summaryrefslogtreecommitdiff
path: root/storage/xtradb/btr/btr0cur.cc
blob: 728d14e6309ee409de09dc349c309cfac8aa20a2 (plain)
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
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
1001
1002
1003
1004
1005
1006
1007
1008
1009
1010
1011
1012
1013
1014
1015
1016
1017
1018
1019
1020
1021
1022
1023
1024
1025
1026
1027
1028
1029
1030
1031
1032
1033
1034
1035
1036
1037
1038
1039
1040
1041
1042
1043
1044
1045
1046
1047
1048
1049
1050
1051
1052
1053
1054
1055
1056
1057
1058
1059
1060
1061
1062
1063
1064
1065
1066
1067
1068
1069
1070
1071
1072
1073
1074
1075
1076
1077
1078
1079
1080
1081
1082
1083
1084
1085
1086
1087
1088
1089
1090
1091
1092
1093
1094
1095
1096
1097
1098
1099
1100
1101
1102
1103
1104
1105
1106
1107
1108
1109
1110
1111
1112
1113
1114
1115
1116
1117
1118
1119
1120
1121
1122
1123
1124
1125
1126
1127
1128
1129
1130
1131
1132
1133
1134
1135
1136
1137
1138
1139
1140
1141
1142
1143
1144
1145
1146
1147
1148
1149
1150
1151
1152
1153
1154
1155
1156
1157
1158
1159
1160
1161
1162
1163
1164
1165
1166
1167
1168
1169
1170
1171
1172
1173
1174
1175
1176
1177
1178
1179
1180
1181
1182
1183
1184
1185
1186
1187
1188
1189
1190
1191
1192
1193
1194
1195
1196
1197
1198
1199
1200
1201
1202
1203
1204
1205
1206
1207
1208
1209
1210
1211
1212
1213
1214
1215
1216
1217
1218
1219
1220
1221
1222
1223
1224
1225
1226
1227
1228
1229
1230
1231
1232
1233
1234
1235
1236
1237
1238
1239
1240
1241
1242
1243
1244
1245
1246
1247
1248
1249
1250
1251
1252
1253
1254
1255
1256
1257
1258
1259
1260
1261
1262
1263
1264
1265
1266
1267
1268
1269
1270
1271
1272
1273
1274
1275
1276
1277
1278
1279
1280
1281
1282
1283
1284
1285
1286
1287
1288
1289
1290
1291
1292
1293
1294
1295
1296
1297
1298
1299
1300
1301
1302
1303
1304
1305
1306
1307
1308
1309
1310
1311
1312
1313
1314
1315
1316
1317
1318
1319
1320
1321
1322
1323
1324
1325
1326
1327
1328
1329
1330
1331
1332
1333
1334
1335
1336
1337
1338
1339
1340
1341
1342
1343
1344
1345
1346
1347
1348
1349
1350
1351
1352
1353
1354
1355
1356
1357
1358
1359
1360
1361
1362
1363
1364
1365
1366
1367
1368
1369
1370
1371
1372
1373
1374
1375
1376
1377
1378
1379
1380
1381
1382
1383
1384
1385
1386
1387
1388
1389
1390
1391
1392
1393
1394
1395
1396
1397
1398
1399
1400
1401
1402
1403
1404
1405
1406
1407
1408
1409
1410
1411
1412
1413
1414
1415
1416
1417
1418
1419
1420
1421
1422
1423
1424
1425
1426
1427
1428
1429
1430
1431
1432
1433
1434
1435
1436
1437
1438
1439
1440
1441
1442
1443
1444
1445
1446
1447
1448
1449
1450
1451
1452
1453
1454
1455
1456
1457
1458
1459
1460
1461
1462
1463
1464
1465
1466
1467
1468
1469
1470
1471
1472
1473
1474
1475
1476
1477
1478
1479
1480
1481
1482
1483
1484
1485
1486
1487
1488
1489
1490
1491
1492
1493
1494
1495
1496
1497
1498
1499
1500
1501
1502
1503
1504
1505
1506
1507
1508
1509
1510
1511
1512
1513
1514
1515
1516
1517
1518
1519
1520
1521
1522
1523
1524
1525
1526
1527
1528
1529
1530
1531
1532
1533
1534
