summaryrefslogtreecommitdiffhomepage
path: root/src/nxt_mem_zone.c
blob: a3ba370076c0dd4589ec2b3428e8852687292803 (plain) (blame)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948

/*
 * Copyright (C) Igor Sysoev
 * Copyright (C) NGINX, Inc.
 */

#include <nxt_main.h>


#define NXT_MEM_ZONE_PAGE_FREE      0
/*
 * A page was never allocated before so it should be filled with
 * junk on the first time allocation if memory debugging is enabled.
 */
#define NXT_MEM_ZONE_PAGE_FRESH     1

/* An entire page is currently used, no chunks inside the page. */
#define NXT_MEM_ZONE_PAGE_USED      2


typedef struct nxt_mem_zone_page_s  nxt_mem_zone_page_t;

struct nxt_mem_zone_page_s {
    /*
     * A size of page chunks if value is greater than or equal to 16.
     * Otherwise it is used to mark page state: NXT_MEM_ZONE_PAGE_FREE,
     * NXT_MEM_ZONE_PAGE_FRESH, and NXT_MEM_ZONE_PAGE_USED.
     */
    uint16_t               size;

    /* A number of free chunks of a chunked page. */
    uint16_t               chunks;

    union {
        /* A chunk bitmap if a number of chunks is lesser than 32. */
        uint8_t            map[4];
        /*
         * The count is a number of successive occupied pages in the first
         * page.  In the next occupied pages and in all free pages the count
         * is zero, because a number of successive free pages is stored in
         * free block size resided in beginning of the first free page.
         */
        uint32_t           count;
    } u;

    /* Used for slot list of pages with free chunks. */
    nxt_mem_zone_page_t    *next;

    /*
     * Used to link of all pages including free, chunked and occupied
     * pages to coalesce free pages.
     */
    nxt_queue_link_t       link;
};


typedef struct {
    uint32_t               size;
    uint32_t               chunks;
    uint32_t               start;
    uint32_t               map_size;
    nxt_mem_zone_page_t    *pages;
} nxt_mem_zone_slot_t;


typedef struct {
    NXT_RBTREE_NODE        (node);
    uint32_t               size;
} nxt_mem_zone_free_block_t;


struct nxt_mem_zone_s {
    nxt_thread_spinlock_t  lock;
    nxt_mem_zone_page_t    *pages;
    nxt_mem_zone_page_t    sentinel_page;
    nxt_rbtree_t           free_pages;

    uint32_t               page_size_shift;
    uint32_t               page_size_mask;
    uint32_t               max_chunk_size;
    uint32_t               small_bitmap_min_size;

    u_char                 *start;
    u_char                 *end;

    nxt_mem_zone_slot_t    slots[];
};


#define nxt_mem_zone_page_addr(zone, page)                                    \
    (void *) (zone->start + ((page - zone->pages) << zone->page_size_shift))


#define nxt_mem_zone_addr_page(zone, addr)                                    \
    &zone->pages[((u_char *) addr - zone->start) >> zone->page_size_shift]


#define nxt_mem_zone_page_is_free(page)                                       \
    (page->size < NXT_MEM_ZONE_PAGE_USED)


#define nxt_mem_zone_page_is_chunked(page)                                    \
    (page->size >= 16)


#define nxt_mem_zone_page_bitmap(zone, slot)                                  \
    (slot->size < zone->small_bitmap_min_size)


#define nxt_mem_zone_set_chunk_free(map, chunk)                               \
    map[chunk / 8] &= ~(0x80 >> (chunk & 7))


#define nxt_mem_zone_chunk_is_free(map, chunk)                                \
    ((map[chunk / 8] & (0x80 >> (chunk & 7))) == 0)


#define nxt_mem_zone_fresh_junk(p, size)                                      \
    nxt_memset((p), 0xA5, size)


