xref: /openbmc/linux/fs/ext4/extents_status.c (revision 4bb1e4e7)
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