1535
1536
1537
1538
1539
1540
1541
1542
1543
1544
1545
1546
1547
1548
1549
1550
1551
1552
1553
1554
1555
1556
1557
1558
1559
1560
1561
1562
1563
1564
1565
1566
1567
1568
1569
1570
1571
1572
1573
1574
1575
1576
1577
1578
1579
1580
1581
1582
1583
1584
1585
1586
1587
1588
1589
1590
1591
1592
1593
1594
1595
1596
1597
1598
1599
1600
1601
1602
1603
1604
1605
1606
1607
1608
1609
1610
1611
1612
1613
1614
1615
1616
1617
1618
1619
1620
1621
1622
1623
1624
1625
1626
1627
1628
1629
1630
1631
1632
1633
1634
1635
1636
1637
1638
1639
1640
1641
1642
1643
1644
1645
1646
1647
1648
1649
1650
1651
1652
1653
1654
1655
1656
1657
1658
1659
1660
1661
1662
1663
1664
1665
1666
1667
1668
1669
1670
1671
1672
1673
1674
1675
1676
1677
1678
1679
1680
1681
1682
1683
1684
1685
1686
1687
1688
1689
1690
1691
1692
1693
1694
1695
1696
1697
1698
1699
1700
1701
1702
1703
1704
1705
1706
1707
1708
1709
1710
1711
1712
1713
1714
1715
1716
1717
1718
1719
1720
1721
1722
1723
1724
1725
1726
1727
1728
1729
1730
1731
1732
1733
1734
1735
1736
1737
1738
1739
1740
1741
1742
1743
1744
1745
1746
1747
1748
1749
1750
1751
1752
1753
1754
1755
1756
1757
1758
1759
1760
1761
1762
1763
1764
1765
1766
1767
1768
1769
1770
1771
1772
1773
1774
1775
1776
1777
1778
1779
1780
1781
1782
1783
1784
1785
1786
1787
1788
1789
1790
1791
1792
1793
1794
1795
1796
1797
1798
1799
1800
1801
1802
1803
1804
1805
1806
1807
1808
1809
1810
1811
1812
1813
1814
1815
1816
1817
1818
1819
1820
1821
1822
1823
1824
1825
1826
1827
1828
1829
1830
1831
1832
1833
1834
1835
1836
1837
1838
1839
1840
1841
1842
1843
1844
1845
1846
1847
1848
1849
1850
1851
1852
1853
1854
1855
1856
1857
1858
1859
1860
1861
1862
1863
1864
1865
1866
1867
1868
1869
1870
1871
1872
1873
1874
1875
1876
1877
1878
1879
1880
1881
1882
1883
1884
1885
1886
1887
1888
1889
1890
1891
1892
1893
1894
1895
1896
1897
1898
1899
1900
1901
1902
1903
1904
1905
1906
1907
1908
1909
1910
1911
1912
1913
1914
1915
1916
1917
1918
1919
1920
1921
1922
1923
1924
1925
1926
1927
1928
1929
1930
1931
1932
1933
1934
1935
1936
1937
1938
1939
1940
1941
1942
1943
1944
1945
1946
1947
1948
1949
1950
1951
1952
1953
1954
1955
1956
1957
1958
1959
1960
1961
1962
1963
1964
1965
1966
1967
1968
1969
1970
1971
1972
1973
1974
1975
1976
1977
1978
1979
1980
1981
1982
1983
1984
1985
1986
1987
1988
1989
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
2011
2012
2013
2014
2015
2016
2017
2018
2019
2020
2021
2022
2023
2024
2025
2026
2027
2028
2029
2030
2031
2032
2033
2034
2035
2036
2037
2038
2039
2040
2041
2042
2043
2044
2045
2046
2047
2048
2049
2050
2051
2052
2053
2054
2055
2056
2057
2058
2059
2060
2061
2062
2063
2064
2065
2066
2067
2068
2069
2070
2071
2072
2073
2074
2075
2076
2077
2078
2079
2080
2081
2082
2083
2084
2085
2086
2087
2088
2089
2090
2091
2092
2093
2094
2095
2096
2097
2098
2099
2100
2101
2102
2103
2104
2105
2106
2107
2108
2109
2110
2111
2112
2113
2114
2115
2116
2117
2118
2119
2120
2121
2122
2123
2124
2125
2126