#define nxt_mem_zone_free_junk(p, size)                                       \
    nxt_memset((p), 0x5A, size)


static uint32_t nxt_mem_zone_pages(u_char *start, size_t zone_size,
    nxt_uint_t page_size);
static void *nxt_mem_zone_slots_init(nxt_mem_zone_t *zone,
    nxt_uint_t page_size);
static void nxt_mem_zone_slot_init(nxt_mem_zone_slot_t *slot,
    nxt_uint_t page_size);
static intptr_t nxt_mem_zone_rbtree_compare(nxt_rbtree_node_t *node1,
    nxt_rbtree_node_t *node2);
static void *nxt_mem_zone_alloc_small(nxt_mem_zone_t *zone,
    nxt_mem_zone_slot_t *slot, size_t size);
static nxt_uint_t nxt_mem_zone_alloc_chunk(uint8_t *map, nxt_uint_t offset,
    nxt_uint_t size);
static void *nxt_mem_zone_alloc_large(nxt_mem_zone_t *zone, size_t alignment,
    size_t size);
static nxt_mem_zone_page_t *nxt_mem_zone_alloc_pages(nxt_mem_zone_t *zone,
    size_t alignment, uint32_t pages);
static nxt_mem_zone_free_block_t *
    nxt_mem_zone_find_free_block(nxt_mem_zone_t *zone, nxt_rbtree_node_t *node,
    uint32_t alignment, uint32_t pages);
static const char *nxt_mem_zone_free_chunk(nxt_mem_zone_t *zone,
    nxt_mem_zone_page_t *page, void *p);
static void nxt_mem_zone_free_pages(nxt_mem_zone_t *zone,
    nxt_mem_zone_page_t *page, nxt_uint_t count);


static nxt_log_moderation_t  nxt_mem_zone_log_moderation = {
    NXT_LOG_ALERT, 2, "mem_zone_alloc() failed, not enough memory",
    NXT_LOG_MODERATION
};


nxt_mem_zone_t *
nxt_mem_zone_init(u_char *start, size_t zone_size, nxt_uint_t page_size)
{
    uint32_t                   pages;
    nxt_uint_t                 n;
    nxt_mem_zone_t             *zone;
    nxt_mem_zone_page_t        *page;
    nxt_mem_zone_free_block_t  *block;

    if (nxt_slow_path((page_size & (page_size - 1)) != 0)) {
        nxt_thread_log_alert("mem zone page size must be a power of 2");
        return NULL;
    }

    pages = nxt_mem_zone_pages(start, zone_size, page_size);
    if (pages == 0) {
        return NULL;
    }

    zone = (nxt_mem_zone_t *) start;

    /* The function returns address after all slots. */
    page = nxt_mem_zone_slots_init(zone, page_size);

    zone->pages = page;

    for (n = 0; n < pages; n++) {
        page[n].size = NXT_MEM_ZONE_PAGE_FRESH;
    }

    /*
     * A special sentinel page entry marked as used does not correspond
     * to a real page.  The entry simplifies neighbour queue nodes check
     * in nxt_mem_zone_free_pages().
     */
    zone->sentinel_page.size = NXT_MEM_ZONE_PAGE_USED;
    nxt_queue_sentinel(&zone->sentinel_page.link);
    nxt_queue_insert_after(&zone->sentinel_page.link, &page->link);

    /* rbtree of free pages. */

    nxt_rbtree_init(&zone->free_pages, nxt_mem_zone_rbtree_compare);

    block = (nxt_mem_zone_free_block_t *) zone->start;
    block->size = pages;

    nxt_rbtree_insert(&zone->free_pages, &block->node);

    return zone;
}


static uint32_t
nxt_mem_zone_pages(u_char *start, size_t zone_size, nxt_uint_t page_size)
{
    u_char          *end;
    size_t          reserved;
    nxt_uint_t      n, pages, size, chunks, last;
    nxt_mem_zone_t  *zone;

    /*
     * Find all maximum chunk sizes which zone page can be split on
     * with minimum 16-byte step.
     */
    last = page_size / 16;
    n = 0;
    size = 32;

    do {
        chunks = page_size / size;

        if (last != chunks) {
            last = chunks;
            n++;
        }

        size += 16;

    } while (chunks > 1);

