1 // SPDX-License-Identifier: GPL-2.0 2 /* 3 * fs/ext4/extents_status.c 4 * 5 * Written by Yongqiang Yang <xiaoqiangnk@gmail.com> 6 * Modified by 7 * Allison Henderson <achender@linux.vnet.ibm.com> 8 * Hugh Dickins <hughd@google.com> 9 * Zheng Liu <wenqing.lz@taobao.com> 10 * 11 * Ext4 extents status tree core functions. 12 */ 13 #include <linux/list_sort.h> 14 #include <linux/proc_fs.h> 15 #include <linux/seq_file.h> 16 #include "ext4.h" 17 18 #include <trace/events/ext4.h> 19 20 /* 21 * According to previous discussion in Ext4 Developer Workshop, we 22 * will introduce a new structure called io tree to track all extent 23 * status in order to solve some problems that we have met 24 * (e.g. Reservation space warning), and provide extent-level locking. 25 * Delay extent tree is the first step to achieve this goal. It is 26 * original built by Yongqiang Yang. At that time it is called delay 27 * extent tree, whose goal is only track delayed extents in memory to 28 * simplify the implementation of fiemap and bigalloc, and introduce 29 * lseek SEEK_DATA/SEEK_HOLE support. That is why it is still called 30 * delay extent tree at the first commit. But for better understand 31 * what it does, it has been rename to extent status tree. 32 * 33 * Step1: 34 * Currently the first step has been done. All delayed extents are 35 * tracked in the tree. It maintains the delayed extent when a delayed 36 * allocation is issued, and the delayed extent is written out or 37 * invalidated. Therefore the implementation of fiemap and bigalloc 38 * are simplified, and SEEK_DATA/SEEK_HOLE are introduced. 39 * 40 * The following comment describes the implemenmtation of extent 41 * status tree and future works. 42 * 43 * Step2: 44 * In this step all extent status are tracked by extent status tree. 45 * Thus, we can first try to lookup a block mapping in this tree before 46 * finding it in extent tree. Hence, single extent cache can be removed 47 * because extent status tree can do a better job. Extents in status 48 * tree are loaded on-demand. Therefore, the extent status tree may not 49 * contain all of the extents in a file. Meanwhile we define a shrinker 50 * to reclaim memory from extent status tree because fragmented extent 51 * tree will make status tree cost too much memory. written/unwritten/- 52 * hole extents in the tree will be reclaimed by this shrinker when we 53 * are under high memory pressure. Delayed extents will not be 54 * reclimed because fiemap, bigalloc, and seek_data/hole need it. 55 */ 56 57 /* 58 * Extent status tree implementation for ext4. 59 * 60 * 61 * ========================================================================== 62 * Extent status tree tracks all extent status. 63 * 64 * 1. Why we need to implement extent status tree? 65 * 66 * Without extent status tree, ext4 identifies a delayed extent by looking 67 * up page cache, this has several deficiencies - complicated, buggy, 68 * and inefficient code. 69 * 70 * FIEMAP, SEEK_HOLE/DATA, bigalloc, and writeout all need to know if a 71 * block or a range of blocks are belonged to a delayed extent. 72 * 73 * Let us have a look at how they do without extent status tree. 74 * -- FIEMAP 75 * FIEMAP looks up page cache to identify delayed allocations from holes. 76 * 77 * -- SEEK_HOLE/DATA 78 * SEEK_HOLE/DATA has the same problem as FIEMAP. 79 * 80 * -- bigalloc 81 * bigalloc looks up page cache to figure out if a block is 82 * already under delayed allocation or not to determine whether 83 * quota reserving is needed for the cluster. 84 * 85 * -- writeout 86 * Writeout looks up whole page cache to see if a buffer is 87 * mapped, If there are not very many delayed buffers, then it is 88 * time consuming. 89 * 90 * With extent status tree implementation, FIEMAP, SEEK_HOLE/DATA, 91 * bigalloc and writeout can figure out if a block or a range of 92 * blocks is under delayed allocation(belonged to a delayed extent) or 93 * not by searching the extent tree. 94 * 95 * 96 * ========================================================================== 97 * 2. Ext4 extent status tree impelmentation 98 * 99 * -- extent 100 * A extent is a range of blocks which are contiguous logically and 101 * physically. Unlike extent in extent tree, this extent in ext4 is 102 * a in-memory struct, there is no corresponding on-disk data. There 103 * is no limit on length of extent, so an extent can contain as many 104 * blocks as they are contiguous logically and physically. 105 * 106 * -- extent status tree 107 * Every inode has an extent status tree and all allocation blocks 108 * are added to the tree with different status. The extent in the 109 * tree are ordered by logical block no. 110 * 111 * -- operations on a extent status tree 112 * There are three important operations on a delayed extent tree: find 113 * next extent, adding a extent(a range of blocks) and removing a extent. 114 * 115 * -- race on a extent status tree 116 * Extent status tree is protected by inode->i_es_lock. 117 * 118 * -- memory consumption 119 * Fragmented extent tree will make extent status tree cost too much 120 * memory. Hence, we will reclaim written/unwritten/hole extents from 121 * the tree under a heavy memory pressure. 122 * 123 * 124 * ========================================================================== 125 * 3. Performance analysis 126 * 127 * -- overhead 128 * 1. There is a cache extent for write access, so if writes are 129 * not very random, adding space operaions are in O(1) time. 130 * 131 * -- gain 132 * 2. Code is much simpler, more readable, more maintainable and 133 * more efficient. 134 * 135 * 136 * ========================================================================== 137 * 4. TODO list 138 * 139 * -- Refactor delayed space reservation 140 * 141 * -- Extent-level locking 142 */ 143 144 static struct kmem_cache *ext4_es_cachep; 145 static struct kmem_cache *ext4_pending_cachep; 146 147 static int __es_insert_extent(struct inode *inode, struct extent_status *newes); 148 static int __es_remove_extent(struct inode *inode, ext4_lblk_t lblk, 149 ext4_lblk_t end, int *reserved); 150 static int es_reclaim_extents(struct ext4_inode_info *ei, int *nr_to_scan); 151 static int __es_shrink(struct ext4_sb_info *sbi, int nr_to_scan, 152 struct ext4_inode_info *locked_ei); 153 static void __revise_pending(struct inode *inode, ext4_lblk_t lblk, 154 ext4_lblk_t len); 155 156 int __init ext4_init_es(void) 157 { 158 ext4_es_cachep = kmem_cache_create("ext4_extent_status", 159 sizeof(struct extent_status), 160 0, (SLAB_RECLAIM_ACCOUNT), NULL); 161 if (ext4_es_cachep == NULL) 162 return -ENOMEM; 163 return 0; 164 } 165 166 void ext4_exit_es(void) 167 { 168 kmem_cache_destroy(ext4_es_cachep); 169 } 170 171 void ext4_es_init_tree(struct ext4_es_tree *tree) 172 { 173 tree->root = RB_ROOT; 174 tree->cache_es = NULL; 175 } 176 177 #ifdef ES_DEBUG__ 178 static void ext4_es_print_tree(struct inode *inode) 179 { 180 struct ext4_es_tree *tree; 181 struct rb_node *node; 182 183 printk(KERN_DEBUG "status extents for inode %lu:", inode->i_ino); 184 tree = &EXT4_I(inode)->i_es_tree; 185 node = rb_first(&tree->root); 186 while (node) { 187 struct extent_status *es; 188 es = rb_entry(node, struct extent_status, rb_node); 189 printk(KERN_DEBUG " [%u/%u) %llu %x", 190 es->es_lblk, es->es_len, 191 ext4_es_pblock(es), ext4_es_status(es)); 192 node = rb_next(node); 193 } 194 printk(KERN_DEBUG "\n"); 195 } 196 #else 197 #define ext4_es_print_tree(inode) 198 #endif 199 200 static inline ext4_lblk_t ext4_es_end(struct extent_status *es) 201 { 202 BUG_ON(es->es_lblk + es->es_len < es->es_lblk); 203 return es->es_lblk + es->es_len - 1; 204 } 205 206 /* 207 * search through the tree for an delayed extent with a given offset. If 208 * it can't be found, try to find next extent. 