2127
2128
2129
2130
2131
2132
2133
2134
2135
2136
2137
2138
2139
2140
2141
2142
2143
2144
2145
2146
2147
2148
2149
2150
2151
2152
2153
2154
2155
2156
2157
2158
2159
2160
2161
2162
2163
2164
2165
2166
2167
2168
2169
2170
2171
2172
2173
2174
2175
2176
2177
2178
2179
2180
2181
2182
2183
2184
2185
2186
2187
2188
2189
2190
2191
2192
2193
2194
2195
2196
2197
2198
2199
2200
2201
2202
2203
2204
2205
2206
2207
2208
2209
2210
2211
2212
2213
2214
2215
2216
2217
2218
2219
2220
2221
2222
2223
2224
2225
2226
2227
2228
2229
2230
2231
2232
2233
2234
2235
2236
2237
2238
2239
2240
2241
2242
2243
2244
2245
2246
2247
2248
2249
2250
2251
2252
2253
2254
2255
2256
2257
2258
2259
2260
2261
2262
2263
2264
2265
2266
2267
2268
2269
2270
2271
2272
2273
2274
2275
2276
2277
2278
2279
2280
2281
2282
2283
2284
2285
2286
2287
2288
2289
2290
2291
2292
2293
2294
2295
2296
2297
2298
2299
2300
2301
2302
2303
2304
2305
2306
2307
2308
2309
2310
2311
2312
2313
2314
2315
2316
2317
2318
2319
2320
2321
2322
2323
2324
2325
2326
2327
2328
2329
2330
2331
2332
2333
2334
2335
2336
2337
2338
2339
2340
2341
2342
2343
2344
2345
2346
2347
2348
2349
2350
2351
2352
2353
2354
2355
2356
2357
2358
2359
2360
2361
2362
2363
2364
2365
2366
2367
2368
2369
2370
2371
2372
2373
2374
2375
2376
2377
2378
2379
2380
2381
2382
2383
2384
2385
2386
2387
2388
2389
2390
2391
2392
2393
2394
2395
2396
2397
2398
2399
2400
2401
2402
2403
2404
2405
2406
2407
2408
2409
2410
2411
2412
2413
2414
2415
2416
2417
2418
2419
2420
2421
2422
2423
2424
2425
2426
2427
2428
2429
2430
2431
2432
2433
2434
2435
2436
2437
2438
2439
2440
2441
2442
2443
2444
2445
2446
2447
2448
2449
2450
2451
2452
2453
2454
2455
2456
2457
2458
2459
2460
2461
2462
2463
2464
2465
2466
2467
2468
2469
2470
2471
2472
2473
2474
2475
2476
2477
2478
2479
2480
2481
2482
2483
2484
2485
2486
2487
2488
2489
2490
2491
2492
2493
2494
2495
2496
2497
2498
2499
2500
2501
2502
2503
2504
2505
2506
2507
2508
2509
2510
2511
2512
2513
2514
2515
2516
2517
2518
2519
2520
2521
2522
2523
2524
2525
2526
2527
2528
2529
2530
2531
2532
2533
2534
2535
2536
2537
2538
2539
2540
2541
2542
2543
2544
2545
2546
2547
2548
2549
2550
2551
2552
2553
2554
2555
2556
2557
2558
2559
2560
2561
2562
2563
2564
2565
2566
2567
2568
2569
2570
2571
2572
2573
2574
2575
2576
2577
2578
2579
2580
2581
2582
2583
2584
2585
2586
2587
2588
2589
2590
2591
2592
2593
2594
2595
2596
2597
2598
2599
2600
2601
2602
2603
2604
2605
2606
2607
2608
2609
2610
2611
2612
2613
2614
2615
2616
2617
2618
2619
2620
2621
2622
2623
2624
2625
2626
2627
2628
2629
2630
2631
2632
2633
2634
2635
2636
2637
2638
2639
2640
2641
2642
2643
2644
2645
2646
2647
2648
2649
2650
2651
2652
2653
2654
2655
2656
2657
2658
2659
2660
2661
2662
2663
2664
2665
2666
2667
2668
2669
2670
2671
2672
2673
2674
2675
2676
2677
2678
2679
2680
2681
2682
2683
2684
2685
2686
2687
2688
2689
2690
2691
2692
2693
2694
2695
2696
2697
2698
2699
2700
2701
2702
2703
2704
2705
2706
2707