    /*
     * Find number of usable zone pages except zone bookkeeping data,
     * slots, and pages entries.
     */
    reserved = sizeof(nxt_mem_zone_t) + (n * sizeof(nxt_mem_zone_slot_t));

    end = nxt_trunc_ptr(start + zone_size, page_size);
    zone_size = end - start;

    pages = (zone_size - reserved) / (page_size + sizeof(nxt_mem_zone_page_t));

    if (reserved > zone_size || pages == 0) {
        nxt_thread_log_alert("mem zone size is too small: %uz", zone_size);
        return 0;
    }

    reserved += pages * sizeof(nxt_mem_zone_page_t);
    nxt_memzero(start, reserved);

    zone = (nxt_mem_zone_t *) start;

    zone->start = nxt_align_ptr(start + reserved, page_size);
    zone->end = end;

    nxt_thread_log_debug("mem zone pages: %uD, unused:%z", pages,
                         end - (zone->start + pages * page_size));

    /*
     * If a chunk size is lesser than zone->small_bitmap_min_size
     * bytes, a page's chunk bitmap is larger than 32 bits and the
     * bimap is placed at the start of the page.
     */
    zone->small_bitmap_min_size = page_size / 32;

    zone->page_size_mask = page_size - 1;
    zone->max_chunk_size = page_size / 2;

    n = zone->max_chunk_size;

    do {
        zone->page_size_shift++;
        n /= 2;
    } while (n != 0);

    return (uint32_t) pages;
}


static void *
nxt_mem_zone_slots_init(nxt_mem_zone_t *zone, nxt_uint_t page_size)
{
    nxt_uint_t           n, size, chunks;
    nxt_mem_zone_slot_t  *slot;

    slot = zone->slots;

    slot[0].chunks = page_size / 16;
    slot[0].size = 16;

    n = 0;
    size = 32;

    for ( ;; ) {
        chunks = page_size / size;

        if (slot[n].chunks != chunks) {

            nxt_mem_zone_slot_init(&slot[n], page_size);

            nxt_thread_log_debug(
                           "mem zone size:%uD chunks:%uD start:%uD map:%uD",
                           slot[n].size, slot[n].chunks + 1,
                           slot[n].start, slot[n].map_size);

            n++;

            if (chunks == 1) {
                return &slot[n];
            }
        }

        slot[n].chunks = chunks;
        slot[n].size = size;
        size += 16;
    }
}


static void
nxt_mem_zone_slot_init(nxt_mem_zone_slot_t *slot, nxt_uint_t page_size)
{
    /*
     * Calculate number of bytes required to store a chunk bitmap
     * and align it to 4 bytes.
     */
    slot->map_size = nxt_align_size(((slot->chunks + 7) / 8), 4);

    /* If chunk size is not a multiple of zone page size, there
     * is surplus space which can be used for the chunk's bitmap.
     */
    slot->start = page_size - slot->chunks * slot->size;

    /* slot->chunks should be one less than actual number of chunks. */
    slot->chunks--;

    if (slot->map_size > 4) {
        /* A page's chunks bitmap is placed at the start of the page. */

        if (slot->start < slot->map_size) {
            /*
             * There is no surplus space or the space is too
             * small for chunks bitmap, so use the first chunks.
             */
            if (slot->size < slot->map_size) {
                /* The first chunks are occupied by bitmap. */
                slot->chunks -= slot->map_size / slot->size;
                slot->start = nxt_align_size(slot->map_size, 16);

            } else {
                /* The first chunk is occupied by bitmap. */
                slot->chunks--;
                slot->start = slot->size;
            }
        }
    }
}


