xref: /openbmc/linux/fs/ext4/extents_status.c (revision faffb083)
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(extent_status, SLAB_RECLAIM_ACCOUNT);
159 	if (ext4_es_cachep == NULL)
160 		return -ENOMEM;
161 	return 0;
162 }
163 
164 void ext4_exit_es(void)
165 {
166 	kmem_cache_destroy(ext4_es_cachep);
167 }
168 
169 void ext4_es_init_tree(struct ext4_es_tree *tree)
170 {
171 	tree->root = RB_ROOT;
172 	tree->cache_es = NULL;
173 }
174 
175 #ifdef ES_DEBUG__
176 static void ext4_es_print_tree(struct inode *inode)
177 {
178 	struct ext4_es_tree *tree;
179 	struct rb_node *node;
180 
181 	printk(KERN_DEBUG "status extents for inode %lu:", inode->i_ino);
182 	tree = &EXT4_I(inode)->i_es_tree;
183 	node = rb_first(&tree->root);
184 	while (node) {
185 		struct extent_status *es;
186 		es = rb_entry(node, struct extent_status, rb_node);
187 		printk(KERN_DEBUG " [%u/%u) %llu %x",
188 		       es->es_lblk, es->es_len,
189 		       ext4_es_pblock(es), ext4_es_status(es));
190 		node = rb_next(node);
191 	}
192 	printk(KERN_DEBUG "\n");
193 }
194 #else
195 #define ext4_es_print_tree(inode)
196 #endif
197 
198 static inline ext4_lblk_t ext4_es_end(struct extent_status *es)
199 {
200 	BUG_ON(es->es_lblk + es->es_len < es->es_lblk);
201 	return es->es_lblk + es->es_len - 1;
202 }
203 
204 /*
205  * search through the tree for an delayed extent with a given offset.  If
206  * it can't be found, try to find next extent.
207  */
208 static struct extent_status *__es_tree_search(struct rb_root *root,
209 					      ext4_lblk_t lblk)
210 {
211 	struct rb_node *node = root->rb_node;
212 	struct extent_status *es = NULL;
213 
214 	while (node) {
215 		es = rb_entry(node, struct extent_status, rb_node);
216 		if (lblk < es->es_lblk)
217 			node = node->rb_left;
218 		else if (lblk > ext4_es_end(es))
219 			node = node->rb_right;
220 		else
221 			return es;
222 	}
223 
224 	if (es && lblk < es->es_lblk)
225 		return es;
226 
227 	if (es && lblk > ext4_es_end(es)) {
228 		node = rb_next(&es->rb_node);
229 		return node ? rb_entry(node, struct extent_status, rb_node) :
230 			      NULL;
231 	}
232 
233 	return NULL;
234 }
235 
236 /*
237  * ext4_es_find_extent_range - find extent with specified status within block
238  *                             range or next extent following block range in
239  *                             extents status tree
240  *
241  * @inode - file containing the range
242  * @matching_fn - pointer to function that matches extents with desired status
243  * @lblk - logical block defining start of range
244  * @end - logical block defining end of range
245  * @es - extent found, if any
246  *
247  * Find the first extent within the block range specified by @lblk and @end
248  * in the extents status tree that satisfies @matching_fn.  If a match
249  * is found, it's returned in @es.  If not, and a matching extent is found
250  * beyond the block range, it's returned in @es.  If no match is found, an
251  * extent is returned in @es whose es_lblk, es_len, and es_pblk components
252  * are 0.
253  */
254 static void __es_find_extent_range(struct inode *inode,
255 				   int (*matching_fn)(struct extent_status *es),
256 				   ext4_lblk_t lblk, ext4_lblk_t end,
257 				   struct extent_status *es)
258 {
259 	struct ext4_es_tree *tree = NULL;
260 	struct extent_status *es1 = NULL;
261 	struct rb_node *node;
262 
263 	WARN_ON(es == NULL);
264 	WARN_ON(end < lblk);
265 
266 	tree = &EXT4_I(inode)->i_es_tree;
267 
268 	/* see if the extent has been cached */
269 	es->es_lblk = es->es_len = es->es_pblk = 0;
270 	if (tree->cache_es) {
271 		es1 = tree->cache_es;
272 		if (in_range(lblk, es1->es_lblk, es1->es_len)) {
273 			es_debug("%u cached by [%u/%u) %llu %x\n",
274 				 lblk, es1->es_lblk, es1->es_len,
275 				 ext4_es_pblock(es1), ext4_es_status(es1));
276 			goto out;
277 		}
278 	}
279 
280 	es1 = __es_tree_search(&tree->root, lblk);
281 
282 out:
283 	if (es1 && !matching_fn(es1)) {
284 		while ((node = rb_next(&es1->rb_node)) != NULL) {
285 			es1 = rb_entry(node, struct extent_status, rb_node);
286 			if (es1->es_lblk > end) {
287 				es1 = NULL;
288 				break;
289 			}
290 			if (matching_fn(es1))
291 				break;
292 		}
293 	}
294 
295 	if (es1 && matching_fn(es1)) {
296 		tree->cache_es = es1;
297 		es->es_lblk = es1->es_lblk;
298 		es->es_len = es1->es_len;
299 		es->es_pblk = es1->es_pblk;
300 	}
301 
302 }
303 
304 /*
305  * Locking for __es_find_extent_range() for external use
306  */
307 void ext4_es_find_extent_range(struct inode *inode,
308 			       int (*matching_fn)(struct extent_status *es),
309 			       ext4_lblk_t lblk, ext4_lblk_t end,
310 			       struct extent_status *es)
311 {
312 	if (EXT4_SB(inode->i_sb)->s_mount_state & EXT4_FC_REPLAY)
313 		return;
314 
315 	trace_ext4_es_find_extent_range_enter(inode, lblk);
316 
317 	read_lock(&EXT4_I(inode)->i_es_lock);
318 	__es_find_extent_range(inode, matching_fn, lblk, end, es);
319 	read_unlock(&EXT4_I(inode)->i_es_lock);
320 
321 	trace_ext4_es_find_extent_range_exit(inode, es);
322 }
323 
324 /*
325  * __es_scan_range - search block range for block with specified status
326  *                   in extents status tree
327  *
328  * @inode - file containing the range
329  * @matching_fn - pointer to function that matches extents with desired status
330  * @lblk - logical block defining start of range
331  * @end - logical block defining end of range
332  *
333  * Returns true if at least one block in the specified block range satisfies
334  * the criterion specified by @matching_fn, and false if not.  If at least
335  * one extent has the specified status, then there is at least one block
336  * in the cluster with that status.  Should only be called by code that has
337  * taken i_es_lock.
338  */
339 static bool __es_scan_range(struct inode *inode,
340 			    int (*matching_fn)(struct extent_status *es),
341 			    ext4_lblk_t start, ext4_lblk_t end)
342 {
343 	struct extent_status es;
344 
345 	__es_find_extent_range(inode, matching_fn, start, end, &es);
346 	if (es.es_len == 0)
347 		return false;   /* no matching extent in the tree */
348 	else if (es.es_lblk <= start &&
349 		 start < es.es_lblk + es.es_len)
350 		return true;
351 	else if (start <= es.es_lblk && es.es_lblk <= end)
352 		return true;
353 	else
354 		return false;
355 }
356 /*
357  * Locking for __es_scan_range() for external use
358  */
359 bool ext4_es_scan_range(struct inode *inode,
360 			int (*matching_fn)(struct extent_status *es),
361 			ext4_lblk_t lblk, ext4_lblk_t end)
362 {
363 	bool ret;
364 
365 	if (EXT4_SB(inode->i_sb)->s_mount_state & EXT4_FC_REPLAY)
366 		return false;
367 
368 	read_lock(&EXT4_I(inode)->i_es_lock);
369 	ret = __es_scan_range(inode, matching_fn, lblk, end);
370 	read_unlock(&EXT4_I(inode)->i_es_lock);
371 
372 	return ret;
373 }
374 
375 /*
376  * __es_scan_clu - search cluster for block with specified status in
377  *                 extents status tree
378  *
379  * @inode - file containing the cluster
380  * @matching_fn - pointer to function that matches extents with desired status
381  * @lblk - logical block in cluster to be searched
382  *
383  * Returns true if at least one extent in the cluster containing @lblk
384  * satisfies the criterion specified by @matching_fn, and false if not.  If at
385  * least one extent has the specified status, then there is at least one block
386  * in the cluster with that status.  Should only be called by code that has
387  * taken i_es_lock.