2708
2709
2710
2711
2712
2713
2714
2715
2716
2717
2718
2719
2720
2721
2722
2723
2724
2725
2726
2727
2728
2729
2730
2731
2732
2733
2734
2735
2736
2737
2738
2739
2740
2741
2742
2743
2744
2745
2746
2747
2748
2749
2750
2751
2752
2753
2754
2755
2756
2757
2758
2759
2760
2761
2762
2763
2764
2765
2766
2767
2768
2769
2770
2771
2772
2773
2774
2775
2776
2777
2778
2779
2780
2781
2782
2783
2784
2785
2786
2787
2788
2789
2790
2791
2792
2793
2794
2795
2796
2797
2798
2799
2800
2801
2802
2803
2804
2805
2806
2807
2808
2809
2810
2811
2812
2813
2814
2815
2816
2817
2818
2819
2820
2821
2822
2823
2824
2825
2826
2827
2828
2829
2830
2831
2832
2833
2834
2835
2836
2837
2838
2839
2840
2841
2842
2843
2844
2845
2846
2847
2848
2849
2850
2851
2852
2853
2854
2855
2856
2857
2858
2859
2860
2861
2862
2863
2864
2865
2866
2867
2868
2869
2870
2871
2872
2873
2874
2875
2876
2877
2878
2879
2880
2881
2882
2883
2884
2885
2886
2887
2888
2889
2890
2891
2892
2893
2894
2895
2896
2897
2898
2899
2900
2901
2902
2903
2904
2905
2906
2907
2908
2909
2910
2911
2912
2913
2914
2915
2916
2917
2918
2919
2920
2921
2922
2923
2924
2925
2926
2927
2928
2929
2930
2931
2932
2933
2934
2935
2936
2937
2938
2939
2940
2941
2942
2943
2944
2945
2946
2947
2948
2949
2950
2951
2952
2953
2954
2955
2956
2957
2958
2959
2960
2961
2962
2963
2964
2965
2966
2967
2968
2969
2970
2971
2972
2973
2974
2975
2976
2977
2978
2979
2980
2981
2982
2983
2984
2985
2986
2987
2988
2989
2990
2991
2992
2993
2994
2995
2996
2997
2998
2999
3000
3001
3002
3003
3004
3005
3006
3007
3008
3009
3010
3011
3012
3013
3014
3015
3016
3017
3018
3019
3020
3021
3022
3023
3024
3025
3026
3027
3028
3029
3030
3031
3032
3033
3034
3035
3036
3037
3038
3039
3040
3041
3042
3043
3044
3045
3046
3047
3048
3049
3050
3051
3052
3053
3054
3055
3056
3057
3058
3059
3060
3061
3062
3063
3064
3065
3066
3067
3068
3069
3070
3071
3072
3073
3074
3075
3076
3077
3078
3079
3080
3081
3082
3083
3084
3085
3086
3087
3088
3089
3090
3091
3092
3093
3094
3095
3096
3097
3098
3099
3100
3101
3102
3103
3104
3105
3106
3107
3108
3109
3110
3111
3112
3113
3114
3115
3116
3117
3118
3119
3120
3121
3122
3123
3124
3125
3126
3127
3128
3129
3130
3131
3132
3133
3134
3135
3136
3137
3138
3139
3140
3141
3142
3143
3144
3145
3146
3147
3148
3149
3150
3151
3152
3153
3154
3155
3156
3157
3158
3159
3160
3161
3162
3163
3164
3165
3166
3167
3168
3169
3170
3171
3172
3173
3174
3175
3176
3177
3178
3179
3180
3181
3182
3183
3184
3185
3186
3187
3188
3189
3190
3191
3192
3193
3194
3195
3196
3197
3198
3199
3200
3201
3202
3203
3204
3205
3206
3207
3208
3209
3210
3211
3212
3213
3214
3215
3216
3217
3218
3219
3220
3221
3222
3223
3224
3225
3226
3227
3228
3229
3230
3231
3232
3233
3234
3235
3236
3237
3238
3239
3240
3241
3242
3243
3244
3245
3246
3247
3248
3249
3250
3251
3252
3253
3254
3255
3256
3257
3258
3259
3260
3261
3262
3263
3264
3265
3266
3267
3268
3269
3270
3271
3272
3273
3274
3275
3276
3277
3278
3279
3280
3281
3282
3283
3284
3285
3286
3287
3288