/*
 * Round up to the next highest power of 2.  The algorithm is
 * described in "Bit Twiddling Hacks" by Sean Eron Anderson.
 */

nxt_inline uint32_t
nxt_next_highest_power_of_two(uint32_t n)
{
    n--;
    n |= n >> 1;
    n |= n >> 2;
    n |= n >> 4;
    n |= n >> 8;
    n |= n >> 16;
    n++;

    return n;
}


static intptr_t
nxt_mem_zone_rbtree_compare(nxt_rbtree_node_t *node1, nxt_rbtree_node_t *node2)
{
    u_char                     *start1, *end1, *start2, *end2;
    uint32_t                   n, size, size1, size2;
    nxt_mem_zone_free_block_t  *block1, *block2;

    block1 = (nxt_mem_zone_free_block_t *) node1;
    block2 = (nxt_mem_zone_free_block_t *) node2;

    size1 = block1->size;
    size2 = block2->size;

    /*
     * This subtractions do not overflow if number of pages of a free
     * block is below 2^31-1.  This allows to use blocks up to 128G if
     * a zone page size is just 64 bytes.
     */
    n = size1 - size2;

    if (n != 0) {
        return n;
    }

    /*
     * Sort equally sized blocks by their capability to allocate memory with
     * alignment equal to the size rounded the previous higest power of 2.
     */

    /* Round the size to the previous higest power of two. */
    size = nxt_next_highest_power_of_two(size1) >> 1;

    /* Align the blocks' start and end to the rounded size. */
    start1 = nxt_align_ptr(block1, size);
    end1 = nxt_trunc_ptr((u_char *) block1 + size1, size);

    start2 = nxt_align_ptr(block2, size);
    end2 = nxt_trunc_ptr((u_char *) block2 + size2, size);

    return (end1 - start1) - (end2 - start2);
}


void *
nxt_mem_zone_zalloc(nxt_mem_zone_t *zone, size_t size)
{
    void  *p;

    p = nxt_mem_zone_align(zone, 1, size);

    if (nxt_fast_path(p != NULL)) {
        nxt_memzero(p, size);
    }

    return p;
}


void *
nxt_mem_zone_align(nxt_mem_zone_t *zone, size_t alignment, size_t size)
{
    void                 *p;
    nxt_mem_zone_slot_t  *slot;

    if (nxt_slow_path((alignment - 1) & alignment) != 0) {
        /* Alignment must be a power of 2. */
        return NULL;
    }

    if (size <= zone->max_chunk_size && alignment <= zone->max_chunk_size) {
        /* All chunks are aligned to 16. */

        if (alignment > 16) {
            /*
             * Chunks which size is power of 2 are aligned to the size.
             * So allocation size should be increased to the next highest
             * power of two.  This can waste memory, but a main consumer
             * of aligned allocations is lvlhsh which anyway allocates
             * memory with alignment equal to size.
             */
            size = nxt_next_highest_power_of_two(size);
            size = nxt_max(size, alignment);
        }

        /*
         * Find a zone slot with appropriate chunk size.
         * This operation can be performed without holding lock.
         */
        for (slot = zone->slots; slot->size < size; slot++) { /* void */ }

        nxt_thread_log_debug("mem zone alloc: @%uz:%uz chunk:%uD",
                             alignment, size, slot->size);

        nxt_thread_spin_lock(&zone->lock);

        p = nxt_mem_zone_alloc_small(zone, slot, size);

    } else {

        nxt_thread_log_debug("mem zone alloc: @%uz:%uz", alignment, size);

        nxt_thread_spin_lock(&zone->lock);

        p = nxt_mem_zone_alloc_large(zone, alignment, size);
    }

    nxt_thread_spin_unlock(&zone->lock);

    if (nxt_fast_path(p != NULL)) {
        nxt_thread_log_debug("mem zone alloc: %p", p);

    } else {
        nxt_log_alert_moderate(&nxt_mem_zone_log_moderation, nxt_thread_log(),
                    "nxt_mem_zone_alloc(%uz, %uz) failed, not enough memory",
                    alignment, size);
    }

    return p;
}


static void *
nxt_mem_zone_alloc_small(nxt_mem_zone_t *zone, nxt_mem_zone_slot_t *slot,
    size_t size)
{
    u_char               *p;
    uint8_t              *map;
    nxt_mem_zone_page_t  *page;

    page = slot->pages;

    if (nxt_fast_path(page != NULL)) {

        p = nxt_mem_zone_page_addr(zone, page);

        if (nxt_mem_zone_page_bitmap(zone, slot)) {
            /* A page's chunks bitmap is placed at the start of the page. */
            map = p;