209 */ 210 static struct extent_status *__es_tree_search(struct rb_root *root, 211 ext4_lblk_t lblk) 212 { 213 struct rb_node *node = root->rb_node; 214 struct extent_status *es = NULL; 215 216 while (node) { 217 es = rb_entry(node, struct extent_status, rb_node); 218 if (lblk < es->es_lblk) 219 node = node->rb_left; 220 else if (lblk > ext4_es_end(es)) 221 node = node->rb_right; 222 else 223 return es; 224 } 225 226 if (es && lblk < es->es_lblk) 227 return es; 228 229 if (es && lblk > ext4_es_end(es)) { 230 node = rb_next(&es->rb_node); 231 return node ? rb_entry(node, struct extent_status, rb_node) : 232 NULL; 233 } 234 235 return NULL; 236 } 237 238 /* 239 * ext4_es_find_extent_range - find extent with specified status within block 240 * range or next extent following block range in 241 * extents status tree 242 * 243 * @inode - file containing the range 244 * @matching_fn - pointer to function that matches extents with desired status 245 * @lblk - logical block defining start of range 246 * @end - logical block defining end of range 247 * @es - extent found, if any 248 * 249 * Find the first extent within the block range specified by @lblk and @end 250 * in the extents status tree that satisfies @matching_fn. If a match 251 * is found, it's returned in @es. If not, and a matching extent is found 252 * beyond the block range, it's returned in @es. If no match is found, an 253 * extent is returned in @es whose es_lblk, es_len, and es_pblk components 254 * are 0. 255 */ 256 static void __es_find_extent_range(struct inode *inode, 257 int (*matching_fn)(struct extent_status *es), 258 ext4_lblk_t lblk, ext4_lblk_t end, 259 struct extent_status *es) 260 { 261 struct ext4_es_tree *tree = NULL; 262 struct extent_status *es1 = NULL; 263 struct rb_node *node; 264 265 WARN_ON(es == NULL); 266 WARN_ON(end < lblk); 267 268 tree = &EXT4_I(inode)->i_es_tree; 269 270 /* see if the extent has been cached */ 271 es->es_lblk = es->es_len = es->es_pblk = 0; 272 if (tree->cache_es) { 273 es1 = tree->cache_es; 274 if (in_range(lblk, es1->es_lblk, es1->es_len)) { 275 es_debug("%u cached by [%u/%u) %llu %x\n", 276 lblk, es1->es_lblk, es1->es_len, 277 ext4_es_pblock(es1), ext4_es_status(es1)); 278 goto out; 279 } 280 } 281 282 es1 = __es_tree_search(&tree->root, lblk); 283 284 out: 285 if (es1 && !matching_fn(es1)) { 286 while ((node = rb_next(&es1->rb_node)) != NULL) { 287 es1 = rb_entry(node, struct extent_status, rb_node); 288 if (es1->es_lblk > end) { 289 es1 = NULL; 290 break; 291 } 292 if (matching_fn(es1)) 293 break; 294 } 295 } 296 297 if (es1 && matching_fn(es1)) { 298 tree->cache_es = es1; 299 es->es_lblk = es1->es_lblk; 300 es->es_len = es1->es_len; 301 es->es_pblk = es1->es_pblk; 302 } 303 304 } 305 306 /* 307 * Locking for __es_find_extent_range() for external use 308 */ 309 void ext4_es_find_extent_range(struct inode *inode, 310 int (*matching_fn)(struct extent_status *es), 311 ext4_lblk_t lblk, ext4_lblk_t end, 312 struct extent_status *es) 313 { 314 trace_ext4_es_find_extent_range_enter(inode, lblk); 315 316 read_lock(&EXT4_I(inode)->i_es_lock); 317 __es_find_extent_range(inode, matching_fn, lblk, end, es); 318 read_unlock(&EXT4_I(inode)->i_es_lock); 319 320 trace_ext4_es_find_extent_range_exit(inode, es); 321 } 322 323 /* 324 * __es_scan_range - search block range for block with specified status 325 * in extents status tree 326 * 327 * @inode - file containing the range 328 * @matching_fn - pointer to function that matches extents with desired status 329 * @lblk - logical block defining start of range 330 * @end - logical block defining end of range 331 * 332 * Returns true if at least one block in the specified block range satisfies 333 * the criterion specified by @matching_fn, and false if not. If at least 334 * one extent has the specified status, then there is at least one block 335 * in the cluster with that status. Should only be called by code that has 336 * taken i_es_lock. 337 */ 338 static bool __es_scan_range(struct inode *inode, 339 int (*matching_fn)(struct extent_status *es), 340 ext4_lblk_t start, ext4_lblk_t end) 341 { 342 struct extent_status es; 343 344 __es_find_extent_range(inode, matching_fn, start, end, &es); 345 if (es.es_len == 0) 346 return false; /* no matching extent in the tree */ 347 else if (es.es_lblk <= start && 348 start < es.es_lblk + es.es_len) 349 return true; 350 else if (start <= es.es_lblk && es.es_lblk <= end) 351 return true; 352 else 353 return false; 354 } 355 /* 356 * Locking for __es_scan_range() for external use 357 */ 358 bool ext4_es_scan_range(struct inode *inode, 359 int (*matching_fn)(struct extent_status *es), 360 ext4_lblk_t lblk, ext4_lblk_t end) 361 { 362 bool ret; 363 364 read_lock(&EXT4_I(inode)->i_es_lock); 365 ret = __es_scan_range(inode, matching_fn, lblk, end); 366 read_unlock(&EXT4_I(inode)->i_es_lock); 367 368 return ret; 369 } 370 371 /* 372 * __es_scan_clu - search cluster for block with specified status in 373 * extents status tree 374 * 375 * @inode - file containing the cluster 376 * @matching_fn - pointer to function that matches extents with desired status 377 * @lblk - logical block in cluster to be searched 378 * 379 * Returns true if at least one extent in the cluster containing @lblk 380 * satisfies the criterion specified by @matching_fn, and false if not. If at 381 * least one extent has the specified status, then there is at least one block 382 * in the cluster with that status. Should only be called by code that has 383 * taken i_es_lock. 384 */ 385 static bool __es_scan_clu(struct inode *inode, 386 int (*matching_fn)(struct extent_status *es), 387 ext4_lblk_t lblk) 388 { 389 struct ext4_sb_info *sbi = EXT4_SB(inode->i_sb); 390 ext4_lblk_t lblk_start, lblk_end; 391 392 lblk_start = EXT4_LBLK_CMASK(sbi, lblk); 393 lblk_end = lblk_start + sbi->s_cluster_ratio - 1; 394 395 return __es_scan_range(inode, matching_fn, lblk_start, lblk_end); 396 } 397 398 /* 399 * Locking for __es_scan_clu() for external use 400 */ 401 bool ext4_es_scan_clu(struct inode *inode, 402 int (*matching_fn)(struct extent_status *es), 403 ext4_lblk_t lblk) 404 { 405 bool ret; 406 407 read_lock(&EXT4_I(inode)->i_es_lock); 408 ret = __es_scan_clu(inode, matching_fn, lblk); 409 read_unlock(&EXT4_I(inode)->i_es_lock); 410 411 return ret; 412 } 413 414 static void ext4_es_list_add(struct inode *inode) 415 { 416 struct ext4_inode_info *ei = EXT4_I(inode); 417 struct ext4_sb_info *sbi = EXT4_SB(inode->i_sb); 418 419 if (!list_empty(&ei->i_es_list)) 420 return; 421 422 spin_lock(&sbi->s_es_lock); 423 if (list_empty(&ei->i_es_list)) { 424 list_add_tail(&ei->i_es_list, &sbi->s_es_list); 425 sbi->s_es_nr_inode++; 426 } 427 spin_unlock(&sbi->s_es_lock); 428 } 429 430 static void ext4_es_list_del(struct inode *inode) 431 { 432 struct ext4_inode_info *ei = EXT4_I(inode); 433 struct ext4_sb_info *sbi = EXT4_SB(inode->i_sb); 434 435 spin_lock(&sbi->s_es_lock); 436 if (!list_empty(&ei->i_es_list)) { 437 list_del_init(&ei->i_es_list); 438 sbi->s_es_nr_inode--; 439 WARN_ON_ONCE(sbi->s_es_nr_inode < 0); 440 } 441 spin_unlock(&sbi->s_es_lock); 442 } 443 444 static struct extent_status * 445 ext4_es_alloc_extent(struct inode *inode, ext4_lblk_t lblk, ext4_lblk_t len, 446 ext4_fsblk_t pblk) 447 { 448 struct extent_status *es; 449 es = kmem_cache_alloc(ext4_es_cachep, GFP_ATOMIC); 450 if (es == NULL) 451 return NULL; 452 es->es_lblk = lblk; 453 es->es_len = len; 454 es->es_pblk = pblk; 455 456 /* 457 * We don't count delayed extent because we never try to reclaim them 458 */ 459 if (!ext4_es_is_delayed(es)) { 460 if (!EXT4_I(inode)->i_es_shk_nr++) 461 ext4_es_list_add(inode); 462 percpu_counter_inc(&EXT4_SB(inode->i_sb)-> 463 s_es_stats.es_stats_shk_cnt); 464 } 465 466 EXT4_I(inode)->i_es_all_nr++; 467 percpu_counter_inc(&EXT4_SB(inode->i_sb)->s_es_stats.es_stats_all_cnt); 468 469 return es; 470 } 471 472 static void ext4_es_free_extent(struct inode *inode, struct extent_status *es) 473 { 474 EXT4_I(inode)->i_es_all_nr--; 475 percpu_counter_dec(&EXT4_SB(inode->i_sb)->s_es_stats.es_stats_all_cnt); 476 477 /* Decrease the shrink counter when this es is not delayed */ 478 if (!ext4_es_is_delayed(es)) { 479 BUG_ON(EXT4_I(inode)->i_es_shk_nr == 0); 480 if (!