3289
3290
3291
3292
3293
3294
3295
3296
3297
3298
3299
3300
3301
3302
3303
3304
3305
3306
3307
3308
3309
3310
3311
3312
3313
3314
3315
3316
3317
3318
3319
3320
3321
3322
3323
3324
3325
3326
3327
3328
3329
3330
3331
3332
3333
3334
3335
3336
3337
3338
3339
3340
3341
3342
3343
3344
3345
3346
3347
3348
3349
3350
3351
3352
3353
3354
3355
3356
3357
3358
3359
3360
3361
3362
3363
3364
3365
3366
3367
3368
3369
3370
3371
3372
3373
3374
3375
3376
3377
3378
3379
3380
3381
3382
3383
3384
3385
3386
3387
3388
3389
3390
3391
3392
3393
3394
3395
3396
3397
3398
3399
3400
3401
3402
3403
3404
3405
3406
3407
3408
3409
3410
3411
3412
3413
3414
3415
3416
3417
3418
3419
3420
3421
3422
3423
3424
3425
3426
3427
3428
3429
3430
3431
3432
3433
3434
3435
3436
3437
3438
3439
3440
3441
3442
3443
3444
3445
3446
3447
3448
3449
3450
3451
3452
3453
3454
3455
3456
3457
3458
3459
3460
3461
3462
3463
3464
3465
3466
3467
3468
3469
3470
3471
3472
3473
3474
3475
3476
3477
3478
3479
3480
3481
3482
3483
3484
3485
3486
3487
3488
3489
3490
3491
3492
3493
3494
3495
3496
3497
3498
3499
3500
3501
3502
3503
3504
3505
3506
3507
3508
3509
3510
3511
3512
3513
3514
3515
3516
3517
3518
3519
3520
3521
3522
3523
3524
3525
3526
3527
3528
3529
3530
3531
3532
3533
3534
3535
3536
3537
3538
3539
3540
3541
3542
3543
3544
3545
3546
3547
3548
3549
3550
3551
3552
3553
3554
3555
3556
3557
3558
3559
3560
3561
3562
3563
3564
3565
3566
3567
3568
3569
3570
3571
3572
3573
3574
3575
3576
3577
3578
3579
3580
3581
3582
3583
3584
3585
3586
3587
3588
3589
3590
3591
3592
3593
3594
3595
3596
3597
3598
3599
3600
3601
3602
3603
3604
3605
3606
3607
3608
3609
3610
3611
3612
3613
3614
3615
3616
3617
3618
3619
3620
3621
3622
3623
3624
3625
3626
3627
3628
3629
3630
3631
3632
3633
3634
3635
3636
3637
3638
3639
3640
3641
3642
3643
3644
3645
3646
3647
3648
3649
3650
3651
3652
3653
3654
3655
3656
3657
3658
3659
3660
3661
3662
3663
3664
3665
3666
3667
3668
3669
3670
3671
3672
3673
3674
3675
3676
3677
3678
3679
3680
3681
3682
3683
3684
3685
3686
3687
3688
3689
3690
3691
3692
3693
3694
3695
3696
3697
3698
3699
3700
3701
3702
3703
3704
3705
3706
3707
3708
3709
3710
3711
3712
3713
3714
3715
3716
3717
3718
3719
3720
3721
3722
3723
3724
3725
3726
3727
3728
3729
3730
3731
3732
3733
3734
3735
3736
3737
3738
3739
3740
3741
3742
3743
3744
3745
3746
3747
3748
3749
3750
3751
3752
3753
3754
3755
3756
3757
3758
3759
3760
3761
3762
3763
3764
3765
3766
3767
3768
3769
3770
3771
3772
3773
3774
3775
3776
3777
3778
3779
3780
3781
3782
3783
3784
3785
3786
3787
3788
3789
3790
3791
3792
3793
3794
3795
3796
3797
3798
3799
3800
3801
3802
3803
3804
3805
3806
3807
3808
3809
3810
3811
3812
3813
3814
3815
3816
3817
3818
3819
3820
3821
3822
3823
3824
3825
3826
3827
3828
3829
3830
3831
3832
3833
3834
3835
3836
3837
3838
3839
3840
3841
3842
3843
3844
3845
3846
3847
3848
3849
3850
3851
3852
3853
3854
3855
3856
3857
3858
3859
3860
3861
3862
3863
3864
3865
3866
3867
3868
3869