388  */
389 static bool __es_scan_clu(struct inode *inode,
390 			  int (*matching_fn)(struct extent_status *es),
391 			  ext4_lblk_t lblk)
392 {
393 	struct ext4_sb_info *sbi = EXT4_SB(inode->i_sb);
394 	ext4_lblk_t lblk_start, lblk_end;
395 
396 	lblk_start = EXT4_LBLK_CMASK(sbi, lblk);
397 	lblk_end = lblk_start + sbi->s_cluster_ratio - 1;
398 
399 	return __es_scan_range(inode, matching_fn, lblk_start, lblk_end);
400 }
401 
402 /*
403  * Locking for __es_scan_clu() for external use
404  */
405 bool ext4_es_scan_clu(struct inode *inode,
406 		      int (*matching_fn)(struct extent_status *es),
407 		      ext4_lblk_t lblk)
408 {
409 	bool ret;
410 
411 	if (EXT4_SB(inode->i_sb)->s_mount_state & EXT4_FC_REPLAY)
412 		return false;
413 
414 	read_lock(&EXT4_I(inode)->i_es_lock);
415 	ret = __es_scan_clu(inode, matching_fn, lblk);
416 	read_unlock(&EXT4_I(inode)->i_es_lock);
417 
418 	return ret;
419 }
420 
421 static void ext4_es_list_add(struct inode *inode)
422 {
423 	struct ext4_inode_info *ei = EXT4_I(inode);
424 	struct ext4_sb_info *sbi = EXT4_SB(inode->i_sb);
425 
426 	if (!list_empty(&ei->i_es_list))
427 		return;
428 
429 	spin_lock(&sbi->s_es_lock);
430 	if (list_empty(&ei->i_es_list)) {
431 		list_add_tail(&ei->i_es_list, &sbi->s_es_list);
432 		sbi->s_es_nr_inode++;
433 	}
434 	spin_unlock(&sbi->s_es_lock);
435 }
436 
437 static void ext4_es_list_del(struct inode *inode)
438 {
439 	struct ext4_inode_info *ei = EXT4_I(inode);
440 	struct ext4_sb_info *sbi = EXT4_SB(inode->i_sb);
441 
442 	spin_lock(&sbi->s_es_lock);
443 	if (!list_empty(&ei->i_es_list)) {
444 		list_del_init(&ei->i_es_list);
445 		sbi->s_es_nr_inode--;
446 		WARN_ON_ONCE(sbi->s_es_nr_inode < 0);
447 	}
448 	spin_unlock(&sbi->s_es_lock);
449 }
450 
451 static struct extent_status *
452 ext4_es_alloc_extent(struct inode *inode, ext4_lblk_t lblk, ext4_lblk_t len,
453 		     ext4_fsblk_t pblk)
454 {
455 	struct extent_status *es;
456 	es = kmem_cache_alloc(ext4_es_cachep, GFP_ATOMIC);
457 	if (es == NULL)
458 		return NULL;
459 	es->es_lblk = lblk;
460 	es->es_len = len;
461 	es->es_pblk = pblk;
462 
463 	/*
464 	 * We don't count delayed extent because we never try to reclaim them
465 	 */
466 	if (!ext4_es_is_delayed(es)) {
467 		if (!EXT4_I(inode)->i_es_shk_nr++)
468 			ext4_es_list_add(inode);
469 		percpu_counter_inc(&EXT4_SB(inode->i_sb)->
470 					s_es_stats.es_stats_shk_cnt);
471 	}
472 
473 	EXT4_I(inode)->i_es_all_nr++;
474 	percpu_counter_inc(&EXT4_SB(inode->i_sb)->s_es_stats.es_stats_all_cnt);
475 
476 	return es;
477 }
478 
479 static void ext4_es_free_extent(struct inode *inode, struct extent_status *es)
480 {
481 	EXT4_I(inode)->i_es_all_nr--;
482 	percpu_counter_dec(&EXT4_SB(inode->i_sb)->s_es_stats.es_stats_all_cnt);
483 
484 	/* Decrease the shrink counter when this es is not delayed */
485 	if (!ext4_es_is_delayed(es)) {
486 		BUG_ON(EXT4_I(inode)->i_es_shk_nr == 0);
487 		if (!--EXT4_I(inode)->i_es_shk_nr)
488 			ext4_es_list_del(inode);
489 		percpu_counter_dec(&EXT4_SB(inode->i_sb)->
490 					s_es_stats.es_stats_shk_cnt);
491 	}
492 
493 	kmem_cache_free(ext4_es_cachep, es);
494 }
495 
496 /*
497  * Check whether or not two extents can be merged
498  * Condition:
499  *  - logical block number is contiguous
500  *  - physical block number is contiguous
501  *  - status is equal
502  */
503 static int ext4_es_can_be_merged(struct extent_status *es1,
504 				 struct extent_status *es2)
505 {
506 	if (ext4_es_type(es1) != ext4_es_type(es2))
507 		return 0;
508 
509 	if (((__u64) es1->es_len) + es2->es_len > EXT_MAX_BLOCKS) {
510 		pr_warn("ES assertion failed when merging extents. "
511 			"The sum of lengths of es1 (%d) and es2 (%d) "
512 			"is bigger than allowed file size (%d)\n",
513 			es1->es_len, es2->es_len, EXT_MAX_BLOCKS);
514 		WARN_ON(1);
515 		return 0;
516 	}
517 
518 	if (((__u64) es1->es_lblk) + es1->es_len != es2->es_lblk)
519 		return 0;
520 
521 	if ((ext4_es_is_written(es1) || ext4_es_is_unwritten(es1)) &&
522 	    (ext4_es_pblock(es1) + es1->es_len == ext4_es_pblock(es2)))
523 		return 1;
524 
525 	if (ext4_es_is_hole(es1))
526 		return 1;
527 
528 	/* we need to check delayed extent is without unwritten status */
529 	if (ext4_es_is_delayed(es1) && !ext4_es_is_unwritten(es1))
530 		return 1;
531 
532 	return 0;
533 }
534 
535 static struct extent_status *
536 ext4_es_try_to_merge_left(struct inode *inode, struct extent_status *es)
537 {
538 	struct ext4_es_tree *tree = &EXT4_I(inode)->i_es_tree;
539 	struct extent_status *es1;
540 	struct rb_node *node;
541 
542 	node = rb_prev(&es->rb_node);
543 	if (!node)
544 		return es;
545 
546 	es1 = rb_entry(node, struct extent_status, rb_node);
547 	if (ext4_es_can_be_merged(es1, es)) {
548 		es1->es_len += es->es_len;
549 		if (ext4_es_is_referenced(es))
550 			ext4_es_set_referenced(es1);
551 		rb_erase(&es->rb_node, &tree->root);
552 		ext4_es_free_extent(inode, es);
553 		es = es1;
554 	}
555 
556 	return es;
557 }
558 
559 static struct extent_status *
560 ext4_es_try_to_merge_right(struct inode *inode, struct extent_status *es)
561 {
562 	struct ext4_es_tree *tree = &EXT4_I(inode)->i_es_tree;
563 	struct extent_status *es1;
564 	struct rb_node *node;
565 
566 	node = rb_next(&es->rb_node);
567 	if (!node)
568 		return es;
569 
570 	es1 = rb_entry(node, struct extent_status, rb_node);
571 	if (ext4_es_can_be_merged(es, es1)) {
572 		es->es_len += es1->es_len;
573 		if (ext4_es_is_referenced(es1))
574 			ext4_es_set_referenced(es);
575 		rb_erase(node, &tree->root);
576 		ext4_es_free_extent(inode, es1);
577 	}
578 
579 	return es;
580 }
581 
582 #ifdef ES_AGGRESSIVE_TEST
583 #include "ext4_extents.h"	/* Needed when ES_AGGRESSIVE_TEST is defined */
584 
585 static void ext4_es_insert_extent_ext_check(struct inode *inode,
586 					    struct extent_status *es)
587 {
588 	struct ext4_ext_path *path = NULL;
589 	struct ext4_extent *ex;
590 	ext4_lblk_t ee_block;
591 	ext4_fsblk_t ee_start;
592 	unsigned short ee_len;
593 	int depth, ee_status, es_status;
594 
595 	path = ext4_find_extent(inode, es->es_lblk, NULL, EXT4_EX_NOCACHE);
596 	if (IS_ERR(path))
597 		return;
598 
599 	depth = ext_depth(inode);
600 	ex = path[depth].p_ext;
601 
602 	if (ex) {
603 
604 		ee_block = le32_to_cpu(ex->ee_block);
605 		ee_start = ext4_ext_pblock(ex);
606 		ee_len = ext4_ext_get_actual_len(ex);
607 
608 		ee_status = ext4_ext_is_unwritten(ex) ? 1 : 0;
609 		es_status = ext4_es_is_unwritten(es) ? 1 : 0;
610 
611 		/*
612 		 * Make sure ex and es are not overlap when we try to insert
613 		 * a delayed/hole extent.
614 		 */
615 		if (!ext4_es_is_written(es) && !ext4_es_is_unwritten(es)) {
616 			if (in_range(es->es_lblk, ee_block, ee_len)) {
617 				pr_warn("ES insert assertion failed for "
618 					"inode: %lu we can find an extent "
619 					"at block [%d/%d/%llu/%c], but we "
620 					"want to add a delayed/hole extent "
621 					"[%d/%d/%llu/%x]\n",
622 					inode->i_ino, ee_block, ee_len,
623 					ee_start, ee_status ? 'u' : 'w',
624 					es->es_lblk, es->es_len,
625 					ext4_es_pblock(es), ext4_es_status(es));
626 			}
627 			goto out;
628 		}
629 
630 		/*
631 		 * We don't check ee_block == es->es_lblk, etc. because es
632 		 * might be a part of whole extent, vice versa.
633 		 */
634 		if (es->es_lblk < ee_block ||
635 		    ext4_es_pblock(es) != ee_start + es->es_lblk - ee_block) {
636 			pr_warn("ES insert assertion failed for inode: %lu "
637 				"ex_status [%d/%d/%llu/%c] != "
638 				"es_status [%d/%d/%llu/%c]\n", inode->i_ino,
639 				ee_block, ee_len, ee_start,
640 				ee_status ? 'u' : 'w', es->es_lblk, es->es_len,
641 				ext4_es_pblock(es), es_status ? 'u' : 'w');
642 			goto out;
643 		}
644 
645 		if (ee_status ^ es_status) {
646 			pr_warn("ES insert assertion failed for inode: %lu "
647 				"ex_status [%d/%d/%llu/%c] != "
648 				"es_status [%d/%d/%llu/%c]\n", inode->i_ino,
649 				ee_block, ee_len, ee_start,
650 				ee_status ? 'u' : 'w', es->es_lblk, es->es_len,
651 				ext4_es_pblock(es), es_status ? 'u' : 'w');
652 		}
653 	} else {
654 		/*
655 		 * We can't find an extent on disk.  So we need to make sure
656 		 * that we don't want to add an written/unwritten extent.
657 		 */
658 		if (!ext4_es_is_delayed(es) && !ext4_es_is_hole(es)) {
659 			pr_warn("ES insert assertion failed for inode: %lu "
660 				"can't find an extent at block %d but we want "
661 				"to add a written/unwritten extent "
662 				"[%d/%d/%llu/%x]\n", inode->i_ino,
663 				es->es_lblk, es->es_lblk, es->es_len,
664 				ext4_es_pblock(es), ext4_es_status(es));
665 		}
666 	}
667 out:
668 	ext4_free_ext_path(path);
669 }
670 
671 static void ext4_es_insert_extent_ind_check(struct inode *inode,
672 					    struct extent_status *es)
673 {
674 	struct ext4_map_blocks map;
675 	int retval;
676 
677 	/*
678 	 * Here we call ext4_ind_map_blocks to lookup a block mapping because
679 	 * 'Indirect' structure is defined in indirect.c.  So we couldn't
680 	 * access direct/indirect tree from outside.  It is too dirty to define
681 	 * this function in indirect.c file.