--EXT4_I(inode)->i_es_shk_nr) 481 ext4_es_list_del(inode); 482 percpu_counter_dec(&EXT4_SB(inode->i_sb)-> 483 s_es_stats.es_stats_shk_cnt); 484 } 485 486 kmem_cache_free(ext4_es_cachep, es); 487 } 488 489 /* 490 * Check whether or not two extents can be merged 491 * Condition: 492 * - logical block number is contiguous 493 * - physical block number is contiguous 494 * - status is equal 495 */ 496 static int ext4_es_can_be_merged(struct extent_status *es1, 497 struct extent_status *es2) 498 { 499 if (ext4_es_type(es1) != ext4_es_type(es2)) 500 return 0; 501 502 if (((__u64) es1->es_len) + es2->es_len > EXT_MAX_BLOCKS) { 503 pr_warn("ES assertion failed when merging extents. " 504 "The sum of lengths of es1 (%d) and es2 (%d) " 505 "is bigger than allowed file size (%d)\n", 506 es1->es_len, es2->es_len, EXT_MAX_BLOCKS); 507 WARN_ON(1); 508 return 0; 509 } 510 511 if (((__u64) es1->es_lblk) + es1->es_len != es2->es_lblk) 512 return 0; 513 514 if ((ext4_es_is_written(es1) || ext4_es_is_unwritten(es1)) && 515 (ext4_es_pblock(es1) + es1->es_len == ext4_es_pblock(es2))) 516 return 1; 517 518 if (ext4_es_is_hole(es1)) 519 return 1; 520 521 /* we need to check delayed extent is without unwritten status */ 522 if (ext4_es_is_delayed(es1) && !ext4_es_is_unwritten(es1)) 523 return 1; 524 525 return 0; 526 } 527 528 static struct extent_status * 529 ext4_es_try_to_merge_left(struct inode *inode, struct extent_status *es) 530 { 531 struct ext4_es_tree *tree = &EXT4_I(inode)->i_es_tree; 532 struct extent_status *es1; 533 struct rb_node *node; 534 535 node = rb_prev(&es->rb_node); 536 if (!node) 537 return es; 538 539 es1 = rb_entry(node, struct extent_status, rb_node); 540 if (ext4_es_can_be_merged(es1, es)) { 541 es1->es_len += es->es_len; 542 if (ext4_es_is_referenced(es)) 543 ext4_es_set_referenced(es1); 544 rb_erase(&es->rb_node, &tree->root); 545 ext4_es_free_extent(inode, es); 546 es = es1; 547 } 548 549 return es; 550 } 551 552 static struct extent_status * 553 ext4_es_try_to_merge_right(struct inode *inode, struct extent_status *es) 554 { 555 struct ext4_es_tree *tree = &EXT4_I(inode)->i_es_tree; 556 struct extent_status *es1; 557 struct rb_node *node; 558 559 node = rb_next(&es->rb_node); 560 if (!node) 561 return es; 562 563 es1 = rb_entry(node, struct extent_status, rb_node); 564 if (ext4_es_can_be_merged(es, es1)) { 565 es->es_len += es1->es_len; 566 if (ext4_es_is_referenced(es1)) 567 ext4_es_set_referenced(es); 568 rb_erase(node, &tree->root); 569 ext4_es_free_extent(inode, es1); 570 } 571 572 return es; 573 } 574 575 #ifdef ES_AGGRESSIVE_TEST 576 #include "ext4_extents.h" /* Needed when ES_AGGRESSIVE_TEST is defined */ 577 578 static void ext4_es_insert_extent_ext_check(struct inode *inode, 579 struct extent_status *es) 580 { 581 struct ext4_ext_path *path = NULL; 582 struct ext4_extent *ex; 583 ext4_lblk_t ee_block; 584 ext4_fsblk_t ee_start; 585 unsigned short ee_len; 586 int depth, ee_status, es_status; 587 588 path = ext4_find_extent(inode, es->es_lblk, NULL, EXT4_EX_NOCACHE); 589 if (IS_ERR(path)) 590 return; 591 592 depth = ext_depth(inode); 593 ex = path[depth].p_ext; 594 595 if (ex) { 596 597 ee_block = le32_to_cpu(ex->ee_block); 598 ee_start = ext4_ext_pblock(ex); 599 ee_len = ext4_ext_get_actual_len(ex); 600 601 ee_status = ext4_ext_is_unwritten(ex) ? 1 : 0; 602 es_status = ext4_es_is_unwritten(es) ? 1 : 0; 603 604 /* 605 * Make sure ex and es are not overlap when we try to insert 606 * a delayed/hole extent. 607 */ 608 if (!ext4_es_is_written(es) && !ext4_es_is_unwritten(es)) { 609 if (in_range(es->es_lblk, ee_block, ee_len)) { 610 pr_warn("ES insert assertion failed for " 611 "inode: %lu we can find an extent " 612 "at block [%d/%d/%llu/%c], but we " 613 "want to add a delayed/hole extent " 614 "[%d/%d/%llu/%x]\n", 615 inode->i_ino, ee_block, ee_len, 616 ee_start, ee_status ? 'u' : 'w', 617 es->es_lblk, es->es_len, 618 ext4_es_pblock(es), ext4_es_status(es)); 619 } 620 goto out; 621 } 622 623 /* 624 * We don't check ee_block == es->es_lblk, etc. because es 625 * might be a part of whole extent, vice versa. 626 */ 627 if (es->es_lblk < ee_block || 628 ext4_es_pblock(es) != ee_start + es->es_lblk - ee_block) { 629 pr_warn("ES insert assertion failed for inode: %lu " 630 "ex_status [%d/%d/%llu/%c] != " 631 "es_status [%d/%d/%llu/%c]\n", inode->i_ino, 632 ee_block, ee_len, ee_start, 633 ee_status ? 'u' : 'w', es->es_lblk, es->es_len, 634 ext4_es_pblock(es), es_status ? 'u' : 'w'); 635 goto out; 636 } 637 638 if (ee_status ^ es_status) { 639 pr_warn("ES insert assertion failed for inode: %lu " 640 "ex_status [%d/%d/%llu/%c] != " 641 "es_status [%d/%d/%llu/%c]\n", inode->i_ino, 642 ee_block, ee_len, ee_start, 643 ee_status ? 'u' : 'w', es->es_lblk, es->es_len, 644 ext4_es_pblock(es), es_status ? 'u' : 'w'); 645 } 646 } else { 647 /* 648 * We can't find an extent on disk. So we need to make sure 649 * that we don't want to add an written/unwritten extent. 650 */ 651 if (!ext4_es_is_delayed(es) && !ext4_es_is_hole(es)) { 652 pr_warn("ES insert assertion failed for inode: %lu " 653 "can't find an extent at block %d but we want " 654 "to add a written/unwritten extent " 655 "[%d/%d/%llu/%x]\n", inode->i_ino, 656 es->es_lblk, es->es_lblk, es->es_len, 657 ext4_es_pblock(es), ext4_es_status(es)); 658 } 659 } 660 out: 661 ext4_ext_drop_refs(path); 662 kfree(path); 663 } 664 665 static void ext4_es_insert_extent_ind_check(struct inode *inode, 666 struct extent_status *es) 667 { 668 struct ext4_map_blocks map; 669 int retval; 670 671 /* 672 * Here we call ext4_ind_map_blocks to lookup a block mapping because 673 * 'Indirect' structure is defined in indirect.c. So we couldn't 674 * access direct/indirect tree from outside. It is too dirty to define 675 * this function in indirect.c file. 676 */ 677 678 map.m_lblk = es->es_lblk; 679 map.m_len = es->es_len; 680 681 retval = ext4_ind_map_blocks(NULL, inode, &map, 0); 682 if (retval > 0) { 683 if (ext4_es_is_delayed(es) || ext4_es_is_hole(es)) { 684 /* 685 * We want to add a delayed/hole extent but this 686 * block has been allocated. 687 */ 688 pr_warn("ES insert assertion failed for inode: %lu " 689 "We can find blocks but we want to add a " 690 "delayed/hole extent [%d/%d/%llu/%x]\n", 691 inode->i_ino, es->es_lblk, es->es_len, 692 ext4_es_pblock(es), ext4_es_status(es)); 693 return; 694 } else if (ext4_es_is_written(es)) { 695 if (retval != es->es_len) { 696 pr_warn("ES insert assertion failed for " 697 "inode: %lu retval %d != es_len %d\n", 698 inode->i_ino, retval, es->es_len); 699 return; 700 } 701 if (map.m_pblk != ext4_es_pblock(es)) { 702 pr_warn("ES insert assertion failed for " 703 "inode: %lu m_pblk %llu != " 704 "es_pblk %llu\n", 705 inode->i_ino, map.m_pblk, 706 ext4_es_pblock(es)); 707 return; 708 } 709 } else { 710 /* 711 * We don't need to check unwritten extent because 712 * indirect-based file doesn't have it. 713 */ 714 BUG(); 715 } 716 } else if (retval == 0) { 717 if (ext4_es_is_written(es)) { 718 pr_warn("ES insert assertion failed for inode: %lu " 719 "We can't find the block but we want to add " 720 "a written extent [%d/%d/%llu/%x]\n", 721 inode->i_ino, es->es_lblk, es->es_len, 722 ext4_es_pblock(es), ext4_es_status(es)); 723 return; 724 } 725 } 726 } 727 728 static inline void ext4_es_insert_extent_check(struct inode *inode, 729 struct extent_status *es) 730 { 731 /* 732 * We don't need to worry about the race condition because 733 * caller takes i_data_sem locking. 734 */ 735 BUG_ON(!rwsem_is_locked(&EXT4_I(inode)->i_data_sem)); 736 if (ext4_test_inode_flag(inode, EXT4_INODE_EXTENTS)) 737 ext4_es_insert_extent_ext_check(inode, es); 738 else 739 ext4_es_insert_extent_ind_check(inode, es); 740 } 741 #else 742 static inline void ext4_es_insert_extent_check(struct inode *inode, 743 struct extent_status *es) 744 { 745 } 746 #endif 747 748 static int __es_insert_extent(struct inode *inode, struct extent_status *newes) 749 { 750 struct ext4_es_tree *tree = &EXT4_I(inode)->i_es_tree; 751 struct rb_node **p = &tree->root.rb_node; 752 struct rb_node *parent = NULL; 753 struct extent_status *es; 754 755 while (*p) { 756 parent = *p; 757 es = rb_entry(parent, struct extent_status, rb_node); 758 759 if (newes->es_lblk < es->es_lblk) { 760 if (ext4_es_can_be_merged(newes, es)) { 761 /* 762 * Here we can modify es_lblk directly 763 * because it isn't overlapped. 764 */ 765 es->es_lblk = newes->es_lblk; 766 es->es_len += newes->es_len; 767 if (ext4_es_is_written(es) || 768 ext4_es_is_unwritten(es)) 769 ext4_es_store_pblock(es, 770 newes->es_pblk); 771 es = ext4_es_try_to_merge_left(inode, es); 772 goto out; 773 } 774 p = &(*p)->rb_left; 775 } else if (newes->es_lblk > ext4_es_end(es)) { 776 if (ext4_es_can_be_merged(es, newes)) { 777 es->es_len += newes->es_len; 778 es = ext4_es_try_to_merge_right(inode, es); 779 goto out; 780 } 781 p = &(*p)->rb_right; 782 } else { 783 BUG(); 784 return -EINVAL; 785 } 786 } 787 788 es = ext4_es_alloc_extent(inode, newes->es_lblk, newes->es_len, 789 newes->es_pblk); 790 if (!es) 791 return -ENOMEM; 792 rb_link_node(&es->rb_node, parent, p); 793 rb_insert_color(&es->rb_node, &tree->root); 794 795 out: 796 tree->cache_es = es; 797 return 0; 798 } 799 800 /* 801 * ext4_es_insert_extent() adds information to an inode's extent 802 * status tree. 803 * 804 * Return 0 on success, error code on failure. 