3870
3871
3872
3873
3874
3875
3876
3877
3878
3879
3880
3881
3882
3883
3884
3885
3886
3887
3888
3889
3890
3891
3892
3893
3894
3895
3896
3897
3898
3899
3900
3901
3902
3903
3904
3905
3906
3907
3908
3909
3910
3911
3912
3913
3914
3915
3916
3917
3918
3919
3920
3921
3922
3923
3924
3925
3926
3927
3928
3929
3930
3931
3932
3933
3934
3935
3936
3937
3938
3939
3940
3941
3942
3943
3944
3945
3946
3947
3948
3949
3950
3951
3952
3953
3954
3955
3956
3957
3958
3959
3960
3961
3962
3963
3964
3965
3966
3967
3968
3969
3970
3971
3972
3973
3974
3975
3976
3977
3978
3979
3980
3981
3982
3983
3984
3985
3986
3987
3988
3989
3990
3991
3992
3993
3994
3995
3996
3997
3998
3999
4000
4001
4002
4003
4004
4005
4006
4007
4008
4009
4010
4011
4012
4013
4014
4015
4016
4017
4018
4019
4020
4021
4022
4023
4024
4025
4026
4027
4028
4029
4030
4031
4032
4033
4034
4035
4036
4037
4038
4039
4040
4041
4042
4043
4044
4045
4046
4047
4048
4049
4050
4051
4052
4053
4054
4055
4056
4057
4058
4059
4060
4061
4062
4063
4064
4065
4066
4067
4068
4069
4070
4071
4072
4073
4074
4075
4076
4077
4078
4079
4080
4081
4082
4083
4084
4085
4086
4087
4088
4089
4090
4091
4092
4093
4094
4095
4096
4097
4098
4099
4100
4101
4102
4103
4104
4105
4106
4107
4108
4109
4110
4111
4112
4113
4114
4115
4116
4117
4118
4119
4120
4121
4122
4123
4124
4125
4126
4127
4128
4129
4130
4131
4132
4133
4134
4135
4136
4137
4138
4139
4140
4141
4142
4143
4144
4145
4146
4147
4148
4149
4150
4151
4152
4153
4154
4155
4156
4157
4158
4159
4160
4161
4162
4163
4164
4165
4166
4167
4168
4169
4170
4171
4172
4173
4174
4175
4176
4177
4178
4179
4180
4181
4182
4183
4184
4185
4186
4187
4188
4189
4190
4191
4192
4193
4194
4195
4196
4197
4198
4199
4200
4201
4202
4203
4204
4205
4206
4207
4208
4209
4210
4211
4212
4213
4214
4215
4216
4217
4218
4219
4220
4221
4222
4223
4224
4225
4226
4227
4228
4229
4230
4231
4232
4233
4234
4235
4236
4237
4238
4239
4240
4241
4242
4243
4244
4245
4246
4247
4248
4249
4250
4251
4252
4253
4254
4255
4256
4257
4258
4259
4260
4261
4262
4263
4264
4265
4266
4267
4268
4269
4270
4271
4272
4273
4274
4275
4276
4277
4278
4279
4280
4281
4282
4283
4284
4285
4286
4287
4288
4289
4290
4291
4292
4293
4294
4295
4296
4297
4298
4299
4300
4301
4302
4303
4304
4305
4306
4307
4308
4309
4310
4311
4312
4313
4314
4315
4316
4317
4318
4319
4320
4321
4322
4323
4324
4325
4326
4327
4328
4329
4330
4331
4332
4333
4334
4335
4336
4337
4338
4339
4340
4341
4342
4343
4344
4345
4346
4347
4348
4349
4350
4351
4352
4353
4354
4355
4356
4357
4358
4359
4360
4361
4362
4363
4364
4365
4366
4367
4368
4369
4370
4371
4372
4373
4374
4375
4376
4377
4378
4379
4380
4381
4382
4383
4384
4385
4386
4387
4388
4389
4390
4391
4392
4393
4394
4395
4396
4397
4398
4399
4400
4401
4402
4403
4404
4405
4406
4407
4408
4409
4410
4411
4412
4413
4414
4415
4416
4417
4418
4419
4420
4421
4422
4423
4424
4425
4426
4427
4428
4429
4430
4431
4432
4433
4434
4435
4436
4437
4438
4439
4440
4441
4442
4443
4444
4445
4446
4447
4448
4449
4450