682 	 */
683 
684 	map.m_lblk = es->es_lblk;
685 	map.m_len = es->es_len;
686 
687 	retval = ext4_ind_map_blocks(NULL, inode, &map, 0);
688 	if (retval > 0) {
689 		if (ext4_es_is_delayed(es) || ext4_es_is_hole(es)) {
690 			/*
691 			 * We want to add a delayed/hole extent but this
692 			 * block has been allocated.
693 			 */
694 			pr_warn("ES insert assertion failed for inode: %lu "
695 				"We can find blocks but we want to add a "
696 				"delayed/hole extent [%d/%d/%llu/%x]\n",
697 				inode->i_ino, es->es_lblk, es->es_len,
698 				ext4_es_pblock(es), ext4_es_status(es));
699 			return;
700 		} else if (ext4_es_is_written(es)) {
701 			if (retval != es->es_len) {
702 				pr_warn("ES insert assertion failed for "
703 					"inode: %lu retval %d != es_len %d\n",
704 					inode->i_ino, retval, es->es_len);
705 				return;
706 			}
707 			if (map.m_pblk != ext4_es_pblock(es)) {
708 				pr_warn("ES insert assertion failed for "
709 					"inode: %lu m_pblk %llu != "
710 					"es_pblk %llu\n",
711 					inode->i_ino, map.m_pblk,
712 					ext4_es_pblock(es));
713 				return;
714 			}
715 		} else {
716 			/*
717 			 * We don't need to check unwritten extent because
718 			 * indirect-based file doesn't have it.
719 			 */
720 			BUG();
721 		}
722 	} else if (retval == 0) {
723 		if (ext4_es_is_written(es)) {
724 			pr_warn("ES insert assertion failed for inode: %lu "
725 				"We can't find the block but we want to add "
726 				"a written extent [%d/%d/%llu/%x]\n",
727 				inode->i_ino, es->es_lblk, es->es_len,
728 				ext4_es_pblock(es), ext4_es_status(es));
729 			return;
730 		}
731 	}
732 }
733 
734 static inline void ext4_es_insert_extent_check(struct inode *inode,
735 					       struct extent_status *es)
736 {
737 	/*
738 	 * We don't need to worry about the race condition because
739 	 * caller takes i_data_sem locking.
740 	 */
741 	BUG_ON(!rwsem_is_locked(&EXT4_I(inode)->i_data_sem));
742 	if (ext4_test_inode_flag(inode, EXT4_INODE_EXTENTS))
743 		ext4_es_insert_extent_ext_check(inode, es);
744 	else
745 		ext4_es_insert_extent_ind_check(inode, es);
746 }
747 #else
748 static inline void ext4_es_insert_extent_check(struct inode *inode,
749 					       struct extent_status *es)
750 {
751 }
752 #endif
753 
754 static int __es_insert_extent(struct inode *inode, struct extent_status *newes)
755 {
756 	struct ext4_es_tree *tree = &EXT4_I(inode)->i_es_tree;
757 	struct rb_node **p = &tree->root.rb_node;
758 	struct rb_node *parent = NULL;
759 	struct extent_status *es;
760 
761 	while (*p) {
762 		parent = *p;
763 		es = rb_entry(parent, struct extent_status, rb_node);
764 
765 		if (newes->es_lblk < es->es_lblk) {
766 			if (ext4_es_can_be_merged(newes, es)) {
767 				/*
768 				 * Here we can modify es_lblk directly
769 				 * because it isn't overlapped.
770 				 */
771 				es->es_lblk = newes->es_lblk;
772 				es->es_len += newes->es_len;
773 				if (ext4_es_is_written(es) ||
774 				    ext4_es_is_unwritten(es))
775 					ext4_es_store_pblock(es,
776 							     newes->es_pblk);
777 				es = ext4_es_try_to_merge_left(inode, es);
778 				goto out;
779 			}
780 			p = &(*p)->rb_left;
781 		} else if (newes->es_lblk > ext4_es_end(es)) {
782 			if (ext4_es_can_be_merged(es, newes)) {
783 				es->es_len += newes->es_len;
784 				es = ext4_es_try_to_merge_right(inode, es);
785 				goto out;
786 			}
787 			p = &(*p)->rb_right;
788 		} else {
789 			BUG();
790 			return -EINVAL;
791 		}
792 	}
793 
794 	es = ext4_es_alloc_extent(inode, newes->es_lblk, newes->es_len,
795 				  newes->es_pblk);
796 	if (!es)
797 		return -ENOMEM;
798 	rb_link_node(&es->rb_node, parent, p);
799 	rb_insert_color(&es->rb_node, &tree->root);
800 
801 out:
802 	tree->cache_es = es;
803 	return 0;
804 }
805 
806 /*
807  * ext4_es_insert_extent() adds information to an inode's extent
808  * status tree.
809  *
810  * Return 0 on success, error code on failure.
811  */
812 int ext4_es_insert_extent(struct inode *inode, ext4_lblk_t lblk,
813 			  ext4_lblk_t len, ext4_fsblk_t pblk,
814 			  unsigned int status)
815 {
816 	struct extent_status newes;
817 	ext4_lblk_t end = lblk + len - 1;
818 	int err = 0;
819 	struct ext4_sb_info *sbi = EXT4_SB(inode->i_sb);
820 
821 	if (EXT4_SB(inode->i_sb)->s_mount_state & EXT4_FC_REPLAY)
822 		return 0;
823 
824 	es_debug("add [%u/%u) %llu %x to extent status tree of inode %lu\n",
825 		 lblk, len, pblk, status, inode->i_ino);
826 
827 	if (!len)
828 		return 0;
829 
830 	BUG_ON(end < lblk);
831 
832 	if ((status & EXTENT_STATUS_DELAYED) &&
833 	    (status & EXTENT_STATUS_WRITTEN)) {
834 		ext4_warning(inode->i_sb, "Inserting extent [%u/%u] as "
835 				" delayed and written which can potentially "
836 				" cause data loss.", lblk, len);
837 		WARN_ON(1);
838 	}
839 
840 	newes.es_lblk = lblk;
841 	newes.es_len = len;
842 	ext4_es_store_pblock_status(&newes, pblk, status);
843 	trace_ext4_es_insert_extent(inode, &newes);
844 
845 	ext4_es_insert_extent_check(inode, &newes);
846 
847 	write_lock(&EXT4_I(inode)->i_es_lock);
848 	err = __es_remove_extent(inode, lblk, end, NULL);
849 	if (err != 0)
850 		goto error;
851 retry:
852 	err = __es_insert_extent(inode, &newes);
853 	if (err == -ENOMEM && __es_shrink(EXT4_SB(inode->i_sb),
854 					  128, EXT4_I(inode)))
855 		goto retry;
856 	if (err == -ENOMEM && !ext4_es_is_delayed(&newes))
857 		err = 0;
858 
859 	if (sbi->s_cluster_ratio > 1 && test_opt(inode->i_sb, DELALLOC) &&
860 	    (status & EXTENT_STATUS_WRITTEN ||
861 	     status & EXTENT_STATUS_UNWRITTEN))
862 		__revise_pending(inode, lblk, len);
863 
864 error:
865 	write_unlock(&EXT4_I(inode)->i_es_lock);
866 
867 	ext4_es_print_tree(inode);
868 
869 	return err;
870 }
871 
872 /*
873  * ext4_es_cache_extent() inserts information into the extent status
874  * tree if and only if there isn't information about the range in
875  * question already.
876  */
877 void ext4_es_cache_extent(struct inode *inode, ext4_lblk_t lblk,
878 			  ext4_lblk_t len, ext4_fsblk_t pblk,
879 			  unsigned int status)
880 {
881 	struct extent_status *es;
882 	struct extent_status newes;
883 	ext4_lblk_t end = lblk + len - 1;
884 
885 	if (EXT4_SB(inode->i_sb)->s_mount_state & EXT4_FC_REPLAY)
886 		return;
887 
888 	newes.es_lblk = lblk;
889 	newes.es_len = len;
890 	ext4_es_store_pblock_status(&newes, pblk, status);
891 	trace_ext4_es_cache_extent(inode, &newes);
892 
893 	if (!len)
894 		return;
895 
896 	BUG_ON(end < lblk);
897 
898 	write_lock(&EXT4_I(inode)->i_es_lock);
899 
900 	es = __es_tree_search(&EXT4_I(inode)->i_es_tree.root, lblk);
901 	if (!es || es->es_lblk > end)
902 		__es_insert_extent(inode, &newes);
903 	write_unlock(&EXT4_I(inode)->i_es_lock);
904 }
905 
906 /*
907  * ext4_es_lookup_extent() looks up an extent in extent status tree.
908  *
909  * ext4_es_lookup_extent is called by ext4_map_blocks/ext4_da_map_blocks.