805 */ 806 int ext4_es_insert_extent(struct inode *inode, ext4_lblk_t lblk, 807 ext4_lblk_t len, ext4_fsblk_t pblk, 808 unsigned int status) 809 { 810 struct extent_status newes; 811 ext4_lblk_t end = lblk + len - 1; 812 int err = 0; 813 struct ext4_sb_info *sbi = EXT4_SB(inode->i_sb); 814 815 es_debug("add [%u/%u) %llu %x to extent status tree of inode %lu\n", 816 lblk, len, pblk, status, inode->i_ino); 817 818 if (!len) 819 return 0; 820 821 BUG_ON(end < lblk); 822 823 if ((status & EXTENT_STATUS_DELAYED) && 824 (status & EXTENT_STATUS_WRITTEN)) { 825 ext4_warning(inode->i_sb, "Inserting extent [%u/%u] as " 826 " delayed and written which can potentially " 827 " cause data loss.", lblk, len); 828 WARN_ON(1); 829 } 830 831 newes.es_lblk = lblk; 832 newes.es_len = len; 833 ext4_es_store_pblock_status(&newes, pblk, status); 834 trace_ext4_es_insert_extent(inode, &newes); 835 836 ext4_es_insert_extent_check(inode, &newes); 837 838 write_lock(&EXT4_I(inode)->i_es_lock); 839 err = __es_remove_extent(inode, lblk, end, NULL); 840 if (err != 0) 841 goto error; 842 retry: 843 err = __es_insert_extent(inode, &newes); 844 if (err == -ENOMEM && __es_shrink(EXT4_SB(inode->i_sb), 845 128, EXT4_I(inode))) 846 goto retry; 847 if (err == -ENOMEM && !ext4_es_is_delayed(&newes)) 848 err = 0; 849 850 if (sbi->s_cluster_ratio > 1 && test_opt(inode->i_sb, DELALLOC) && 851 (status & EXTENT_STATUS_WRITTEN || 852 status & EXTENT_STATUS_UNWRITTEN)) 853 __revise_pending(inode, lblk, len); 854 855 error: 856 write_unlock(&EXT4_I(inode)->i_es_lock); 857 858 ext4_es_print_tree(inode); 859 860 return err; 861 } 862 863 /* 864 * ext4_es_cache_extent() inserts information into the extent status 865 * tree if and only if there isn't information about the range in 866 * question already. 867 */ 868 void ext4_es_cache_extent(struct inode *inode, ext4_lblk_t lblk, 869 ext4_lblk_t len, ext4_fsblk_t pblk, 870 unsigned int status) 871 { 872 struct extent_status *es; 873 struct extent_status newes; 874 ext4_lblk_t end = lblk + len - 1; 875 876 newes.es_lblk = lblk; 877 newes.es_len = len; 878 ext4_es_store_pblock_status(&newes, pblk, status); 879 trace_ext4_es_cache_extent(inode, &newes); 880 881 if (!len) 882 return; 883 884 BUG_ON(end < lblk); 885 886 write_lock(&EXT4_I(inode)->i_es_lock); 887 888 es = __es_tree_search(&EXT4_I(inode)->i_es_tree.root, lblk); 889 if (!es || es->es_lblk > end) 890 __es_insert_extent(inode, &newes); 891 write_unlock(&EXT4_I(inode)->i_es_lock); 892 } 893 894 /* 895 * ext4_es_lookup_extent() looks up an extent in extent status tree. 896 * 897 * ext4_es_lookup_extent is called by ext4_map_blocks/ext4_da_map_blocks. 898 * 899 * Return: 1 on found, 0 on not 900 */ 901 int ext4_es_lookup_extent(struct inode *inode, ext4_lblk_t lblk, 902 ext4_lblk_t *next_lblk, 903 struct extent_status *es) 904 { 905 struct ext4_es_tree *tree; 906 struct ext4_es_stats *stats; 907 struct extent_status *es1 = NULL; 908 struct rb_node *node; 909 int found = 0; 910 911 trace_ext4_es_lookup_extent_enter(inode, lblk); 912 es_debug("lookup extent in block %u\n", lblk); 913 914 tree = &EXT4_I(inode)->i_es_tree; 915 read_lock(&EXT4_I(inode)->i_es_lock); 916 917 /* find extent in cache firstly */ 918 es->es_lblk = es->es_len = es->es_pblk = 0; 919 if (tree->cache_es) { 920 es1 = tree->cache_es; 921 if (in_range(lblk, es1->es_lblk, es1->es_len)) { 922 es_debug("%u cached by [%u/%u)\n", 923 lblk, es1->es_lblk, es1->es_len); 924 found = 1; 925 goto out; 926 } 927 } 928 929 node = tree->root.rb_node; 930 while (node) { 931 es1 = rb_entry(node, struct extent_status, rb_node); 932 if (lblk < es1->es_lblk) 933 node = node->rb_left; 934 else if (lblk > ext4_es_end(es1)) 935 node = node->rb_right; 936 else { 937 found = 1; 938 break; 939 } 940 } 941 942 out: 943 stats = &EXT4_SB(inode->i_sb)->s_es_stats; 944 if (found) { 945 BUG_ON(!es1); 946 es->es_lblk = es1->es_lblk; 947 es->es_len = es1->es_len; 948 es->es_pblk = es1->es_pblk; 949 if (!ext4_es_is_referenced(es1)) 950 ext4_es_set_referenced(es1); 951 percpu_counter_inc(&stats->es_stats_cache_hits); 952 if (next_lblk) { 953 node = rb_next(&es1->rb_node); 954 if (node) { 955 es1 = rb_entry(node, struct extent_status, 956 rb_node); 957 *next_lblk = es1->es_lblk; 958 } else 959 *next_lblk = 0; 960 } 961 } else { 962 percpu_counter_inc(&stats->es_stats_cache_misses); 963 } 964 965 read_unlock(&EXT4_I(inode)->i_es_lock); 966 967 trace_ext4_es_lookup_extent_exit(inode, es, found); 968 return found; 969 } 970 971 struct rsvd_count { 972 int ndelonly; 973 bool first_do_lblk_found; 974 ext4_lblk_t first_do_lblk; 975 ext4_lblk_t last_do_lblk; 976 struct extent_status *left_es; 977 bool partial; 978 ext4_lblk_t lclu; 979 }; 980 981 /* 982 * init_rsvd - initialize reserved count data before removing block range 983 * in file from extent status tree 984 * 985 * @inode - file containing range 986 * @lblk - first block in range 987 * @es - pointer to first extent in range 988 * @rc - pointer to reserved count data 989 * 990 * Assumes es is not NULL 991 */ 992 static void init_rsvd(struct inode *inode, ext4_lblk_t lblk, 993 struct extent_status *es, struct rsvd_count *rc) 994 { 995 struct ext4_sb_info *sbi = EXT4_SB(inode->i_sb); 996 struct rb_node *node; 997 998 rc->ndelonly = 0; 999 1000 /* 1001 * for bigalloc, note the first delonly block in the range has not 1002 * been found, record the extent containing the block to the left of 1003 * the region to be removed, if any, and note that there's no partial 1004 * cluster to track 1005 */ 1006 if (sbi->s_cluster_ratio > 1) { 1007 rc->first_do_lblk_found = false; 1008 if (lblk > es->es_lblk) { 1009 rc->left_es = es; 1010 } else { 1011 node = rb_prev(&es->rb_node); 1012 rc->left_es = node ? rb_entry(node, 1013 struct extent_status, 1014 rb_node) : NULL; 1015 } 1016 rc->partial = false; 1017 } 1018 } 1019 1020 /* 1021 * count_rsvd - count the clusters containing delayed and not unwritten 1022 * (delonly) blocks in a range within an extent and add to 1023 * the running tally in rsvd_count 1024 * 1025 * @inode - file containing extent 1026 * @lblk - first block in range 1027 * @len - length of range in blocks 1028 * @es - pointer to extent containing clusters to be counted 1029 * @rc - pointer to reserved count data 1030 * 1031 * Tracks partial clusters found at the beginning and end of extents so 1032 * they aren't overcounted when they span adjacent extents 1033 */ 1034 static void count_rsvd(struct inode *inode, ext4_lblk_t lblk, long len, 1035 struct extent_status *es, struct rsvd_count *rc) 1036 { 1037 struct ext4_sb_info *sbi = EXT4_SB(inode->i_sb); 1038 ext4_lblk_t i, end, nclu; 1039 1040 if (!ext4_es_is_delonly(es)) 1041 return; 1042 1043 WARN_ON(len <= 0); 1044 1045 if (sbi->s_cluster_ratio == 1) { 1046 rc->ndelonly += (int) len; 1047 return; 1048 } 1049 1050 /* bigalloc */ 1051 1052 i = (lblk < es->es_lblk) ? es->es_lblk : lblk; 1053 end = lblk + (ext4_lblk_t) len - 1; 1054 end = (end > ext4_es_end(es)) ? ext4_es_end(es) : end; 1055 1056 /* record the first block of the first delonly extent seen */ 1057 if (rc->first_do_lblk_found == false) { 1058 rc->first_do_lblk = i; 1059 rc->first_do_lblk_found = true; 1060 } 1061 1062 /* update the last lblk in the region seen so far */ 1063 rc->last_do_lblk = end; 1064 1065 /* 1066 * if we're tracking a partial cluster and the current extent 1067 * doesn't start with it, count it and stop tracking 1068 */ 1069 if (rc->partial && (rc->lclu != EXT4_B2C(sbi, i))) { 1070 rc->ndelonly++; 1071 rc->partial = false; 1072 } 1073 1074 /* 1075 * if the first cluster doesn't start on a cluster boundary but 1076 * ends on one, count it 1077 */ 1078 if (EXT4_LBLK_COFF(sbi, i) != 0) { 1079 if (end >= EXT4_LBLK_CFILL(sbi, i)) { 1080 rc->ndelonly++; 1081 rc->partial = false; 1082 i = EXT4_LBLK_CFILL(sbi, i) + 1; 1083 } 1084 } 1085 1086 /* 1087 * if the current cluster starts on a cluster boundary, count the 1088 * number of whole delonly clusters in the extent 