4451
4452
4453
4454
4455
4456
4457
4458
4459
4460
4461
4462
4463
4464
4465
4466
4467
4468
4469
4470
4471
4472
4473
4474
4475
4476
4477
4478
4479
4480
4481
4482
4483
4484
4485
4486
4487
4488
4489
4490
4491
4492
4493
4494
4495
4496
4497
4498
4499
4500
4501
4502
4503
4504
4505
4506
4507
4508
4509
4510
4511
4512
4513
4514
4515
4516
4517
4518
4519
4520
4521
4522
4523
4524
4525
4526
4527
4528
4529
4530
4531
4532
4533
4534
4535
4536
4537
4538
4539
4540
4541
4542
4543
4544
4545
4546
4547
4548
4549
4550
4551
4552
4553
4554
4555
4556
4557
4558
4559
4560
4561
4562
4563
4564
4565
4566
4567
4568
4569
4570
4571
4572
4573
4574
4575
4576
4577
4578
4579
4580
4581
4582
4583
4584
4585
4586
4587
4588
4589
4590
4591
4592
4593
4594
4595
4596
4597
4598
4599
4600
4601
4602
4603
4604
4605
4606
4607
4608
4609
4610
4611
4612
4613
4614
4615
4616
4617
4618
4619
4620
4621
4622
4623
4624
4625
4626
4627
4628
4629
4630
4631
4632
4633
4634
4635
4636
4637
4638
4639
4640
4641
4642
4643
4644
4645
4646
4647
4648
4649
4650
4651
4652
4653
4654
4655
4656
4657
4658
4659
4660
4661
4662
4663
4664
4665
4666
4667
4668
4669
4670
4671
4672
4673
4674
4675
4676
4677
4678
4679
4680
4681
4682
4683
4684
4685
4686
4687
4688
4689
4690
4691
4692
4693
4694
4695
4696
4697
4698
4699
4700
4701
4702
4703
4704
4705
4706
4707
4708
4709
4710
4711
4712
4713
4714
4715
4716
4717
4718
4719
4720
4721
4722
4723
4724
4725
4726
4727
4728
4729
4730
4731
4732
4733
4734
4735
4736
4737
4738
4739
4740
4741
4742
4743
4744
4745
4746
4747
4748
4749
4750
4751
4752
4753
4754
4755
4756
4757
4758
4759
4760
4761
4762
4763
4764
4765
4766
4767
4768
4769
4770
4771
4772
4773
4774
4775
4776
4777
4778
4779
4780
4781
4782
4783
4784
4785
4786
4787
4788
4789
4790
4791
4792
4793
4794
4795
4796
4797
4798
4799
4800
4801
4802
4803
4804
4805
4806
4807
4808
4809
4810
4811
4812
4813
4814
4815
4816
4817
4818
4819
4820
4821
4822
4823
4824
4825
4826
4827
4828
4829
4830
4831
4832
4833
4834
4835
4836
4837
4838
4839
4840
4841
4842
4843
4844
4845
4846
4847
4848
4849
4850
4851
4852
4853
4854
4855
4856
4857
4858
4859
4860
4861
4862
4863
4864
4865
4866
4867
4868
4869
4870
4871
4872
4873
4874
4875
4876
4877
4878
4879
4880
4881
4882
4883
4884
4885
4886
4887
4888
4889
4890
4891
4892
4893
4894
4895
4896
4897
4898
4899
4900
4901
4902
4903
4904
4905
4906
4907
4908
4909
4910
4911
4912
4913
4914
4915
4916
4917
4918
4919
4920
4921
4922
4923
4924
4925
4926
4927
4928
4929
4930
4931
4932
4933
4934
4935
4936
4937
4938
4939
4940
4941
4942
4943
4944
4945
4946
4947
4948
4949
4950
4951
4952
4953
4954
4955
4956
4957
4958
4959
4960
4961
4962
4963
4964
4965
4966
4967
4968
4969
4970
4971
4972
4973
4974
4975
4976
4977
4978
4979
4980
4981
4982
4983
4984
4985
4986
4987
4988
4989
4990
4991
4992
4993
4994
4995
4996
4997
4998
4999
5000
5001
5002
5003
5004
5005
5006
5007
5008
5009
5010
5011
5012
5013
5014
5015
5016
5017
5018
5019
5020
5021
5022
5023
5024
5025
5026
5027
5028
5029
5030
5031