        } else {
            map = page->u.map;
        }

        p += nxt_mem_zone_alloc_chunk(map, slot->start, slot->size);

        page->chunks--;

        if (page->chunks == 0) {
            /*
             * Remove full page from the zone slot list of pages with
             * free chunks.
             */
            slot->pages = page->next;
#if (NXT_DEBUG)
            page->next = NULL;
#endif
        }

        return p;
    }

    page = nxt_mem_zone_alloc_pages(zone, 1, 1);

    if (nxt_fast_path(page != NULL)) {

        slot->pages = page;

        page->size = slot->size;
        /* slot->chunks are already one less. */
        page->chunks = slot->chunks;
        page->u.count = 0;
        page->next = NULL;

        p = nxt_mem_zone_page_addr(zone, page);

        if (nxt_mem_zone_page_bitmap(zone, slot)) {
            /* A page's chunks bitmap is placed at the start of the page. */
            map = p;
            nxt_memzero(map, slot->map_size);

        } else {
            map = page->u.map;
        }

        /* Mark the first chunk as busy. */
        map[0] = 0x80;

        return p + slot->start;
    }

    return NULL;
}


static nxt_uint_t
nxt_mem_zone_alloc_chunk(uint8_t *map, nxt_uint_t offset, nxt_uint_t size)
{
    uint8_t     mask;
    nxt_uint_t  n;

    n = 0;

    /* The page must have at least one free chunk. */

    for ( ;; ) {
        /* The bitmap is always aligned to uint32_t. */

        if (*(uint32_t *) &map[n] != 0xFFFFFFFF) {

            do {
                if (map[n] != 0xFF) {

                    mask = 0x80;

                    do {
                        if ((map[n] & mask) == 0) {
                            /* The free chunk is found. */
                            map[n] |= mask;
                            return offset;
                        }

                        offset += size;
                        mask >>= 1;

                    } while (mask != 0);

                } else {
                    /* Fast-forward: all 8 chunks are occupied. */
                    offset += size * 8;
                }

                n++;

            } while (n % 4 != 0);

        } else {
            /* Fast-forward: all 32 chunks are occupied. */
            offset += size * 32;
            n += 4;
        }
    }
}


static void *
nxt_mem_zone_alloc_large(nxt_mem_zone_t *zone, size_t alignment, size_t size)
{
    uint32_t             pages;
    nxt_mem_zone_page_t  *page;

    pages = (size + zone->page_size_mask) >> zone->page_size_shift;

    page = nxt_mem_zone_alloc_pages(zone, alignment, pages);

    if (nxt_fast_path(page != NULL)) {
        return nxt_mem_zone_page_addr(zone, page);
    }

    return NULL;
}


static nxt_mem_zone_page_t *
nxt_mem_zone_alloc_pages(nxt_mem_zone_t *zone, size_t alignment, uint32_t pages)
{
    u_char                     *p;
    size_t                     prev_size;
    uint32_t                   prev_pages, node_pages, next_pages;
    nxt_uint_t                 n;
    nxt_mem_zone_page_t        *prev_page, *page, *next_page;
    nxt_mem_zone_free_block_t  *block, *next_block;

    block = nxt_mem_zone_find_free_block(zone,
                                         nxt_rbtree_root(&zone->free_pages),
                                         alignment, pages);

    if (nxt_slow_path(block == NULL)) {
        return NULL;
    }

    node_pages = block->size;

    nxt_rbtree_delete(&zone->free_pages, &block->node);

    p = nxt_align_ptr(block, alignment);
    page = nxt_mem_zone_addr_page(zone, p);

    prev_size = p - (u_char *) block;

    if (prev_size != 0) {
        prev_pages = prev_size >> zone->page_size_shift;
        node_pages -= prev_pages;

        block->size = prev_pages;
        nxt_rbtree_insert(&zone->free_pages, &block->node);

        prev_page = nxt_mem_zone_addr_page(zone, block);
        nxt_queue_insert_after(&prev_page->link, &page->link);
    }

    next_pages = node_pages - pages;

    if (next_pages != 0) {
        next_page = &page[pages];
        next_block = nxt_mem_zone_page_addr(zone, next_page);
        next_block->size = next_pages;

        nxt_rbtree_insert(&zone->free_pages, &next_block->node);
        nxt_queue_insert_after(&page->link, &next_page->link);
    }