1089 */ 1090 if ((i + sbi->s_cluster_ratio - 1) <= end) { 1091 nclu = (end - i + 1) >> sbi->s_cluster_bits; 1092 rc->ndelonly += nclu; 1093 i += nclu << sbi->s_cluster_bits; 1094 } 1095 1096 /* 1097 * start tracking a partial cluster if there's a partial at the end 1098 * of the current extent and we're not already tracking one 1099 */ 1100 if (!rc->partial && i <= end) { 1101 rc->partial = true; 1102 rc->lclu = EXT4_B2C(sbi, i); 1103 } 1104 } 1105 1106 /* 1107 * __pr_tree_search - search for a pending cluster reservation 1108 * 1109 * @root - root of pending reservation tree 1110 * @lclu - logical cluster to search for 1111 * 1112 * Returns the pending reservation for the cluster identified by @lclu 1113 * if found. If not, returns a reservation for the next cluster if any, 1114 * and if not, returns NULL. 1115 */ 1116 static struct pending_reservation *__pr_tree_search(struct rb_root *root, 1117 ext4_lblk_t lclu) 1118 { 1119 struct rb_node *node = root->rb_node; 1120 struct pending_reservation *pr = NULL; 1121 1122 while (node) { 1123 pr = rb_entry(node, struct pending_reservation, rb_node); 1124 if (lclu < pr->lclu) 1125 node = node->rb_left; 1126 else if (lclu > pr->lclu) 1127 node = node->rb_right; 1128 else 1129 return pr; 1130 } 1131 if (pr && lclu < pr->lclu) 1132 return pr; 1133 if (pr && lclu > pr->lclu) { 1134 node = rb_next(&pr->rb_node); 1135 return node ? rb_entry(node, struct pending_reservation, 1136 rb_node) : NULL; 1137 } 1138 return NULL; 1139 } 1140 1141 /* 1142 * get_rsvd - calculates and returns the number of cluster reservations to be 1143 * released when removing a block range from the extent status tree 1144 * and releases any pending reservations within the range 1145 * 1146 * @inode - file containing block range 1147 * @end - last block in range 1148 * @right_es - pointer to extent containing next block beyond end or NULL 1149 * @rc - pointer to reserved count data 1150 * 1151 * The number of reservations to be released is equal to the number of 1152 * clusters containing delayed and not unwritten (delonly) blocks within 1153 * the range, minus the number of clusters still containing delonly blocks 1154 * at the ends of the range, and minus the number of pending reservations 1155 * within the range. 1156 */ 1157 static unsigned int get_rsvd(struct inode *inode, ext4_lblk_t end, 1158 struct extent_status *right_es, 1159 struct rsvd_count *rc) 1160 { 1161 struct ext4_sb_info *sbi = EXT4_SB(inode->i_sb); 1162 struct pending_reservation *pr; 1163 struct ext4_pending_tree *tree = &EXT4_I(inode)->i_pending_tree; 1164 struct rb_node *node; 1165 ext4_lblk_t first_lclu, last_lclu; 1166 bool left_delonly, right_delonly, count_pending; 1167 struct extent_status *es; 1168 1169 if (sbi->s_cluster_ratio > 1) { 1170 /* count any remaining partial cluster */ 1171 if (rc->partial) 1172 rc->ndelonly++; 1173 1174 if (rc->ndelonly == 0) 1175 return 0; 1176 1177 first_lclu = EXT4_B2C(sbi, rc->first_do_lblk); 1178 last_lclu = EXT4_B2C(sbi, rc->last_do_lblk); 1179 1180 /* 1181 * decrease the delonly count by the number of clusters at the 1182 * ends of the range that still contain delonly blocks - 1183 * these clusters still need to be reserved 1184 */ 1185 left_delonly = right_delonly = false; 1186 1187 es = rc->left_es; 1188 while (es && ext4_es_end(es) >= 1189 EXT4_LBLK_CMASK(sbi, rc->first_do_lblk)) { 1190 if (ext4_es_is_delonly(es)) { 1191 rc->ndelonly--; 1192 left_delonly = true; 1193 break; 1194 } 1195 node = rb_prev(&es->rb_node); 1196 if (!node) 1197 break; 1198 es = rb_entry(node, struct extent_status, rb_node); 1199 } 1200 if (right_es && (!left_delonly || first_lclu != last_lclu)) { 1201 if (end < ext4_es_end(right_es)) { 1202 es = right_es; 1203 } else { 1204 node = rb_next(&right_es->rb_node); 1205 es = node ? rb_entry(node, struct extent_status, 1206 rb_node) : NULL; 1207 } 1208 while (es && es->es_lblk <= 1209 EXT4_LBLK_CFILL(sbi, rc->last_do_lblk)) { 1210 if (ext4_es_is_delonly(es)) { 1211 rc->ndelonly--; 1212 right_delonly = true; 1213 break; 1214 } 1215 node = rb_next(&es->rb_node); 1216 if (!node) 1217 break; 1218 es = rb_entry(node, struct extent_status, 1219 rb_node); 1220 } 1221 } 1222 1223 /* 1224 * Determine the block range that should be searched for 1225 * pending reservations, if any. Clusters on the ends of the 1226 * original removed range containing delonly blocks are 1227 * excluded. They've already been accounted for and it's not 1228 * possible to determine if an associated pending reservation 1229 * should be released with the information available in the 1230 * extents status tree. 1231 */ 1232 if (first_lclu == last_lclu) { 1233 if (left_delonly | right_delonly) 1234 count_pending = false; 1235 else 1236 count_pending = true; 1237 } else { 1238 if (left_delonly) 1239 first_lclu++; 1240 if (right_delonly) 1241 last_lclu--; 1242 if (first_lclu <= last_lclu) 1243 count_pending = true; 1244 else 1245 count_pending = false; 1246 } 1247 1248 /* 1249 * a pending reservation found between first_lclu and last_lclu 1250 * represents an allocated cluster that contained at least one 1251 * delonly block, so the delonly total must be reduced by one 1252 * for each pending reservation found and released 1253 */ 1254 if (count_pending) { 1255 pr = __pr_tree_search(&tree->root, first_lclu); 1256 while (pr && pr->lclu <= last_lclu) { 1257 rc->ndelonly--; 1258 node = rb_next(&pr->rb_node); 1259 rb_erase(&pr->rb_node, &tree->root); 1260 kmem_cache_free(ext4_pending_cachep, pr); 1261 if (!node) 1262 break; 1263 pr = rb_entry(node, struct pending_reservation, 1264 rb_node); 1265 } 1266 } 1267 } 1268 return rc->ndelonly; 1269 } 1270 1271 1272 /* 1273 * __es_remove_extent - removes block range from extent status tree 1274 * 1275 * @inode - file containing range 1276 * @lblk - first block in range 1277 * @end - last block in range 1278 * @reserved - number of cluster reservations released 1279 * 1280 * If @reserved is not NULL and delayed allocation is enabled, counts 1281 * block/cluster reservations freed by removing range and if bigalloc 1282 * enabled cancels pending reservations as needed. Returns 0 on success, 1283 * error code on failure. 1284 */ 1285 static int __es_remove_extent(struct inode *inode, ext4_lblk_t lblk, 1286 ext4_lblk_t end, int *reserved) 1287 { 1288 struct ext4_es_tree *tree = &EXT4_I(inode)->i_es_tree; 1289 struct rb_node *node; 1290 struct extent_status *es; 1291 struct extent_status orig_es; 1292 ext4_lblk_t len1, len2; 1293 ext4_fsblk_t block; 1294 int err; 1295 bool count_reserved = true; 1296 struct rsvd_count rc; 1297 1298 if (reserved == NULL || !test_opt(inode->i_sb, DELALLOC)) 1299 count_reserved = false; 1300 retry: 1301 err = 0; 1302 1303 es = __es_tree_search(&tree->root, lblk); 1304 if (!es) 1305 goto out; 1306 if (es->es_lblk > end) 1307 goto out; 1308 1309 /* Simply invalidate cache_es. */ 1310 tree->cache_es = NULL; 1311 if (count_reserved) 1312 init_rsvd(inode, lblk, es, &rc); 1313 1314 orig_es.es_lblk = es->es_lblk; 1315 orig_es.es_len = es->es_len; 1316 orig_es.es_pblk = es->es_pblk; 1317 1318 len1 = lblk > es->es_lblk ? lblk - es->es_lblk : 0; 1319 len2 = ext4_es_end(es) > end ? ext4_es_end(es) - end : 0; 1320 if (len1 > 0) 1321 es->es_len = len1; 1322 if (len2 > 0) { 1323 if (len1 > 0) { 1324 struct extent_status newes; 1325 1326 newes.es_lblk = end + 1; 1327 newes.es_len = len2; 1328 block = 0x7FDEADBEEFULL; 1329 if (ext4_es_is_written(&orig_es) || 1330 ext4_es_is_unwritten(&orig_es)) 1331 block = ext4_es_pblock(&orig_es) + 1332 orig_es.es_len - len2; 1333 ext4_es_store_pblock_status(&newes, block, 1334 ext4_es_status(&orig_es)); 1335 err = __es_insert_extent(inode, &newes); 1336 if (err) { 1337 es->es_lblk = orig_es.es_lblk; 1338 es->es_len = orig_es.es_len; 1339 if ((err == -ENOMEM) && 1340 __es_shrink(EXT4_SB(inode->i_sb), 1341 128, EXT4_I(inode))) 1342 goto retry; 1343 goto out; 1344 } 1345 } else { 1346 es->es_lblk = end + 1; 1347 es->es_len = len2; 1348 if (ext4_es_is_written(es) || 1349 ext4_es_is_unwritten(es)) { 1350 block = orig_es.es_pblk + orig_es.es_len - len2; 1351 ext4_es_store_pblock(es, block); 1352 } 1353 } 1354 if (count_reserved) 1355 count_rsvd(inode, lblk, orig_es.es_len - len1 - len2, 1356 &orig_es, &rc); 1357 goto out; 1358 } 1359 1360 if (len1 > 0) { 1361 if (count_reserved) 1362 