910  *
911  * Return: 1 on found, 0 on not
912  */
913 int ext4_es_lookup_extent(struct inode *inode, ext4_lblk_t lblk,
914 			  ext4_lblk_t *next_lblk,
915 			  struct extent_status *es)
916 {
917 	struct ext4_es_tree *tree;
918 	struct ext4_es_stats *stats;
919 	struct extent_status *es1 = NULL;
920 	struct rb_node *node;
921 	int found = 0;
922 
923 	if (EXT4_SB(inode->i_sb)->s_mount_state & EXT4_FC_REPLAY)
924 		return 0;
925 
926 	trace_ext4_es_lookup_extent_enter(inode, lblk);
927 	es_debug("lookup extent in block %u\n", lblk);
928 
929 	tree = &EXT4_I(inode)->i_es_tree;
930 	read_lock(&EXT4_I(inode)->i_es_lock);
931 
932 	/* find extent in cache firstly */
933 	es->es_lblk = es->es_len = es->es_pblk = 0;
934 	if (tree->cache_es) {
935 		es1 = tree->cache_es;
936 		if (in_range(lblk, es1->es_lblk, es1->es_len)) {
937 			es_debug("%u cached by [%u/%u)\n",
938 				 lblk, es1->es_lblk, es1->es_len);
939 			found = 1;
940 			goto out;
941 		}
942 	}
943 
944 	node = tree->root.rb_node;
945 	while (node) {
946 		es1 = rb_entry(node, struct extent_status, rb_node);
947 		if (lblk < es1->es_lblk)
948 			node = node->rb_left;
949 		else if (lblk > ext4_es_end(es1))
950 			node = node->rb_right;
951 		else {
952 			found = 1;
953 			break;
954 		}
955 	}
956 
957 out:
958 	stats = &EXT4_SB(inode->i_sb)->s_es_stats;
959 	if (found) {
960 		BUG_ON(!es1);
961 		es->es_lblk = es1->es_lblk;
962 		es->es_len = es1->es_len;
963 		es->es_pblk = es1->es_pblk;
964 		if (!ext4_es_is_referenced(es1))
965 			ext4_es_set_referenced(es1);
966 		percpu_counter_inc(&stats->es_stats_cache_hits);
967 		if (next_lblk) {
968 			node = rb_next(&es1->rb_node);
969 			if (node) {
970 				es1 = rb_entry(node, struct extent_status,
971 					       rb_node);
972 				*next_lblk = es1->es_lblk;
973 			} else
974 				*next_lblk = 0;
975 		}
976 	} else {
977 		percpu_counter_inc(&stats->es_stats_cache_misses);
978 	}
979 
980 	read_unlock(&EXT4_I(inode)->i_es_lock);
981 
982 	trace_ext4_es_lookup_extent_exit(inode, es, found);
983 	return found;
984 }
985 
986 struct rsvd_count {
987 	int ndelonly;
988 	bool first_do_lblk_found;
989 	ext4_lblk_t first_do_lblk;
990 	ext4_lblk_t last_do_lblk;
991 	struct extent_status *left_es;
992 	bool partial;
993 	ext4_lblk_t lclu;
994 };
995 
996 /*
997  * init_rsvd - initialize reserved count data before removing block range
998  *	       in file from extent status tree
999  *
1000  * @inode - file containing range
1001  * @lblk - first block in range
1002  * @es - pointer to first extent in range
1003  * @rc - pointer to reserved count data
1004  *
1005  * Assumes es is not NULL
1006  */
1007 static void init_rsvd(struct inode *inode, ext4_lblk_t lblk,
1008 		      struct extent_status *es, struct rsvd_count *rc)
1009 {
1010 	struct ext4_sb_info *sbi = EXT4_SB(inode->i_sb);
1011 	struct rb_node *node;
1012 
1013 	rc->ndelonly = 0;
1014 
1015 	/*
1016 	 * for bigalloc, note the first delonly block in the range has not
1017 	 * been found, record the extent containing the block to the left of
1018 	 * the region to be removed, if any, and note that there's no partial
1019 	 * cluster to track
1020 	 */
1021 	if (sbi->s_cluster_ratio > 1) {
1022 		rc->first_do_lblk_found = false;
1023 		if (lblk > es->es_lblk) {
1024 			rc->left_es = es;
1025 		} else {
1026 			node = rb_prev(&es->rb_node);
1027 			rc->left_es = node ? rb_entry(node,
1028 						      struct extent_status,
1029 						      rb_node) : NULL;
1030 		}
1031 		rc->partial = false;
1032 	}
1033 }
1034 
1035 /*
1036  * count_rsvd - count the clusters containing delayed and not unwritten
1037  *		(delonly) blocks in a range within an extent and add to
1038  *	        the running tally in rsvd_count
1039  *
1040  * @inode - file containing extent
1041  * @lblk - first block in range
1042  * @len - length of range in blocks
1043  * @es - pointer to extent containing clusters to be counted
1044  * @rc - pointer to reserved count data
1045  *
1046  * Tracks partial clusters found at the beginning and end of extents so
1047  * they aren't overcounted when they span adjacent extents
1048  */
1049 static void count_rsvd(struct inode *inode, ext4_lblk_t lblk, long len,
1050 		       struct extent_status *es, struct rsvd_count *rc)
1051 {
1052 	struct ext4_sb_info *sbi = EXT4_SB(inode->i_sb);
1053 	ext4_lblk_t i, end, nclu;
1054 
1055 	if (!ext4_es_is_delonly(es))
1056 		return;
1057 
1058 	WARN_ON(len <= 0);
1059 
1060 	if (sbi->s_cluster_ratio == 1) {
1061 		rc->ndelonly += (int) len;
1062 		return;
1063 	}
1064 
1065 	/* bigalloc */
1066 
1067 	i = (lblk < es->es_lblk) ? es->es_lblk : lblk;
1068 	end = lblk + (ext4_lblk_t) len - 1;
1069 	end = (end > ext4_es_end(es)) ? ext4_es_end(es) : end;
1070 
1071 	/* record the first block of the first delonly extent seen */
1072 	if (!rc->first_do_lblk_found) {
1073 		rc->first_do_lblk = i;
1074 		rc->first_do_lblk_found = true;
1075 	}
1076 
1077 	/* update the last lblk in the region seen so far */
1078 	rc->last_do_lblk = end;
1079 
1080 	/*
1081 	 * if we're tracking a partial cluster and the current extent
1082 	 * doesn't start with it, count it and stop tracking
1083 	 */
1084 	if (rc->partial && (rc->lclu != EXT4_B2C(sbi, i))) {
1085 		rc->ndelonly++;
1086 		rc->partial = false;
1087 	}
1088 
1089 	/*
1090 	 * if the first cluster doesn't start on a cluster boundary but
1091 	 * ends on one, count it
1092 	 */
1093 	if (EXT4_LBLK_COFF(sbi, i) != 0) {
1094 		if (end >= EXT4_LBLK_CFILL(sbi, i)) {
1095 			rc->ndelonly++;
1096 			rc->partial = false;
1097 			i = EXT4_LBLK_CFILL(sbi, i) + 1;
1098 		}
1099 	}
1100 
1101 	/*
1102 	 * if the current cluster starts on a cluster boundary, count the
1103 	 * number of whole delonly clusters in the extent
1104 	 */
1105 	if ((i + sbi->s_cluster_ratio - 1) <= end) {
1106 		nclu = (end - i + 1) >> sbi->s_cluster_bits;
1107 		rc->ndelonly += nclu;
1108 		i += nclu << sbi->s_cluster_bits;
1109 	}
1110 
1111 	/*
1112 	 * start tracking a partial cluster if there's a partial at the end
1113 	 * of the current extent and we're not already tracking one
1114 	 */
1115 	if (!rc->partial && i <= end) {
1116 		rc->partial = true;
1117 		rc->lclu = EXT4_B2C(sbi, i);
1118 	}
1119 }
1120 
1121 /*
1122  * __pr_tree_search - search for a pending cluster reservation
1123  *
1124  * @root - root of pending reservation tree
1125  * @lclu - logical cluster to search for
1126  *
1127  * Returns the pending reservation for the cluster identified by @lclu
1128  * if found.  If not, returns a reservation for the next cluster if any,
1129  * and if not, returns NULL.
1130  */
1131 static struct pending_reservation *__pr_tree_search(struct rb_root *root,
1132 						    ext4_lblk_t lclu)
1133 {
1134 	struct rb_node *node = root->rb_node;
1135 	struct pending_reservation *pr = NULL;
1136 
1137 	while (node) {
1138 		pr = rb_entry(node, struct pending_reservation, rb_node);
1139 		if (lclu < pr->lclu)
1140 			node = node->rb_left;
1141 		else if (lclu > pr->lclu)
1142 			node = node->rb_right;
1143 		else
1144 			return pr;
1145 	}
1146 	if (pr && lclu < pr->lclu)
1147 		return pr;
1148 	if (pr && lclu > pr->lclu) {
1149 		node = rb_next(&pr->rb_node);
1150 		return node ? rb_entry(node, struct pending_reservation,
1151 				       rb_node) : NULL;
1152 	}
1153 	return NULL;
1154 }
1155 
1156 /*
1157  * get_rsvd - calculates and returns the number of cluster reservations to be
1158  *	      released when removing a block range from the extent status tree
1159  *	      and releases any pending reservations within the range
1160  *
1161  * @inode - file containing block range
1162  * @end - last block in range
1163  * @right_es - pointer to extent containing next block beyond end or NULL
1164  * @rc - pointer to reserved count data
1165  *
1166  * The number of reservations to be released is equal to the number of
1167  * clusters containing delayed and not unwritten (delonly) blocks within
1168  * the range, minus the number of clusters still containing delonly blocks
1169  * at the ends of the range, and minus the number of pending reservations
1170  * within the range.