    /* Go through pages after all rbtree operations to not trash CPU cache. */

    page[0].u.count = pages;

    for (n = 0; n < pages; n++) {

        if (page[n].size == NXT_MEM_ZONE_PAGE_FRESH) {
            nxt_mem_zone_fresh_junk(nxt_mem_zone_page_addr(zone, &page[n]),
                                    zone->page_size_mask + 1);
        }

        page[n].size = NXT_MEM_ZONE_PAGE_USED;
    }

    return page;
}


/*
 * Free blocks are sorted by size and then if the sizes are equal
 * by aligned allocation capabilty.  The former criterion is just
 * comparison with a requested size and it can be used for iteractive
 * search.  The later criterion cannot be tested only by the requested
 * size and alignment, so recursive in-order tree traversal is required
 * to find a suitable free block.  nxt_mem_zone_find_free_block() uses
 * only recursive in-order tree traversal because anyway the slowest part
 * of the algorithm are CPU cache misses.  Besides the last tail recursive
 * call may be optimized by compiler into iteractive search.
 */

static nxt_mem_zone_free_block_t *
nxt_mem_zone_find_free_block(nxt_mem_zone_t *zone, nxt_rbtree_node_t *node,
    uint32_t alignment, uint32_t pages)
{
    u_char                     *aligned, *end;
    nxt_mem_zone_free_block_t  *block, *free_block;

    if (node == nxt_rbtree_sentinel(&zone->free_pages)) {
        return NULL;
    }

    block = (nxt_mem_zone_free_block_t *) node;

    if (pages <= block->size) {

        free_block = nxt_mem_zone_find_free_block(zone, block->node.left,
                                                  alignment, pages);
        if (free_block != NULL) {
            return free_block;
        }

        aligned = nxt_align_ptr(block, alignment);

        if (pages == block->size) {
            if (aligned == (u_char *) block) {
                /* Exact match. */
                return block;
            }

        } else {  /* pages < block->size */
            aligned += pages << zone->page_size_shift;
            end = nxt_pointer_to(block, block->size << zone->page_size_shift);

            if (aligned <= end) {
                return block;
            }
        }
    }

    return nxt_mem_zone_find_free_block(zone, block->node.right,
                                        alignment, pages);
}


void
nxt_mem_zone_free(nxt_mem_zone_t *zone, void *p)
{
    nxt_uint_t           count;
    const char           *err;
    nxt_mem_zone_page_t  *page;

    nxt_thread_log_debug("mem zone free: %p", p);

    if (nxt_fast_path(zone->start <= (u_char *) p
                      && (u_char *) p < zone->end))
    {
        page = nxt_mem_zone_addr_page(zone, p);

        nxt_thread_spin_lock(&zone->lock);

        if (nxt_mem_zone_page_is_chunked(page)) {
            err = nxt_mem_zone_free_chunk(zone, page, p);

        } else if (nxt_slow_path(nxt_mem_zone_page_is_free(page))) {
            err = "page is already free";

        } else if (nxt_slow_path((uintptr_t) p & zone->page_size_mask) != 0) {
            err = "invalid pointer to chunk";

        } else {
            count = page->u.count;

            if (nxt_fast_path(count != 0)) {
                nxt_mem_zone_free_junk(p, count * zone->page_size_mask + 1);
                nxt_mem_zone_free_pages(zone, page, count);
                err = NULL;

            } else {
                /* Not the first allocated page. */
                err = "pointer to wrong page";
            }
        }

        nxt_thread_spin_unlock(&zone->lock);