count_rsvd(inode, lblk, orig_es.es_len - len1, 1363 &orig_es, &rc); 1364 node = rb_next(&es->rb_node); 1365 if (node) 1366 es = rb_entry(node, struct extent_status, rb_node); 1367 else 1368 es = NULL; 1369 } 1370 1371 while (es && ext4_es_end(es) <= end) { 1372 if (count_reserved) 1373 count_rsvd(inode, es->es_lblk, es->es_len, es, &rc); 1374 node = rb_next(&es->rb_node); 1375 rb_erase(&es->rb_node, &tree->root); 1376 ext4_es_free_extent(inode, es); 1377 if (!node) { 1378 es = NULL; 1379 break; 1380 } 1381 es = rb_entry(node, struct extent_status, rb_node); 1382 } 1383 1384 if (es && es->es_lblk < end + 1) { 1385 ext4_lblk_t orig_len = es->es_len; 1386 1387 len1 = ext4_es_end(es) - end; 1388 if (count_reserved) 1389 count_rsvd(inode, es->es_lblk, orig_len - len1, 1390 es, &rc); 1391 es->es_lblk = end + 1; 1392 es->es_len = len1; 1393 if (ext4_es_is_written(es) || ext4_es_is_unwritten(es)) { 1394 block = es->es_pblk + orig_len - len1; 1395 ext4_es_store_pblock(es, block); 1396 } 1397 } 1398 1399 if (count_reserved) 1400 *reserved = get_rsvd(inode, end, es, &rc); 1401 out: 1402 return err; 1403 } 1404 1405 /* 1406 * ext4_es_remove_extent - removes block range from extent status tree 1407 * 1408 * @inode - file containing range 1409 * @lblk - first block in range 1410 * @len - number of blocks to remove 1411 * 1412 * Reduces block/cluster reservation count and for bigalloc cancels pending 1413 * reservations as needed. Returns 0 on success, error code on failure. 1414 */ 1415 int ext4_es_remove_extent(struct inode *inode, ext4_lblk_t lblk, 1416 ext4_lblk_t len) 1417 { 1418 ext4_lblk_t end; 1419 int err = 0; 1420 int reserved = 0; 1421 1422 trace_ext4_es_remove_extent(inode, lblk, len); 1423 es_debug("remove [%u/%u) from extent status tree of inode %lu\n", 1424 lblk, len, inode->i_ino); 1425 1426 if (!len) 1427 return err; 1428 1429 end = lblk + len - 1; 1430 BUG_ON(end < lblk); 1431 1432 /* 1433 * ext4_clear_inode() depends on us taking i_es_lock unconditionally 1434 * so that we are sure __es_shrink() is done with the inode before it 1435 * is reclaimed. 1436 */ 1437 write_lock(&EXT4_I(inode)->i_es_lock); 1438 err = __es_remove_extent(inode, lblk, end, &reserved); 1439 write_unlock(&EXT4_I(inode)->i_es_lock); 1440 ext4_es_print_tree(inode); 1441 ext4_da_release_space(inode, reserved); 1442 return err; 1443 } 1444 1445 static int __es_shrink(struct ext4_sb_info *sbi, int nr_to_scan, 1446 struct ext4_inode_info *locked_ei) 1447 { 1448 struct ext4_inode_info *ei; 1449 struct ext4_es_stats *es_stats; 1450 ktime_t start_time; 1451 u64 scan_time; 1452 int nr_to_walk; 1453 int nr_shrunk = 0; 1454 int retried = 0, nr_skipped = 0; 1455 1456 es_stats = &sbi->s_es_stats; 1457 start_time = ktime_get(); 1458 1459 retry: 1460 spin_lock(&sbi->s_es_lock); 1461 nr_to_walk = sbi->s_es_nr_inode; 1462 while (nr_to_walk-- > 0) { 1463 if (list_empty(&sbi->s_es_list)) { 1464 spin_unlock(&sbi->s_es_lock); 1465 goto out; 1466 } 1467 ei = list_first_entry(&sbi->s_es_list, struct ext4_inode_info, 1468 i_es_list); 1469 /* Move the inode to the tail */ 1470 list_move_tail(&ei->i_es_list, &sbi->s_es_list); 1471 1472 /* 1473 * Normally we try hard to avoid shrinking precached inodes, 1474 * but we will as a last resort. 1475 */ 1476 if (!retried && ext4_test_inode_state(&ei->vfs_inode, 1477 EXT4_STATE_EXT_PRECACHED)) { 1478 nr_skipped++; 1479 continue; 1480 } 1481 1482 if (ei == locked_ei || !write_trylock(&ei->i_es_lock)) { 1483 nr_skipped++; 1484 continue; 1485 } 1486 /* 1487 * Now we hold i_es_lock which protects us from inode reclaim 1488 * freeing inode under us 1489 */ 1490 spin_unlock(&sbi->s_es_lock); 1491 1492 nr_shrunk += es_reclaim_extents(ei, &nr_to_scan); 1493 write_unlock(&ei->i_es_lock); 1494 1495 if (nr_to_scan <= 0) 1496 goto out; 1497 spin_lock(&sbi->s_es_lock); 1498 } 1499 spin_unlock(&sbi->s_es_lock); 1500 1501 /* 1502 * If we skipped any inodes, and we weren't able to make any 1503 * forward progress, try again to scan precached inodes. 1504 */ 1505 if ((nr_shrunk == 0) && nr_skipped && !retried) { 1506 retried++; 1507 goto retry; 1508 } 1509 1510 if (locked_ei && nr_shrunk == 0) 1511 nr_shrunk = es_reclaim_extents(locked_ei, &nr_to_scan); 1512 1513 out: 1514 scan_time = ktime_to_ns(ktime_sub(ktime_get(), start_time)); 1515 if (likely(es_stats->es_stats_scan_time)) 1516 es_stats->es_stats_scan_time = (scan_time + 1517 es_stats->es_stats_scan_time*3) / 4; 1518 else 1519 es_stats->es_stats_scan_time = scan_time; 1520 if (scan_time > es_stats->es_stats_max_scan_time) 1521 es_stats->es_stats_max_scan_time = scan_time; 1522 if (likely(es_stats->es_stats_shrunk)) 1523 es_stats->es_stats_shrunk = (nr_shrunk + 1524 es_stats->es_stats_shrunk*3) / 4; 1525 else 1526 es_stats->es_stats_shrunk = nr_shrunk; 1527 1528 trace_ext4_es_shrink(sbi->s_sb, nr_shrunk, scan_time, 1529 nr_skipped, retried); 1530 return nr_shrunk; 1531 } 1532 1533 static unsigned long ext4_es_count(struct shrinker *shrink, 1534 struct shrink_control *sc) 1535 { 1536 unsigned long nr; 1537 struct ext4_sb_info *sbi; 1538 1539 sbi = container_of(shrink, struct ext4_sb_info, s_es_shrinker); 1540 nr = percpu_counter_read_positive(&sbi->s_es_stats.es_stats_shk_cnt); 1541 trace_ext4_es_shrink_count(sbi->s_sb, sc->nr_to_scan, nr); 1542 return nr; 1543 } 1544 1545 static unsigned long ext4_es_scan(struct shrinker *shrink, 1546 struct shrink_control *sc) 1547 { 1548 struct ext4_sb_info *sbi = container_of(shrink, 1549 struct ext4_sb_info, s_es_shrinker); 1550 int nr_to_scan = sc->nr_to_scan; 1551 int ret, nr_shrunk; 1552 1553 ret = percpu_counter_read_positive(&sbi->s_es_stats.es_stats_shk_cnt); 1554 trace_ext4_es_shrink_scan_enter(sbi->s_sb, nr_to_scan, ret); 1555 1556 if (!nr_to_scan) 1557 return ret; 1558 1559 nr_shrunk = __es_shrink(sbi, nr_to_scan, NULL); 1560 1561 trace_ext4_es_shrink_scan_exit(sbi->s_sb, nr_shrunk, ret); 1562 return nr_shrunk; 1563 } 1564 1565 int ext4_seq_es_shrinker_info_show(struct seq_file *seq, void *v) 1566 { 1567 struct ext4_sb_info *sbi = EXT4_SB((struct super_block *) seq->private); 1568 struct ext4_es_stats *es_stats = &sbi->s_es_stats; 1569 struct ext4_inode_info *ei, *max = NULL; 1570 unsigned int inode_cnt = 0; 1571 1572 if (v != SEQ_START_TOKEN) 1573 return 0; 1574 1575 /* here we just find an inode that has the max nr. of objects */ 1576 spin_lock(&sbi->s_es_lock); 1577 list_for_each_entry(ei, &sbi->s_es_list, i_es_list) { 1578 inode_cnt++; 1579 if (max && max->i_es_all_nr < ei->i_es_all_nr) 1580 max = ei; 1581 else if (!max) 1582 max = ei; 1583 } 1584 spin_unlock(&sbi->s_es_lock); 1585 1586 seq_printf(seq, "stats:\n %lld objects\n %lld reclaimable objects\n", 1587 percpu_counter_sum_positive(&es_stats->es_stats_all_cnt), 1588 percpu_counter_sum_positive(&es_stats->es_stats_shk_cnt)); 1589 seq_printf(seq, " %lld/%lld cache hits/misses\n", 1590 percpu_counter_sum_positive(&es_stats->es_stats_cache_hits), 1591 percpu_counter_sum_positive(&es_stats->es_stats_cache_misses)); 1592 if (inode_cnt) 1593 seq_printf(seq, " %d inodes on list\n", inode_cnt); 1594 1595 seq_printf(seq, "average:\n %llu us scan time\n", 1596 div_u64(es_stats->es_stats_scan_time, 1000)); 1597 seq_printf(seq, " %lu shrunk objects\n", es_stats->es_stats_shrunk); 1598 if (inode_cnt) 1599 seq_printf(seq, 1600 "maximum:\n %lu inode (%u objects, %u reclaimable)\n" 1601 " %llu us max scan time\n", 1602 max->vfs_inode.i_ino, max->i_es_all_nr, max->i_es_shk_nr, 1603 div_u64(es_stats->es_stats_max_scan_time, 1000)); 1604 1605 return 0; 1606 } 1607 1608 int ext4_es_register_shrinker(struct ext4_sb_info *sbi) 1609 { 1610 int err; 1611 1612 /* Make sure we have enough bits for physical block number */ 1613 BUILD_BUG_ON(ES_SHIFT < 48); 1614 INIT_LIST_HEAD(&sbi->s_es_list); 1615 sbi->s_es_nr_inode = 0; 1616 spin_lock_init(&sbi->s_es_lock); 1617 sbi->s_es_stats.es_stats_shrunk = 0; 1618 err = percpu_counter_init(&sbi->s_es_stats.es_stats_cache_hits, 0, 1619 GFP_KERNEL); 1620 if (err) 1621 return err; 1622 err = percpu_counter_init(&sbi->s_es_stats.es_stats_cache_misses, 0, 1623 GFP_KERNEL); 1624 if (err) 1625 goto err1; 1626 sbi->s_es_stats.es_stats_scan_time = 0; 1627 sbi->s_es_stats.es_stats_max_scan_time = 0; 1628 err = percpu_counter_init(&sbi->s_es_stats.es_stats_all_cnt, 0, GFP_KERNEL); 1629 if (err) 1630 goto err2; 1631 err = percpu_counter_init(&sbi->s_es_stats.es_stats_shk_cnt, 0, GFP_KERNEL); 1632 if (err) 1633 goto err3; 1634 1635 sbi->s_es_shrinker.scan_objects = ext4_es_scan; 1636 sbi->s_es_shrinker.count_objects = ext4_es_count; 1637 sbi->s_es_shrinker.seeks = DEFAULT_SEEKS; 1638 err = register_shrinker(&sbi->s_es_shrinker); 1639 if (err) 1640 goto err4; 1641 1642 return 0; 1643 err4: 1644 percpu_counter_destroy(&sbi->s_es_stats.es_stats_shk_cnt); 1645 err3: 1646 percpu_counter_destroy(&sbi->s_es_stats.es_stats_all_cnt); 1647 err2: 1648 percpu_counter_destroy(&sbi->s_es_stats.es_stats_cache_misses); 1649 err1: 1650 percpu_counter_destroy(&sbi->s_es_stats.es_stats_cache_hits); 1651 return err; 1652 } 1653 1654 void ext4_es_unregister_shrinker(struct ext4_sb_info *sbi) 1655 { 1656 percpu_counter_destroy(&sbi->s_es_stats.es_stats_cache_hits); 1657 percpu_counter_destroy(&sbi->s_es_stats.es_stats_cache_misses); 1658 percpu_counter_destroy(&sbi->s_es_stats.es_stats_all_cnt); 1659 percpu_counter_destroy(&sbi->s_es_stats.es_stats_shk_cnt); 1660 unregister_shrinker(&sbi->s_es_shrinker); 1661 } 1662 1663 /* 1664 * Shrink extents in given inode from ei->i_es_shrink_lblk till end. Scan at 1665 * most *nr_to_scan extents, update *nr_to_scan accordingly. 