1171  */
1172 static unsigned int get_rsvd(struct inode *inode, ext4_lblk_t end,
1173 			     struct extent_status *right_es,
1174 			     struct rsvd_count *rc)
1175 {
1176 	struct ext4_sb_info *sbi = EXT4_SB(inode->i_sb);
1177 	struct pending_reservation *pr;
1178 	struct ext4_pending_tree *tree = &EXT4_I(inode)->i_pending_tree;
1179 	struct rb_node *node;
1180 	ext4_lblk_t first_lclu, last_lclu;
1181 	bool left_delonly, right_delonly, count_pending;
1182 	struct extent_status *es;
1183 
1184 	if (sbi->s_cluster_ratio > 1) {
1185 		/* count any remaining partial cluster */
1186 		if (rc->partial)
1187 			rc->ndelonly++;
1188 
1189 		if (rc->ndelonly == 0)
1190 			return 0;
1191 
1192 		first_lclu = EXT4_B2C(sbi, rc->first_do_lblk);
1193 		last_lclu = EXT4_B2C(sbi, rc->last_do_lblk);
1194 
1195 		/*
1196 		 * decrease the delonly count by the number of clusters at the
1197 		 * ends of the range that still contain delonly blocks -
1198 		 * these clusters still need to be reserved
1199 		 */
1200 		left_delonly = right_delonly = false;
1201 
1202 		es = rc->left_es;
1203 		while (es && ext4_es_end(es) >=
1204 		       EXT4_LBLK_CMASK(sbi, rc->first_do_lblk)) {
1205 			if (ext4_es_is_delonly(es)) {
1206 				rc->ndelonly--;
1207 				left_delonly = true;
1208 				break;
1209 			}
1210 			node = rb_prev(&es->rb_node);
1211 			if (!node)
1212 				break;
1213 			es = rb_entry(node, struct extent_status, rb_node);
1214 		}
1215 		if (right_es && (!left_delonly || first_lclu != last_lclu)) {
1216 			if (end < ext4_es_end(right_es)) {
1217 				es = right_es;
1218 			} else {
1219 				node = rb_next(&right_es->rb_node);
1220 				es = node ? rb_entry(node, struct extent_status,
1221 						     rb_node) : NULL;
1222 			}
1223 			while (es && es->es_lblk <=
1224 			       EXT4_LBLK_CFILL(sbi, rc->last_do_lblk)) {
1225 				if (ext4_es_is_delonly(es)) {
1226 					rc->ndelonly--;
1227 					right_delonly = true;
1228 					break;
1229 				}
1230 				node = rb_next(&es->rb_node);
1231 				if (!node)
1232 					break;
1233 				es = rb_entry(node, struct extent_status,
1234 					      rb_node);
1235 			}
1236 		}
1237 
1238 		/*
1239 		 * Determine the block range that should be searched for
1240 		 * pending reservations, if any.  Clusters on the ends of the
1241 		 * original removed range containing delonly blocks are
1242 		 * excluded.  They've already been accounted for and it's not
1243 		 * possible to determine if an associated pending reservation
1244 		 * should be released with the information available in the
1245 		 * extents status tree.
1246 		 */
1247 		if (first_lclu == last_lclu) {
1248 			if (left_delonly | right_delonly)
1249 				count_pending = false;
1250 			else
1251 				count_pending = true;
1252 		} else {
1253 			if (left_delonly)
1254 				first_lclu++;
1255 			if (right_delonly)
1256 				last_lclu--;
1257 			if (first_lclu <= last_lclu)
1258 				count_pending = true;
1259 			else
1260 				count_pending = false;
1261 		}
1262 
1263 		/*
1264 		 * a pending reservation found between first_lclu and last_lclu
1265 		 * represents an allocated cluster that contained at least one
1266 		 * delonly block, so the delonly total must be reduced by one
1267 		 * for each pending reservation found and released
1268 		 */
1269 		if (count_pending) {
1270 			pr = __pr_tree_search(&tree->root, first_lclu);
1271 			while (pr && pr->lclu <= last_lclu) {
1272 				rc->ndelonly--;
1273 				node = rb_next(&pr->rb_node);
1274 				rb_erase(&pr->rb_node, &tree->root);
1275 				kmem_cache_free(ext4_pending_cachep, pr);
1276 				if (!node)
1277 					break;
1278 				pr = rb_entry(node, struct pending_reservation,
1279 					      rb_node);
1280 			}
1281 		}
1282 	}
1283 	return rc->ndelonly;
1284 }
1285 
1286 
1287 /*
1288  * __es_remove_extent - removes block range from extent status tree
1289  *
1290  * @inode - file containing range
1291  * @lblk - first block in range
1292  * @end - last block in range
1293  * @reserved - number of cluster reservations released
1294  *
1295  * If @reserved is not NULL and delayed allocation is enabled, counts
1296  * block/cluster reservations freed by removing range and if bigalloc
1297  * enabled cancels pending reservations as needed. Returns 0 on success,
1298  * error code on failure.
1299  */
1300 static int __es_remove_extent(struct inode *inode, ext4_lblk_t lblk,
1301 			      ext4_lblk_t end, int *reserved)
1302 {
1303 	struct ext4_es_tree *tree = &EXT4_I(inode)->i_es_tree;
1304 	struct rb_node *node;
1305 	struct extent_status *es;
1306 	struct extent_status orig_es;
1307 	ext4_lblk_t len1, len2;
1308 	ext4_fsblk_t block;
1309 	int err;
1310 	bool count_reserved = true;
1311 	struct rsvd_count rc;
1312 
1313 	if (reserved == NULL || !test_opt(inode->i_sb, DELALLOC))
1314 		count_reserved = false;
1315 retry:
1316 	err = 0;
1317 
1318 	es = __es_tree_search(&tree->root, lblk);
1319 	if (!es)
1320 		goto out;
1321 	if (es->es_lblk > end)
1322 		goto out;
1323 
1324 	/* Simply invalidate cache_es. */
1325 	tree->cache_es = NULL;
1326 	if (count_reserved)
1327 		init_rsvd(inode, lblk, es, &rc);
1328 
1329 	orig_es.es_lblk = es->es_lblk;
1330 	orig_es.es_len = es->es_len;
1331 	orig_es.es_pblk = es->es_pblk;
1332 
1333 	len1 = lblk > es->es_lblk ? lblk - es->es_lblk : 0;
1334 	len2 = ext4_es_end(es) > end ? ext4_es_end(es) - end : 0;
1335 	if (len1 > 0)
1336 		es->es_len = len1;
1337 	if (len2 > 0) {
1338 		if (len1 > 0) {
1339 			struct extent_status newes;
1340 
1341 			newes.es_lblk = end + 1;
1342 			newes.es_len = len2;
1343 			block = 0x7FDEADBEEFULL;
1344 			if (ext4_es_is_written(&orig_es) ||
1345 			    ext4_es_is_unwritten(&orig_es))
1346 				block = ext4_es_pblock(&orig_es) +
1347 					orig_es.es_len - len2;
1348 			ext4_es_store_pblock_status(&newes, block,
1349 						    ext4_es_status(&orig_es));
1350 			err = __es_insert_extent(inode, &newes);
1351 			if (err) {
1352 				es->es_lblk = orig_es.es_lblk;
1353 				es->es_len = orig_es.es_len;
1354 				if ((err == -ENOMEM) &&
1355 				    __es_shrink(EXT4_SB(inode->i_sb),
1356 							128, EXT4_I(inode)))
1357 					goto retry;
1358 				goto out;
1359 			}
1360 		} else {
1361 			es->es_lblk = end + 1;
1362 			es->es_len = len2;
1363 			if (ext4_es_is_written(es) ||
1364 			    ext4_es_is_unwritten(es)) {
1365 				block = orig_es.es_pblk + orig_es.es_len - len2;
1366 				ext4_es_store_pblock(es, block);
1367 			}
1368 		}
1369 		if (count_reserved)
1370 			count_rsvd(inode, lblk, orig_es.es_len - len1 - len2,
1371 				   &orig_es, &rc);
1372 		goto out_get_reserved;
1373 	}
1374 
1375 	if (len1 > 0) {
1376 		if (count_reserved)
1377 			count_rsvd(inode, lblk, orig_es.es_len - len1,
1378 				   &orig_es, &rc);
1379 		node = rb_next(&es->rb_node);
1380 		if (node)
1381 			es = rb_entry(node, struct extent_status, rb_node);
1382 		else
1383 			es = NULL;
1384 	}
1385 
1386 	while (es && ext4_es_end(es) <= end) {
1387 		if (count_reserved)
1388 			count_rsvd(inode, es->es_lblk, es->es_len, es, &rc);
1389 		node = rb_next(&es->rb_node);
1390 		rb_erase(&es->rb_node, &tree->root);
1391 		ext4_es_free_extent(inode, es);
1392 		if (!node) {
1393 			es = NULL;
1394 			break;
1395 		}
1396 		es = rb_entry(node, struct extent_status, rb_node);
1397 	}
1398 
1399 	if (es && es->es_lblk < end + 1) {
1400 		ext4_lblk_t orig_len = es->es_len;
1401 
1402 		len1 = ext4_es_end(es) - end;
1403 		if (count_reserved)
1404 			count_rsvd(inode, es->es_lblk, orig_len - len1,
1405 				   es, &rc);
1406 		es->es_lblk = end + 1;
1407 		es->es_len = len1;
1408 		if (ext4_es_is_written(es) || ext4_es_is_unwritten(es)) {
1409 			block = es->es_pblk + orig_len - len1;
1410 			ext4_es_store_pblock(es, block);
1411 		}
1412 	}
1413 
1414 out_get_reserved:
1415 	if (count_reserved)
1416 		*reserved = get_rsvd(inode, end, es, &rc);
1417 out:
1418 	return err;
1419 }
1420 
1421 /*
1422  * ext4_es_remove_extent - removes block range from extent status tree
1423  *
1424  * @inode - file containing range
1425  * @lblk - first block in range
1426  * @len - number of blocks to remove
1427  *
1428  * Reduces block/cluster reservation count and for bigalloc cancels pending
1429  * reservations as needed. Returns 0 on success, error code on failure.
1430  */
1431 int ext4_es_remove_extent(struct inode *inode, ext4_lblk_t lblk,
1432 			  ext4_lblk_t len)
1433 {
1434 	ext4_lblk_t end;
1435 	int err = 0;
1436 	int reserved = 0;
1437 
1438 	if (EXT4_SB(inode->i_sb)->s_mount_state & EXT4_FC_REPLAY)
1439 		return 0;
1440 
1441 	trace_ext4_es_remove_extent(inode, lblk, len);
1442 	es_debug("remove [%u/%u) from extent status tree of inode %lu\n",
1443 		 lblk, len, inode->i_ino);
1444 
1445 	if (!len)
1446 		return err;
1447 
1448 	end = lblk + len - 1;
1449 	BUG_ON(end < lblk);
1450 
1451 	/*
1452 	 * ext4_clear_inode() depends on us taking i_es_lock unconditionally
1453 	 * so that we are sure __es_shrink() is done with the inode before it
1454 	 * is reclaimed.
