689 1690 nm_i->nat_blkaddr = le32_to_cpu(sb_raw->nat_blkaddr); 1691 1692 /* segment_count_nat includes pair segment so divide to 2. */ 1693 nat_segs = le32_to_cpu(sb_raw->segment_count_nat) >> 1; 1694 nat_blocks = nat_segs << le32_to_cpu(sb_raw->log_blocks_per_seg); 1695 nm_i->max_nid = NAT_ENTRY_PER_BLOCK * nat_blocks; 1696 nm_i->fcnt = 0; 1697 nm_i->nat_cnt = 0; 1698 1699 INIT_LIST_HEAD(&nm_i->free_nid_list); 1700 INIT_RADIX_TREE(&nm_i->nat_root, GFP_ATOMIC); 1701 INIT_LIST_HEAD(&nm_i->nat_entries); 1702 INIT_LIST_HEAD(&nm_i->dirty_nat_entries); 1703 1704 mutex_init(&nm_i->build_lock); 1705 spin_lock_init(&nm_i->free_nid_list_lock); 1706 rwlock_init(&nm_i->nat_tree_lock); 1707 1708 nm_i->next_scan_nid = le32_to_cpu(sbi->ckpt->next_free_nid); 1709 nm_i->bitmap_size = __bitmap_size(sbi, NAT_BITMAP); 1710 version_bitmap = __bitmap_ptr(sbi, NAT_BITMAP); 1711 if (!version_bitmap) 1712 return -EFAULT; 1713 1714 nm_i->nat_bitmap = kmemdup(version_bitmap, nm_i->bitmap_size, 1715 GFP_KERNEL); 1716 if (!nm_i->nat_bitmap) 1717 return -ENOMEM; 1718 return 0; 1719 } 1720 1721 int build_node_manager(struct f2fs_sb_info *sbi) 1722 { 1723 int err; 1724 1725 sbi->nm_info = kzalloc(sizeof(struct f2fs_nm_info), GFP_KERNEL); 1726 if (!sbi->nm_info) 1727 return -ENOMEM; 1728 1729 err = init_node_manager(sbi); 1730 if (err) 1731 return err; 1732 1733 build_free_nids(sbi); 1734 return 0; 1735 } 1736 1737 void destroy_node_manager(struct f2fs_sb_info *sbi) 1738 { 1739 struct f2fs_nm_info *nm_i = NM_I(sbi); 1740 struct free_nid *i, *next_i; 1741 struct nat_entry *natvec[NATVEC_SIZE]; 1742 nid_t nid = 0; 1743 unsigned int found; 1744 1745 if (!nm_i) 1746 return; 1747 1748 /* destroy free nid list */ 1749 spin_lock(&nm_i->free_nid_list_lock); 1750 list_for_each_entry_safe(i, next_i, &nm_i->free_nid_list, list) { 1751 BUG_ON(i->state == NID_ALLOC); 1752 __del_from_free_nid_list(i); 1753 nm_i->fcnt--; 1754 } 1755 BUG_ON(nm_i->fcnt); 1756 spin_unlock(&nm_i->free_nid_list_lock); 1757 1758 /* destroy nat cache */ 1759 write_lock(&nm_i->nat_tree_lock); 1760 while ((found = __gang_lookup_nat_cache(nm_i, 1761 nid, NATVEC_SIZE, natvec))) { 1762 unsigned idx; 1763 for (idx = 0; idx < found; idx++) { 1764 struct nat_entry *e = natvec[idx]; 1765 nid = nat_get_nid(e) + 1; 1766 __del_from_nat_cache(nm_i, e); 1767 } 1768 } 1769 BUG_ON(nm_i->nat_cnt); 1770 write_unlock(&nm_i->nat_tree_lock); 1771 1772 kfree(nm_i->nat_bitmap); 1773 sbi->nm_info = NULL; 1774 kfree(nm_i); 1775 } 1776 1777 int __init create_node_manager_caches(void) 1778 { 1779 nat_entry_slab = f2fs_kmem_cache_create("nat_entry", 1780 sizeof(struct nat_entry), NULL); 1781 if (!nat_entry_slab) 1782 return -ENOMEM; 1783 1784 free_nid_slab = f2fs_kmem_cache_create("free_nid", 1785 sizeof(struct free_nid), NULL); 1786 if (!free_nid_slab) { 1787 kmem_cache_destroy(nat_entry_slab); 1788 return -ENOMEM; 1789 } 1790 return 0; 1791 } 1792 1793 void destroy_node_manager_caches(void) 1794 { 1795 kmem_cache_destroy(free_nid_slab); 1796 kmem_cache_destroy(nat_entry_slab); 1797 } 1798