1666 * 1667 * Return 0 if we hit end of tree / interval, 1 if we exhausted nr_to_scan. 1668 * Increment *nr_shrunk by the number of reclaimed extents. Also update 1669 * ei->i_es_shrink_lblk to where we should continue scanning. 1670 */ 1671 static int es_do_reclaim_extents(struct ext4_inode_info *ei, ext4_lblk_t end, 1672 int *nr_to_scan, int *nr_shrunk) 1673 { 1674 struct inode *inode = &ei->vfs_inode; 1675 struct ext4_es_tree *tree = &ei->i_es_tree; 1676 struct extent_status *es; 1677 struct rb_node *node; 1678 1679 es = __es_tree_search(&tree->root, ei->i_es_shrink_lblk); 1680 if (!es) 1681 goto out_wrap; 1682 1683 while (*nr_to_scan > 0) { 1684 if (es->es_lblk > end) { 1685 ei->i_es_shrink_lblk = end + 1; 1686 return 0; 1687 } 1688 1689 (*nr_to_scan)--; 1690 node = rb_next(&es->rb_node); 1691 /* 1692 * We can't reclaim delayed extent from status tree because 1693 * fiemap, bigallic, and seek_data/hole need to use it. 1694 */ 1695 if (ext4_es_is_delayed(es)) 1696 goto next; 1697 if (ext4_es_is_referenced(es)) { 1698 ext4_es_clear_referenced(es); 1699 goto next; 1700 } 1701 1702 rb_erase(&es->rb_node, &tree->root); 1703 ext4_es_free_extent(inode, es); 1704 (*nr_shrunk)++; 1705 next: 1706 if (!node) 1707 goto out_wrap; 1708 es = rb_entry(node, struct extent_status, rb_node); 1709 } 1710 ei->i_es_shrink_lblk = es->es_lblk; 1711 return 1; 1712 out_wrap: 1713 ei->i_es_shrink_lblk = 0; 1714 return 0; 1715 } 1716 1717 static int es_reclaim_extents(struct ext4_inode_info *ei, int *nr_to_scan) 1718 { 1719 struct inode *inode = &ei->vfs_inode; 1720 int nr_shrunk = 0; 1721 ext4_lblk_t start = ei->i_es_shrink_lblk; 1722 static DEFINE_RATELIMIT_STATE(_rs, DEFAULT_RATELIMIT_INTERVAL, 1723 DEFAULT_RATELIMIT_BURST); 1724 1725 if (ei->i_es_shk_nr == 0) 1726 return 0; 1727 1728 if (ext4_test_inode_state(inode, EXT4_STATE_EXT_PRECACHED) && 1729 __ratelimit(&_rs)) 1730 ext4_warning(inode->i_sb, "forced shrink of precached extents"); 1731 1732 if (!es_do_reclaim_extents(ei, EXT_MAX_BLOCKS, nr_to_scan, &nr_shrunk) && 1733 start != 0) 1734 es_do_reclaim_extents(ei, start - 1, nr_to_scan, &nr_shrunk); 1735 1736 ei->i_es_tree.cache_es = NULL; 1737 return nr_shrunk; 1738 } 1739 1740 /* 1741 * Called to support EXT4_IOC_CLEAR_ES_CACHE. We can only remove 1742 * discretionary entries from the extent status cache. (Some entries 1743 * must be present for proper operations.) 1744 */ 1745 void ext4_clear_inode_es(struct inode *inode) 1746 { 1747 struct ext4_inode_info *ei = EXT4_I(inode); 1748 struct extent_status *es; 1749 struct ext4_es_tree *tree; 1750 struct rb_node *node; 1751 1752 write_lock(&ei->i_es_lock); 1753 tree = &EXT4_I(inode)->i_es_tree; 1754 tree->cache_es = NULL; 1755 node = rb_first(&tree->root); 1756 while (node) { 1757 es = rb_entry(node, struct extent_status, rb_node); 1758 node = rb_next(node); 1759 if (!ext4_es_is_delayed(es)) { 1760 rb_erase(&es->rb_node, &tree->root); 1761 ext4_es_free_extent(inode, es); 1762 } 1763 } 1764 ext4_clear_inode_state(inode, EXT4_STATE_EXT_PRECACHED); 1765 write_unlock(&ei->i_es_lock); 1766 } 1767 1768 #ifdef ES_DEBUG__ 1769 static void ext4_print_pending_tree(struct inode *inode) 1770 { 1771 struct ext4_pending_tree *tree; 1772 struct rb_node *node; 1773 struct pending_reservation *pr; 1774 1775 printk(KERN_DEBUG "pending reservations for inode %lu:", inode->i_ino); 1776 tree = &EXT4_I(inode)->i_pending_tree; 1777 node = rb_first(&tree->root); 1778 while (node) { 1779 pr = rb_entry(node, struct pending_reservation, rb_node); 1780 printk(KERN_DEBUG " %u", pr->lclu); 1781 node = rb_next(node); 1782 } 1783 printk(KERN_DEBUG "\n"); 1784 } 1785 #else 1786 #define ext4_print_pending_tree(inode) 1787 #endif 1788 1789 int __init ext4_init_pending(void) 1790 { 1791 ext4_pending_cachep = kmem_cache_create("ext4_pending_reservation", 1792 sizeof(struct pending_reservation), 1793 0, (SLAB_RECLAIM_ACCOUNT), NULL); 1794 if (ext4_pending_cachep == NULL) 1795 return -ENOMEM; 1796 return 0; 1797 } 1798 1799 void ext4_exit_pending(void) 1800 { 1801 kmem_cache_destroy(ext4_pending_cachep); 1802 } 1803 1804 void ext4_init_pending_tree(struct ext4_pending_tree *tree) 1805 { 1806 tree->root = RB_ROOT; 1807 } 1808 1809 /* 1810 * __get_pending - retrieve a pointer to a pending reservation 1811 * 1812 * @inode - file containing the pending cluster reservation 1813 * @lclu - logical cluster of interest 1814 * 1815 * Returns a pointer to a pending reservation if it's a member of 1816 * the set, and NULL if not. Must be called holding i_es_lock. 1817 */ 1818 static struct pending_reservation *__get_pending(struct inode *inode, 1819 ext4_lblk_t lclu) 1820 { 1821 struct ext4_pending_tree *tree; 1822 struct rb_node *node; 1823 struct pending_reservation *pr = NULL; 1824 1825 tree = &EXT4_I(inode)->i_pending_tree; 1826 node = (&tree->root)->rb_node; 1827 1828 while (node) { 1829 pr = rb_entry(node, struct pending_reservation, rb_node); 1830 if (lclu < pr->lclu) 1831 node = node->rb_left; 1832 else if (lclu > pr->lclu) 1833 node = node->rb_right; 1834 else if (lclu == pr->lclu) 1835 return pr; 1836 } 1837 return NULL; 1838 } 1839 1840 /* 1841 * __insert_pending - adds a pending cluster reservation to the set of 1842 * pending reservations 1843 * 1844 * @inode - file containing the cluster 1845 * @lblk - logical block in the cluster to be added 1846 * 1847 * Returns 0 on successful insertion and -ENOMEM on failure. If the 1848 * pending reservation is already in the set, returns successfully. 1849 */ 1850 static int __insert_pending(struct inode *inode, ext4_lblk_t lblk) 1851 { 1852 struct ext4_sb_info *sbi = EXT4_SB(inode->i_sb); 1853 struct ext4_pending_tree *tree = &EXT4_I(inode)->i_pending_tree; 1854 struct rb_node **p = &tree->root.rb_node; 1855 struct rb_node *parent = NULL; 1856 struct pending_reservation *pr; 1857 ext4_lblk_t lclu; 1858 int ret = 0; 1859 1860 lclu = EXT4_B2C(sbi, lblk); 1861 /* search to find parent for insertion */ 1862 while (*p) { 1863 parent = *p; 1864 pr = rb_entry(parent, struct pending_reservation, rb_node); 1865 1866 if (lclu < pr->lclu) { 1867 p = &(*p)->rb_left; 1868 } else if (lclu > pr->lclu) { 1869 p = &(*p)->rb_right; 1870 } else { 1871 /* pending reservation already inserted */ 1872 goto out; 1873 } 1874 } 1875 1876 pr = kmem_cache_alloc(ext4_pending_cachep, GFP_ATOMIC); 1877 if (pr == NULL) { 1878 ret = -ENOMEM; 1879 goto out; 1880 } 1881 pr->lclu = lclu; 1882 1883 rb_link_node(&pr->rb_node, parent, p); 1884 rb_insert_color(&pr->rb_node, &tree->root); 1885 1886 out: 1887 return ret; 1888 } 1889 1890 /* 1891 * __remove_pending - removes a pending cluster reservation from the set 1892 * of pending reservations 1893 * 1894 * @inode - file containing the cluster 1895 * @lblk - logical block in the pending cluster reservation to be removed 1896 * 1897 * Returns successfully if pending reservation is not a member of the set. 1898 */ 1899 static void __remove_pending(struct inode *inode, ext4_lblk_t lblk) 1900 { 1901 struct ext4_sb_info *sbi = EXT4_SB(inode->i_sb); 1902 struct pending_reservation *pr; 1903 struct ext4_pending_tree *tree; 1904 1905 pr = __get_pending(inode, EXT4_B2C(sbi, lblk)); 1906 if (pr != NULL) { 1907 tree = &EXT4_I(inode)->i_pending_tree; 1908 rb_erase(&pr->rb_node, &tree->root); 1909 kmem_cache_free(ext4_pending_cachep, pr); 1910 } 1911 } 1912 1913 /* 1914 * ext4_remove_pending - removes a pending cluster reservation from the set 1915 * of pending reservations 1916 * 1917 * @inode - file containing the cluster 1918 * @lblk - logical block in the pending cluster reservation to be removed 1919 * 1920 * Locking for external use of __remove_pending. 