1455 	 */
1456 	write_lock(&EXT4_I(inode)->i_es_lock);
1457 	err = __es_remove_extent(inode, lblk, end, &reserved);
1458 	write_unlock(&EXT4_I(inode)->i_es_lock);
1459 	ext4_es_print_tree(inode);
1460 	ext4_da_release_space(inode, reserved);
1461 	return err;
1462 }
1463 
1464 static int __es_shrink(struct ext4_sb_info *sbi, int nr_to_scan,
1465 		       struct ext4_inode_info *locked_ei)
1466 {
1467 	struct ext4_inode_info *ei;
1468 	struct ext4_es_stats *es_stats;
1469 	ktime_t start_time;
1470 	u64 scan_time;
1471 	int nr_to_walk;
1472 	int nr_shrunk = 0;
1473 	int retried = 0, nr_skipped = 0;
1474 
1475 	es_stats = &sbi->s_es_stats;
1476 	start_time = ktime_get();
1477 
1478 retry:
1479 	spin_lock(&sbi->s_es_lock);
1480 	nr_to_walk = sbi->s_es_nr_inode;
1481 	while (nr_to_walk-- > 0) {
1482 		if (list_empty(&sbi->s_es_list)) {
1483 			spin_unlock(&sbi->s_es_lock);
1484 			goto out;
1485 		}
1486 		ei = list_first_entry(&sbi->s_es_list, struct ext4_inode_info,
1487 				      i_es_list);
1488 		/* Move the inode to the tail */
1489 		list_move_tail(&ei->i_es_list, &sbi->s_es_list);
1490 
1491 		/*
1492 		 * Normally we try hard to avoid shrinking precached inodes,
1493 		 * but we will as a last resort.
1494 		 */
1495 		if (!retried && ext4_test_inode_state(&ei->vfs_inode,
1496 						EXT4_STATE_EXT_PRECACHED)) {
1497 			nr_skipped++;
1498 			continue;
1499 		}
1500 
1501 		if (ei == locked_ei || !write_trylock(&ei->i_es_lock)) {
1502 			nr_skipped++;
1503 			continue;
1504 		}
1505 		/*
1506 		 * Now we hold i_es_lock which protects us from inode reclaim
1507 		 * freeing inode under us
1508 		 */
1509 		spin_unlock(&sbi->s_es_lock);
1510 
1511 		nr_shrunk += es_reclaim_extents(ei, &nr_to_scan);
1512 		write_unlock(&ei->i_es_lock);
1513 
1514 		if (nr_to_scan <= 0)
1515 			goto out;
1516 		spin_lock(&sbi->s_es_lock);
1517 	}
1518 	spin_unlock(&sbi->s_es_lock);
1519 
1520 	/*
1521 	 * If we skipped any inodes, and we weren't able to make any
1522 	 * forward progress, try again to scan precached inodes.
1523 	 */
1524 	if ((nr_shrunk == 0) && nr_skipped && !retried) {
1525 		retried++;
1526 		goto retry;
1527 	}
1528 
1529 	if (locked_ei && nr_shrunk == 0)
1530 		nr_shrunk = es_reclaim_extents(locked_ei, &nr_to_scan);
1531 
1532 out:
1533 	scan_time = ktime_to_ns(ktime_sub(ktime_get(), start_time));
1534 	if (likely(es_stats->es_stats_scan_time))
1535 		es_stats->es_stats_scan_time = (scan_time +
1536 				es_stats->es_stats_scan_time*3) / 4;
1537 	else
1538 		es_stats->es_stats_scan_time = scan_time;
1539 	if (scan_time > es_stats->es_stats_max_scan_time)
1540 		es_stats->es_stats_max_scan_time = scan_time;
1541 	if (likely(es_stats->es_stats_shrunk))
1542 		es_stats->es_stats_shrunk = (nr_shrunk +
1543 				es_stats->es_stats_shrunk*3) / 4;
1544 	else
1545 		es_stats->es_stats_shrunk = nr_shrunk;
1546 
1547 	trace_ext4_es_shrink(sbi->s_sb, nr_shrunk, scan_time,
1548 			     nr_skipped, retried);
1549 	return nr_shrunk;
1550 }
1551 
1552 static unsigned long ext4_es_count(struct shrinker *shrink,
1553 				   struct shrink_control *sc)
1554 {
1555 	unsigned long nr;
1556 	struct ext4_sb_info *sbi;
1557 
1558 	sbi = container_of(shrink, struct ext4_sb_info, s_es_shrinker);
1559 	nr = percpu_counter_read_positive(&sbi->s_es_stats.es_stats_shk_cnt);
1560 	trace_ext4_es_shrink_count(sbi->s_sb, sc->nr_to_scan, nr);
1561 	return nr;
1562 }
1563 
1564 static unsigned long ext4_es_scan(struct shrinker *shrink,
1565 				  struct shrink_control *sc)
1566 {
1567 	struct ext4_sb_info *sbi = container_of(shrink,
1568 					struct ext4_sb_info, s_es_shrinker);
1569 	int nr_to_scan = sc->nr_to_scan;
1570 	int ret, nr_shrunk;
1571 
1572 	ret = percpu_counter_read_positive(&sbi->s_es_stats.es_stats_shk_cnt);
1573 	trace_ext4_es_shrink_scan_enter(sbi->s_sb, nr_to_scan, ret);
1574 
1575 	nr_shrunk = __es_shrink(sbi, nr_to_scan, NULL);
1576 
1577 	ret = percpu_counter_read_positive(&sbi->s_es_stats.es_stats_shk_cnt);
1578 	trace_ext4_es_shrink_scan_exit(sbi->s_sb, nr_shrunk, ret);
1579 	return nr_shrunk;
1580 }
1581 
1582 int ext4_seq_es_shrinker_info_show(struct seq_file *seq, void *v)
1583 {
1584 	struct ext4_sb_info *sbi = EXT4_SB((struct super_block *) seq->private);
1585 	struct ext4_es_stats *es_stats = &sbi->s_es_stats;
1586 	struct ext4_inode_info *ei, *max = NULL;
1587 	unsigned int inode_cnt = 0;
1588 
1589 	if (v != SEQ_START_TOKEN)
1590 		return 0;
1591 
1592 	/* here we just find an inode that has the max nr. of objects */
1593 	spin_lock(&sbi->s_es_lock);
1594 	list_for_each_entry(ei, &sbi->s_es_list, i_es_list) {
1595 		inode_cnt++;
1596 		if (max && max->i_es_all_nr < ei->i_es_all_nr)
1597 			max = ei;
1598 		else if (!max)
1599 			max = ei;
1600 	}
1601 	spin_unlock(&sbi->s_es_lock);
1602 
1603 	seq_printf(seq, "stats:\n  %lld objects\n  %lld reclaimable objects\n",
1604 		   percpu_counter_sum_positive(&es_stats->es_stats_all_cnt),
1605 		   percpu_counter_sum_positive(&es_stats->es_stats_shk_cnt));
1606 	seq_printf(seq, "  %lld/%lld cache hits/misses\n",
1607 		   percpu_counter_sum_positive(&es_stats->es_stats_cache_hits),
1608 		   percpu_counter_sum_positive(&es_stats->es_stats_cache_misses));
1609 	if (inode_cnt)
1610 		seq_printf(seq, "  %d inodes on list\n", inode_cnt);
1611 
1612 	seq_printf(seq, "average:\n  %llu us scan time\n",
1613 	    div_u64(es_stats->es_stats_scan_time, 1000));
1614 	seq_printf(seq, "  %lu shrunk objects\n", es_stats->es_stats_shrunk);
1615 	if (inode_cnt)
1616 		seq_printf(seq,
1617 		    "maximum:\n  %lu inode (%u objects, %u reclaimable)\n"
1618 		    "  %llu us max scan time\n",
1619 		    max->vfs_inode.i_ino, max->i_es_all_nr, max->i_es_shk_nr,
1620 		    div_u64(es_stats->es_stats_max_scan_time, 1000));
1621 
1622 	return 0;
1623 }
1624 
1625 int ext4_es_register_shrinker(struct ext4_sb_info *sbi)
1626 {
1627 	int err;
1628 
1629 	/* Make sure we have enough bits for physical block number */
1630 	BUILD_BUG_ON(ES_SHIFT < 48);
1631 	INIT_LIST_HEAD(&sbi->s_es_list);
1632 	sbi->s_es_nr_inode = 0;
1633 	spin_lock_init(&sbi->s_es_lock);
1634 	sbi->s_es_stats.es_stats_shrunk = 0;
1635 	err = percpu_counter_init(&sbi->s_es_stats.es_stats_cache_hits, 0,
1636 				  GFP_KERNEL);
1637 	if (err)
1638 		return err;
1639 	err = percpu_counter_init(&sbi->s_es_stats.es_stats_cache_misses, 0,
1640 				  GFP_KERNEL);
1641 	if (err)
1642 		goto err1;
1643 	sbi->s_es_stats.es_stats_scan_time = 0;
1644 	sbi->s_es_stats.es_stats_max_scan_time = 0;
1645 	err = percpu_counter_init(&sbi->s_es_stats.es_stats_all_cnt, 0, GFP_KERNEL);
1646 	if (err)
1647 		goto err2;
1648 	err = percpu_counter_init(&sbi->s_es_stats.es_stats_shk_cnt, 0, GFP_KERNEL);
1649 	if (err)
1650 		goto err3;
1651 
1652 	sbi->s_es_shrinker.scan_objects = ext4_es_scan;
1653 	sbi->s_es_shrinker.count_objects = ext4_es_count;
1654 	sbi->s_es_shrinker.seeks = DEFAULT_SEEKS;
1655 	err = register_shrinker(&sbi->s_es_shrinker, "ext4-es:%s",
1656 				sbi->s_sb->s_id);
1657 	if (err)
1658 		goto err4;
1659 
1660 	return 0;
1661 err4:
1662 	percpu_counter_destroy(&sbi->s_es_stats.es_stats_shk_cnt);
1663 err3:
1664 	percpu_counter_destroy(&sbi->s_es_stats.es_stats_all_cnt);
1665 err2:
1666 	percpu_counter_destroy(&sbi->s_es_stats.es_stats_cache_misses);
1667 err1:
1668 	percpu_counter_destroy(&sbi->s_es_stats.es_stats_cache_hits);
1669 	return err;
1670 }
1671 
1672 void ext4_es_unregister_shrinker(struct ext4_sb_info *sbi)
1673 {
1674 	percpu_counter_destroy(&sbi->s_es_stats.es_stats_cache_hits);
1675 	percpu_counter_destroy(&sbi->s_es_stats.es_stats_cache_misses);
1676 	percpu_counter_destroy(&sbi->s_es_stats.es_stats_all_cnt);
1677 	percpu_counter_destroy(&sbi->s_es_stats.es_stats_shk_cnt);
1678 	unregister_shrinker(&sbi->s_es_shrinker);
1679 }
1680 
1681 /*
1682  * Shrink extents in given inode from ei->i_es_shrink_lblk till end. Scan at
1683  * most *nr_to_scan extents, update *nr_to_scan accordingly.