    } else {
        err = "pointer is out of zone";
    }

    if (nxt_slow_path(err != NULL)) {
        nxt_thread_log_alert("nxt_mem_zone_free(%p): %s", p, err);
    }
}


static const char *
nxt_mem_zone_free_chunk(nxt_mem_zone_t *zone, nxt_mem_zone_page_t *page,
    void *p)
{
    u_char               *map;
    uint32_t             size, offset, chunk;
    nxt_mem_zone_page_t  *pg, **ppg;
    nxt_mem_zone_slot_t  *slot;

    size = page->size;

    /* Find a zone slot with appropriate chunk size. */
    for (slot = zone->slots; slot->size < size; slot++) { /* void */ }

    offset = (uintptr_t) p & zone->page_size_mask;
    offset -= slot->start;

    chunk = offset / size;

    if (nxt_slow_path(offset != chunk * size)) {
        return "pointer to wrong chunk";
    }

    if (nxt_mem_zone_page_bitmap(zone, slot)) {
        /* A page's chunks bitmap is placed at the start of the page. */
        map = (u_char *) ((uintptr_t) p & ~((uintptr_t) zone->page_size_mask));

    } else {
        map = page->u.map;
    }

    if (nxt_mem_zone_chunk_is_free(map, chunk)) {
        return "chunk is already free";
    }

    nxt_mem_zone_set_chunk_free(map, chunk);

    nxt_mem_zone_free_junk(p, page->size);

    if (page->chunks == 0) {
        page->chunks = 1;

        /* Add the page to the head of slot list of pages with free chunks. */
        page->next = slot->pages;
        slot->pages = page;

    } else if (page->chunks != slot->chunks) {
        page->chunks++;

    } else {

        if (map != page->u.map) {
            nxt_mem_zone_free_junk(map, slot->map_size);
        }

        /*
         * All chunks are free, remove the page from the slot list of pages
         * with free chunks and add the page to the free pages tree.
         */
        ppg = &slot->pages;

        for (pg = slot->pages; pg != NULL; pg = pg->next) {

            if (pg == page) {
                *ppg = page->next;
                break;
            }

            ppg = &pg->next;
        }

        nxt_mem_zone_free_pages(zone, page, 1);
    }

    return NULL;
}


static void
nxt_mem_zone_free_pages(nxt_mem_zone_t *zone, nxt_mem_zone_page_t *page,
    nxt_uint_t count)
{
    nxt_mem_zone_page_t        *prev_page, *next_page;
    nxt_mem_zone_free_block_t  *block, *prev_block, *next_block;

    page->size = NXT_MEM_ZONE_PAGE_FREE;
    page->chunks = 0;
    page->u.count = 0;
    page->next = NULL;

    nxt_memzero(&page[1], (count - 1) * sizeof(nxt_mem_zone_page_t));

    next_page = nxt_queue_link_data(page->link.next, nxt_mem_zone_page_t, link);

    if (nxt_mem_zone_page_is_free(next_page)) {

        /* Coalesce with the next free pages. */

        nxt_queue_remove(&next_page->link);
        nxt_memzero(next_page, sizeof(nxt_mem_zone_page_t));

        next_block = nxt_mem_zone_page_addr(zone, next_page);
        count += next_block->size;
        nxt_rbtree_delete(&zone->free_pages, &next_block->node);
    }

    prev_page = nxt_queue_link_data(page->link.prev, nxt_mem_zone_page_t, link);

    if (nxt_mem_zone_page_is_free(prev_page)) {

        /* Coalesce with the previous free pages. */

        nxt_queue_remove(&page->link);

        prev_block = nxt_mem_zone_page_addr(zone, prev_page);
        count += prev_block->size;
        nxt_rbtree_delete(&zone->free_pages, &prev_block->node);

        prev_block->size = count;
        nxt_rbtree_insert(&zone->free_pages, &prev_block->node);

        return;
    }

    block = nxt_mem_zone_page_addr(zone, page);
    block->size = count;
    nxt_rbtree_insert(&zone->free_pages, &block->node);
}