1921 */ 1922 void ext4_remove_pending(struct inode *inode, ext4_lblk_t lblk) 1923 { 1924 struct ext4_inode_info *ei = EXT4_I(inode); 1925 1926 write_lock(&ei->i_es_lock); 1927 __remove_pending(inode, lblk); 1928 write_unlock(&ei->i_es_lock); 1929 } 1930 1931 /* 1932 * ext4_is_pending - determine whether a cluster has a pending reservation 1933 * on it 1934 * 1935 * @inode - file containing the cluster 1936 * @lblk - logical block in the cluster 1937 * 1938 * Returns true if there's a pending reservation for the cluster in the 1939 * set of pending reservations, and false if not. 1940 */ 1941 bool ext4_is_pending(struct inode *inode, ext4_lblk_t lblk) 1942 { 1943 struct ext4_sb_info *sbi = EXT4_SB(inode->i_sb); 1944 struct ext4_inode_info *ei = EXT4_I(inode); 1945 bool ret; 1946 1947 read_lock(&ei->i_es_lock); 1948 ret = (bool)(__get_pending(inode, EXT4_B2C(sbi, lblk)) != NULL); 1949 read_unlock(&ei->i_es_lock); 1950 1951 return ret; 1952 } 1953 1954 /* 1955 * ext4_es_insert_delayed_block - adds a delayed block to the extents status 1956 * tree, adding a pending reservation where 1957 * needed 1958 * 1959 * @inode - file containing the newly added block 1960 * @lblk - logical block to be added 1961 * @allocated - indicates whether a physical cluster has been allocated for 1962 * the logical cluster that contains the block 1963 * 1964 * Returns 0 on success, negative error code on failure. 1965 */ 1966 int ext4_es_insert_delayed_block(struct inode *inode, ext4_lblk_t lblk, 1967 bool allocated) 1968 { 1969 struct extent_status newes; 1970 int err = 0; 1971 1972 es_debug("add [%u/1) delayed to extent status tree of inode %lu\n", 1973 lblk, inode->i_ino); 1974 1975 newes.es_lblk = lblk; 1976 newes.es_len = 1; 1977 ext4_es_store_pblock_status(&newes, ~0, EXTENT_STATUS_DELAYED); 1978 trace_ext4_es_insert_delayed_block(inode, &newes, allocated); 1979 1980 ext4_es_insert_extent_check(inode, &newes); 1981 1982 write_lock(&EXT4_I(inode)->i_es_lock); 1983 1984 err = __es_remove_extent(inode, lblk, lblk, NULL); 1985 if (err != 0) 1986 goto error; 1987 retry: 1988 err = __es_insert_extent(inode, &newes); 1989 if (err == -ENOMEM && __es_shrink(EXT4_SB(inode->i_sb), 1990 128, EXT4_I(inode))) 1991 goto retry; 1992 if (err != 0) 1993 goto error; 1994 1995 if (allocated) 1996 __insert_pending(inode, lblk); 1997 1998 error: 1999 write_unlock(&EXT4_I(inode)->i_es_lock); 2000 2001 ext4_es_print_tree(inode); 2002 ext4_print_pending_tree(inode); 2003 2004 return err; 2005 } 2006 2007 /* 2008 * __es_delayed_clu - count number of clusters containing blocks that 2009 * are delayed only 2010 * 2011 * @inode - file containing block range 2012 * @start - logical block defining start of range 2013 * @end - logical block defining end of range 2014 * 2015 * Returns the number of clusters containing only delayed (not delayed 2016 * and unwritten) blocks in the range specified by @start and @end. Any 2017 * cluster or part of a cluster within the range and containing a delayed 2018 * and not unwritten block within the range is counted as a whole cluster. 2019 */ 2020 static unsigned int __es_delayed_clu(struct inode *inode, ext4_lblk_t start, 2021 ext4_lblk_t end) 2022 { 2023 struct ext4_es_tree *tree = &EXT4_I(inode)->i_es_tree; 2024 struct extent_status *es; 2025 struct ext4_sb_info *sbi = EXT4_SB(inode->i_sb); 2026 struct rb_node *node; 2027 ext4_lblk_t first_lclu, last_lclu; 2028 unsigned long long last_counted_lclu; 2029 unsigned int n = 0; 2030 2031 /* guaranteed to be unequal to any ext4_lblk_t value */ 2032 last_counted_lclu = ~0ULL; 2033 2034 es = __es_tree_search(&tree->root, start); 2035 2036 while (es && (es->es_lblk <= end)) { 2037 if (ext4_es_is_delonly(es)) { 2038 if (es->es_lblk <= start) 2039 first_lclu = EXT4_B2C(sbi, start); 2040 else 2041 first_lclu = EXT4_B2C(sbi, es->es_lblk); 2042 2043 if (ext4_es_end(es) >= end) 2044 last_lclu = EXT4_B2C(sbi, end); 2045 else 2046 last_lclu = EXT4_B2C(sbi, ext4_es_end(es)); 2047 2048 if (first_lclu == last_counted_lclu) 2049 n += last_lclu - first_lclu; 2050 else 2051 n += last_lclu - first_lclu + 1; 2052 last_counted_lclu = last_lclu; 2053 } 2054 node = rb_next(&es->rb_node); 2055 if (!node) 2056 break; 2057 es = rb_entry(node, struct extent_status, rb_node); 2058 } 2059 2060 return n; 2061 } 2062 2063 /* 2064 * ext4_es_delayed_clu - count number of clusters containing blocks that 2065 * are both delayed and unwritten 2066 * 2067 * @inode - file containing block range 2068 * @lblk - logical block defining start of range 2069 * @len - number of blocks in range 2070 * 2071 * Locking for external use of __es_delayed_clu(). 2072 */ 2073 unsigned int ext4_es_delayed_clu(struct inode *inode, ext4_lblk_t lblk, 2074 ext4_lblk_t len) 2075 { 2076 struct ext4_inode_info *ei = EXT4_I(inode); 2077 ext4_lblk_t end; 2078 unsigned int n; 2079 2080 if (len == 0) 2081 return 0; 2082 2083 end = lblk + len - 1; 2084 WARN_ON(end < lblk); 2085 2086 read_lock(&ei->i_es_lock); 2087 2088 n = __es_delayed_clu(inode, lblk, end); 2089 2090 read_unlock(&ei->i_es_lock); 2091 2092 return n; 2093 } 2094 2095 /* 2096 * __revise_pending - makes, cancels, or leaves unchanged pending cluster 2097 * reservations for a specified block range depending 2098 * upon the presence or absence of delayed blocks 2099 * outside the range within clusters at the ends of the 2100 * range 2101 * 2102 * @inode - file containing the range 2103 * @lblk - logical block defining the start of range 2104 * @len - length of range in blocks 2105 * 2106 * Used after a newly allocated extent is added to the extents status tree. 2107 * Requires that the extents in the range have either written or unwritten 2108 * status. Must be called while holding i_es_lock. 2109 */ 2110 static void __revise_pending(struct inode *inode, ext4_lblk_t lblk, 2111 ext4_lblk_t len) 2112 { 2113 struct ext4_sb_info *sbi = EXT4_SB(inode->i_sb); 2114 ext4_lblk_t end = lblk + len - 1; 2115 ext4_lblk_t first, last; 2116 bool f_del = false, l_del = false; 2117 2118 if (len == 0) 2119 return; 2120 2121 /* 2122 * Two cases - block range within single cluster and block range 2123 * spanning two or more clusters. Note that a cluster belonging 2124 * to a range starting and/or ending on a cluster boundary is treated 2125 * as if it does not contain a delayed extent. The new range may 2126 * have allocated space for previously delayed blocks out to the 2127 * cluster boundary, requiring that any pre-existing pending 2128 * reservation be canceled. Because this code only looks at blocks 2129 * outside the range, it should revise pending reservations 2130 * correctly even if the extent represented by the range can't be 2131 * inserted in the extents status tree due to ENOSPC. 2132 */ 2133 2134 if (EXT4_B2C(sbi, lblk) == EXT4_B2C(sbi, end)) { 2135 first = EXT4_LBLK_CMASK(sbi, lblk); 2136 if (first != lblk) 2137 f_del = __es_scan_range(inode, &ext4_es_is_delonly, 2138 first, lblk - 1); 2139 if (f_del) { 2140 __insert_pending(inode, first); 2141 } else { 2142 last = EXT4_LBLK_CMASK(sbi, end) + 2143 sbi->s_cluster_ratio - 1; 2144 if (last != end) 2145 l_del = __es_scan_range(inode, 2146 &ext4_es_is_delonly, 2147 end + 1, last); 2148 if (l_del) 2149 __insert_pending(inode, last); 2150 else 2151 __remove_pending(inode, last); 2152 } 2153 } else { 2154 first = EXT4_LBLK_CMASK(sbi, lblk); 2155 if (first != lblk) 2156 f_del = __es_scan_range(inode, &ext4_es_is_delonly, 2157 first, lblk - 1); 2158 if (f_del) 2159 __insert_pending(inode, first); 2160 else 2161 __remove_pending(inode, first); 2162 2163 last = EXT4_LBLK_CMASK(sbi, end) + sbi->s_cluster_ratio - 1; 2164 if (last != end) 2165 l_del = __es_scan_range(inode, &ext4_es_is_delonly, 2166 end + 1, last); 2167 if (l_del) 2168 __insert_pending(inode, last); 2169 else 2170 __remove_pending(inode, last); 2171 } 2172 } 2173