1684  *
1685  * Return 0 if we hit end of tree / interval, 1 if we exhausted nr_to_scan.
1686  * Increment *nr_shrunk by the number of reclaimed extents. Also update
1687  * ei->i_es_shrink_lblk to where we should continue scanning.
1688  */
1689 static int es_do_reclaim_extents(struct ext4_inode_info *ei, ext4_lblk_t end,
1690 				 int *nr_to_scan, int *nr_shrunk)
1691 {
1692 	struct inode *inode = &ei->vfs_inode;
1693 	struct ext4_es_tree *tree = &ei->i_es_tree;
1694 	struct extent_status *es;
1695 	struct rb_node *node;
1696 
1697 	es = __es_tree_search(&tree->root, ei->i_es_shrink_lblk);
1698 	if (!es)
1699 		goto out_wrap;
1700 
1701 	while (*nr_to_scan > 0) {
1702 		if (es->es_lblk > end) {
1703 			ei->i_es_shrink_lblk = end + 1;
1704 			return 0;
1705 		}
1706 
1707 		(*nr_to_scan)--;
1708 		node = rb_next(&es->rb_node);
1709 		/*
1710 		 * We can't reclaim delayed extent from status tree because
1711 		 * fiemap, bigallic, and seek_data/hole need to use it.
1712 		 */
1713 		if (ext4_es_is_delayed(es))
1714 			goto next;
1715 		if (ext4_es_is_referenced(es)) {
1716 			ext4_es_clear_referenced(es);
1717 			goto next;
1718 		}
1719 
1720 		rb_erase(&es->rb_node, &tree->root);
1721 		ext4_es_free_extent(inode, es);
1722 		(*nr_shrunk)++;
1723 next:
1724 		if (!node)
1725 			goto out_wrap;
1726 		es = rb_entry(node, struct extent_status, rb_node);
1727 	}
1728 	ei->i_es_shrink_lblk = es->es_lblk;
1729 	return 1;
1730 out_wrap:
1731 	ei->i_es_shrink_lblk = 0;
1732 	return 0;
1733 }
1734 
1735 static int es_reclaim_extents(struct ext4_inode_info *ei, int *nr_to_scan)
1736 {
1737 	struct inode *inode = &ei->vfs_inode;
1738 	int nr_shrunk = 0;
1739 	ext4_lblk_t start = ei->i_es_shrink_lblk;
1740 	static DEFINE_RATELIMIT_STATE(_rs, DEFAULT_RATELIMIT_INTERVAL,
1741 				      DEFAULT_RATELIMIT_BURST);
1742 
1743 	if (ei->i_es_shk_nr == 0)
1744 		return 0;
1745 
1746 	if (ext4_test_inode_state(inode, EXT4_STATE_EXT_PRECACHED) &&
1747 	    __ratelimit(&_rs))
1748 		ext4_warning(inode->i_sb, "forced shrink of precached extents");
1749 
1750 	if (!es_do_reclaim_extents(ei, EXT_MAX_BLOCKS, nr_to_scan, &nr_shrunk) &&
1751 	    start != 0)
1752 		es_do_reclaim_extents(ei, start - 1, nr_to_scan, &nr_shrunk);
1753 
1754 	ei->i_es_tree.cache_es = NULL;
1755 	return nr_shrunk;
1756 }
1757 
1758 /*
1759  * Called to support EXT4_IOC_CLEAR_ES_CACHE.  We can only remove
1760  * discretionary entries from the extent status cache.  (Some entries
1761  * must be present for proper operations.)
1762  */
1763 void ext4_clear_inode_es(struct inode *inode)
1764 {
1765 	struct ext4_inode_info *ei = EXT4_I(inode);
1766 	struct extent_status *es;
1767 	struct ext4_es_tree *tree;
1768 	struct rb_node *node;
1769 
1770 	write_lock(&ei->i_es_lock);
1771 	tree = &EXT4_I(inode)->i_es_tree;
1772 	tree->cache_es = NULL;
1773 	node = rb_first(&tree->root);
1774 	while (node) {
1775 		es = rb_entry(node, struct extent_status, rb_node);
1776 		node = rb_next(node);
1777 		if (!ext4_es_is_delayed(es)) {
1778 			rb_erase(&es->rb_node, &tree->root);
1779 			ext4_es_free_extent(inode, es);
1780 		}
1781 	}
1782 	ext4_clear_inode_state(inode, EXT4_STATE_EXT_PRECACHED);
1783 	write_unlock(&ei->i_es_lock);
1784 }
1785 
1786 #ifdef ES_DEBUG__
1787 static void ext4_print_pending_tree(struct inode *inode)
1788 {
1789 	struct ext4_pending_tree *tree;
1790 	struct rb_node *node;
1791 	struct pending_reservation *pr;
1792 
1793 	printk(KERN_DEBUG "pending reservations for inode %lu:", inode->i_ino);
1794 	tree = &EXT4_I(inode)->i_pending_tree;
1795 	node = rb_first(&tree->root);
1796 	while (node) {
1797 		pr = rb_entry(node, struct pending_reservation, rb_node);
1798 		printk(KERN_DEBUG " %u", pr->lclu);
1799 		node = rb_next(node);
1800 	}
1801 	printk(KERN_DEBUG "\n");
1802 }
1803 #else
1804 #define ext4_print_pending_tree(inode)
1805 #endif
1806 
1807 int __init ext4_init_pending(void)
1808 {
1809 	ext4_pending_cachep = KMEM_CACHE(pending_reservation, SLAB_RECLAIM_ACCOUNT);
1810 	if (ext4_pending_cachep == NULL)
1811 		return -ENOMEM;
1812 	return 0;
1813 }
1814 
1815 void ext4_exit_pending(void)
1816 {
1817 	kmem_cache_destroy(ext4_pending_cachep);
1818 }
1819 
1820 void ext4_init_pending_tree(struct ext4_pending_tree *tree)
1821 {
1822 	tree->root = RB_ROOT;
1823 }
1824 
1825 /*
1826  * __get_pending - retrieve a pointer to a pending reservation
1827  *
1828  * @inode - file containing the pending cluster reservation
1829  * @lclu - logical cluster of interest
1830  *
1831  * Returns a pointer to a pending reservation if it's a member of
1832  * the set, and NULL if not.  Must be called holding i_es_lock.
1833  */
1834 static struct pending_reservation *__get_pending(struct inode *inode,
1835 						 ext4_lblk_t lclu)
1836 {
1837 	struct ext4_pending_tree *tree;
1838 	struct rb_node *node;
1839 	struct pending_reservation *pr = NULL;
1840 
1841 	tree = &EXT4_I(inode)->i_pending_tree;
1842 	node = (&tree->root)->rb_node;
1843 
1844 	while (node) {
1845 		pr = rb_entry(node, struct pending_reservation, rb_node);
1846 		if (lclu < pr->lclu)
1847 			node = node->rb_left;
1848 		else if (lclu > pr->lclu)
1849 			node = node->rb_right;
1850 		else if (lclu == pr->lclu)
1851 			return pr;
1852 	}
1853 	return NULL;
1854 }
1855 
1856 /*
1857  * __insert_pending - adds a pending cluster reservation to the set of
1858  *                    pending reservations
1859  *
1860  * @inode - file containing the cluster
1861  * @lblk - logical block in the cluster to be added
1862  *
1863  * Returns 0 on successful insertion and -ENOMEM on failure.  If the
1864  * pending reservation is already in the set, returns successfully.
1865  */
1866 static int __insert_pending(struct inode *inode, ext4_lblk_t lblk)
1867 {
1868 	struct ext4_sb_info *sbi = EXT4_SB(inode->i_sb);
1869 	struct ext4_pending_tree *tree = &EXT4_I(inode)->i_pending_tree;
1870 	struct rb_node **p = &tree->root.rb_node;
1871 	struct rb_node *parent = NULL;
1872 	struct pending_reservation *pr;
1873 	ext4_lblk_t lclu;
1874 	int ret = 0;
1875 
1876 	lclu = EXT4_B2C(sbi, lblk);
1877 	/* search to find parent for insertion */
1878 	while (*p) {
1879 		parent = *p;
1880 		pr = rb_entry(parent, struct pending_reservation, rb_node);
1881 
1882 		if (lclu < pr->lclu) {
1883 			p = &(*p)->rb_left;
1884 		} else if (lclu > pr->lclu) {
1885 			p = &(*p)->rb_right;
1886 		} else {
1887 			/* pending reservation already inserted */
1888 			goto out;
1889 		}
1890 	}
1891 
1892 	pr = kmem_cache_alloc(ext4_pending_cachep, GFP_ATOMIC);
1893 	if (pr == NULL) {
1894 		ret = -ENOMEM;
1895 		goto out;
1896 	}
1897 	pr->lclu = lclu;
1898 
1899 	rb_link_node(&pr->rb_node, parent, p);
1900 	rb_insert_color(&pr->rb_node, &tree->root);
1901 
1902 out:
1903 	return ret;
1904 }
1905 
1906 /*
1907  * __remove_pending - removes a pending cluster reservation from the set
1908  *                    of pending reservations
1909  *
1910  * @inode - file containing the cluster
1911  * @lblk - logical block in the pending cluster reservation to be removed
1912  *
1913  * Returns successfully if pending reservation is not a member of the set.
1914  */
1915 static void __remove_pending(struct inode *inode, ext4_lblk_t lblk)
1916 {
1917 	struct ext4_sb_info *sbi = EXT4_SB(inode->i_sb);
1918 	struct pending_reservation *pr;
1919 	struct ext4_pending_tree *tree;
1920 
1921 	pr = __get_pending(inode, EXT4_B2C(sbi, lblk));
1922 	if (pr != NULL) {
1923 		tree = &EXT4_I(inode)->i_pending_tree;
1924 		rb_erase(&pr->rb_node, &tree->root);
1925 		kmem_cache_free(ext4_pending_cachep, pr);
1926 	}
1927 }
1928 
1929 /*
1930  * ext4_remove_pending - removes a pending cluster reservation from the set
1931  *                       of pending reservations
1932  *
1933  * @inode - file containing the cluster
1934  * @lblk - logical block in the pending cluster reservation to be removed
1935  *
1936  * Locking for external use of __remove_pending.
1937  */
1938 void ext4_remove_pending(struct inode *inode, ext4_lblk_t lblk)
1939 {
1940 	struct ext4_inode_info *ei = EXT4_I(inode);
1941 
1942 	write_lock(&ei->i_es_lock);
1943 	__remove_pending(inode, lblk);
1944 	write_unlock(&ei->i_es_lock);
1945 }
1946 
1947 /*
1948  * ext4_is_pending - determine whether a cluster has a pending reservation
1949  *                   on it
1950  *
1951  * @inode - file containing the cluster
1952  * @lblk - logical block in the cluster
1953  *
1954  * Returns true if there's a pending reservation for the cluster in the
1955  * set of pending reservations, and false if not.
1956  */
1957 bool ext4_is_pending(struct inode *inode, ext4_lblk_t lblk)
1958 {
1959 	struct ext4_sb_info *sbi = EXT4_SB(inode->i_sb);
1960 	struct ext4_inode_info *ei = EXT4_I(inode);
1961 	bool ret;
1962 
1963 	read_lock(&ei->i_es_lock);
1964 	ret = (bool)(__get_pending(inode, EXT4_B2C(sbi, lblk)) != NULL);
1965 	read_unlock(&ei->i_es_lock);
1966 
1967 	return ret;
1968 }
1969 
1970 /*
1971  * ext4_es_insert_delayed_block - adds a delayed block to the extents status
1972  *                                tree, adding a pending reservation where
1973  *                                needed
1974  *
1975  * @inode - file containing the newly added block
1976  * @lblk - logical block to be added
1977  * @allocated - indicates whether a physical cluster has been allocated for
1978  *              the logical cluster that contains the block
1979  *
1980  * Returns 0 on success, negative error code on failure.
1981  */
1982 int ext4_es_insert_delayed_block(struct inode *inode, ext4_lblk_t lblk,
1983 				 bool allocated)
1984 {
1985 	struct extent_status newes;
1986 	int err = 0;
1987 
1988 	if (EXT4_SB(inode->i_sb)->s_mount_state & EXT4_FC_REPLAY)
1989 		return 0;
1990 
1991 	es_debug("add [%u/1) delayed to extent status tree of inode %lu\n",
1992 		 lblk, inode->i_ino);
1993 
1994 	newes.es_lblk = lblk;
1995 	newes.es_len = 1;
1996 	ext4_es_store_pblock_status(&newes, ~0, EXTENT_STATUS_DELAYED);
1997 	trace_ext4_es_insert_delayed_block(inode, &newes, allocated);
1998 
1999 	ext4_es_insert_extent_check(inode, &newes);
2000 
2001 	write_lock(&EXT4_I(inode)->i_es_lock);
2002 
2003 	err = __es_remove_extent(inode, lblk, lblk, NULL);
2004 	if (err != 0)
2005 		goto error;
2006 retry:
2007 	err = __es_insert_extent(inode, &newes);
2008 	if (err == -ENOMEM && __es_shrink(EXT4_SB(inode->i_sb),
2009 					  128, EXT4_I(inode)))
2010 		goto retry;
2011 	if (err != 0)
2012 		goto error;
2013 
2014 	if (allocated)
2015 		__insert_pending(inode, lblk);
2016 
2017 error:
2018 	write_unlock(&EXT4_I(inode)->i_es_lock);
2019 
2020 	ext4_es_print_tree(inode);
2021 	ext4_print_pending_tree(inode);
2022 
2023 	return err;
2024 }
2025 
2026 /*
2027  * __es_delayed_clu - count number of clusters containing blocks that
2028  *                    are delayed only
2029  *
2030  * @inode - file containing block range
2031  * @start - logical block defining start of range
2032  * @end - logical block defining end of range
2033  *
2034  * Returns the number of clusters containing only delayed (not delayed
2035  * and unwritten) blocks in the range specified by @start and @end.  Any
2036  * cluster or part of a cluster within the range and containing a delayed
2037  * and not unwritten block within the range is counted as a whole cluster.
2038  */
2039 static unsigned int __es_delayed_clu(struct inode *inode, ext4_lblk_t start,
2040 				     ext4_lblk_t end)
2041 {
2042 	struct ext4_es_tree *tree = &EXT4_I(inode)->i_es_tree;
2043 	struct extent_status *es;
2044 	struct ext4_sb_info *sbi = EXT4_SB(inode->i_sb);
2045 	struct rb_node *node;
2046 	ext4_lblk_t first_lclu, last_lclu;
2047 	unsigned long long last_counted_lclu;
2048 	unsigned int n = 0;
2049 
2050 	/* guaranteed to be unequal to any ext4_lblk_t value */
2051 	last_counted_lclu = ~0ULL;
2052 
2053 	es = __es_tree_search(&tree->root, start);
2054 
2055 	while (es && (es->es_lblk <= end)) {
2056 		if (ext4_es_is_delonly(es)) {
2057 			if (es->es_lblk <= start)
2058 				first_lclu = EXT4_B2C(sbi, start);
2059 			else
2060 				first_lclu = EXT4_B2C(sbi, es->es_lblk);
2061 
2062 			if (ext4_es_end(es) >= end)
2063 				last_lclu = EXT4_B2C(sbi, end);
2064 			else
2065 				last_lclu = EXT4_B2C(sbi, ext4_es_end(es));
2066 
2067 			if (first_lclu == last_counted_lclu)
2068 				n += last_lclu - first_lclu;
2069 			else
2070 				n += last_lclu - first_lclu + 1;
2071 			last_counted_lclu = last_lclu;
2072 		}
2073 		node = rb_next(&es->rb_node);
2074 		if (!node)
2075 			break;
2076 		es = rb_entry(node, struct extent_status, rb_node);
2077 	}
2078 
2079 	return n;
2080 }
2081 
2082 /*
2083  * ext4_es_delayed_clu - count number of clusters containing blocks that
2084  *                       are both delayed and unwritten
2085  *
2086  * @inode - file containing block range
2087  * @lblk - logical block defining start of range
2088  * @len - number of blocks in range
2089  *
2090  * Locking for external use of __es_delayed_clu().
2091  */
2092 unsigned int ext4_es_delayed_clu(struct inode *inode, ext4_lblk_t lblk,
2093 				 ext4_lblk_t len)
2094 {
2095 	struct ext4_inode_info *ei = EXT4_I(inode);
2096 	ext4_lblk_t end;
2097 	unsigned int n;
2098 
2099 	if (len == 0)
2100 		return 0;
2101 
2102 	end = lblk + len - 1;
2103 	WARN_ON(end < lblk);
2104 
2105 	read_lock(&ei->i_es_lock);
2106 
2107 	n = __es_delayed_clu(inode, lblk, end);
2108 
2109 	read_unlock(&ei->i_es_lock);
2110 
2111 	return n;
2112 }
2113 
2114 /*
2115  * __revise_pending - makes, cancels, or leaves unchanged pending cluster
2116  *                    reservations for a specified block range depending
2117  *                    upon the presence or absence of delayed blocks
2118  *                    outside the range within clusters at the ends of the
2119  *                    range
2120  *
2121  * @inode - file containing the range
2122  * @lblk - logical block defining the start of range
2123  * @len  - length of range in blocks
2124  *
2125  * Used after a newly allocated extent is added to the extents status tree.
2126  * Requires that the extents in the range have either written or unwritten
2127  * status.  Must be called while holding i_es_lock.
2128  */
2129 static void __revise_pending(struct inode *inode, ext4_lblk_t lblk,
2130 			     ext4_lblk_t len)
2131 {
2132 	struct ext4_sb_info *sbi = EXT4_SB(inode->i_sb);
2133 	ext4_lblk_t end = lblk + len - 1;
2134 	ext4_lblk_t first, last;
2135 	bool f_del = false, l_del = false;
2136 
2137 	if (len == 0)
2138 		return;
2139 
2140 	/*
2141 	 * Two cases - block range within single cluster and block range
2142 	 * spanning two or more clusters.  Note that a cluster belonging
2143 	 * to a range starting and/or ending on a cluster boundary is treated
2144 	 * as if it does not contain a delayed extent.  The new range may
2145 	 * have allocated space for previously delayed blocks out to the
2146 	 * cluster boundary, requiring that any pre-existing pending
2147 	 * reservation be canceled.  Because this code only looks at blocks
2148 	 * outside the range, it should revise pending reservations
2149 	 * correctly even if the extent represented by the range can't be
2150 	 * inserted in the extents status tree due to ENOSPC.
2151 	 */
2152 
2153 	if (EXT4_B2C(sbi, lblk) == EXT4_B2C(sbi, end)) {
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 			last = EXT4_LBLK_CMASK(sbi, end) +
2162 			       sbi->s_cluster_ratio - 1;
2163 			if (last != end)
2164 				l_del = __es_scan_range(inode,
2165 							&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 	} else {
2173 		first = EXT4_LBLK_CMASK(sbi, lblk);
2174 		if (first != lblk)
2175 			f_del = __es_scan_range(inode, &ext4_es_is_delonly,
2176 						first, lblk - 1);
2177 		if (f_del)
2178 			__insert_pending(inode, first);
2179 		else
2180 			__remove_pending(inode, first);
2181 
2182 		last = EXT4_LBLK_CMASK(sbi, end) + sbi->s_cluster_ratio - 1;
2183 		if (last != end)
2184 			l_del = __es_scan_range(inode, &ext4_es_is_delonly,
2185 						end + 1, last);
2186 		if (l_del)
2187 			__insert_pending(inode, last);
2188 		else
2189 			__remove_pending(inode, last);
2190 	}
2191 }
2192