5032
5033
5034
5035
5036
5037
5038
5039
5040
5041
5042
5043
5044
5045
5046
5047
5048
5049
5050
5051
5052
5053
5054
5055
5056
5057
5058
5059
5060
5061
5062
5063
5064
5065
5066
5067
5068
5069
5070
5071
5072
5073
5074
5075
5076
5077
5078
5079
5080
5081
5082
5083
5084
5085
5086
5087
5088
5089
5090
5091
5092
5093
5094
5095
5096
5097
5098
5099
5100
5101
5102
5103
5104
5105
5106
5107
5108
5109
5110
5111
5112
5113
5114
5115
5116
5117
5118
5119
5120
5121
5122
5123
5124
5125
5126
5127
5128
5129
5130
5131
5132
5133
5134
5135
5136
5137
5138
5139
5140
5141
5142
5143
5144
5145
5146
5147
5148
5149
5150
5151
5152
5153
5154
5155
5156
5157
5158
5159
5160
5161
5162
5163
5164
5165
5166
5167
5168
5169
5170
5171
5172
5173
5174
5175
5176
5177
5178
5179
5180
5181
5182
5183
5184
5185
5186
5187
5188
5189
5190
5191
5192
5193
5194
5195
5196
5197
5198
5199
5200
5201
5202
5203
5204
5205
5206
5207
5208
5209
5210
5211
5212
5213
5214
5215
5216
5217
5218
5219
5220
5221
5222
5223
5224
5225
5226
5227
5228
5229
5230
5231
5232
5233
5234
5235
5236
5237
5238
5239
5240
5241
5242
5243
5244
5245
5246
5247
5248
5249
5250
5251
5252
5253
5254
5255
5256
5257
5258
5259
5260
5261
5262
5263
5264
5265
5266
5267
5268
5269
5270
5271
5272
5273
5274
5275
5276
5277
5278
5279
5280
5281
5282
5283
5284
5285
5286
5287
5288
5289
5290
5291
5292
5293
5294
5295
5296
5297
5298
5299
5300
5301
5302
5303
5304
5305
5306
5307
5308
5309
5310
5311
5312
5313
5314
5315
5316
5317
5318
5319
5320
5321
5322
5323
5324
5325
5326
5327
5328
5329
5330
5331
5332
5333
5334
5335
5336
5337
5338
5339
5340
5341
5342
5343
5344
5345
5346
5347
5348
5349
5350
5351
5352
5353
5354
5355
5356
5357
5358
5359
5360
5361
5362
5363
5364
5365
5366
5367
5368
5369
5370
5371
5372
5373
5374
5375
5376
5377
5378
5379
5380
5381
5382
5383
5384
5385
5386
5387
5388
5389
5390
5391
5392
5393
5394
5395
5396
5397
5398
5399
5400
5401
5402
5403
5404
5405
5406
5407
5408
5409
5410
5411
5412
5413
5414
5415
5416
5417
5418
5419
5420
5421
5422
5423
5424
5425
5426
5427
5428
5429
5430
5431
5432
5433
5434
5435
5436
5437
5438
5439
5440
5441
5442
5443
5444
5445
5446
5447
5448
5449
5450
5451
5452
5453
5454
5455
5456
5457
5458
5459
5460
5461
5462
5463
5464
5465
5466
5467
5468
5469
5470
5471
5472
5473
5474
5475
5476
5477
5478
5479
5480
5481
5482
5483
5484
5485
5486
5487
5488
5489
5490
5491
5492
5493
5494
5495
5496
5497
5498
5499
5500
5501
5502
5503
5504
5505
5506
5507
5508
5509
5510
5511
5512
5513
5514
5515
5516
5517
5518
5519
5520
5521
5522
5523
5524
5525
5526
5527
5528
5529
5530
5531
5532
5533
5534
5535
5536
5537
5538
5539
5540
5541
5542
5543
5544
5545
5546
5547
5548
5549
5550
5551
5552
5553
5554
5555
5556
5557
5558
5559
5560
5561
5562
5563
5564
5565
5566
5567
5568
5569
5570
5571
5572
5573
5574
5575
5576
5577
5578
5579
5580
5581
5582
5583
5584
5585
5586
5587
5588
5589
5590
5591
5592
5593
5594
5595
5596
5597
5598
5599
5600
5601
5602
5603
5604
5605
5606
5607
5608
5609
5610
5611
5612
5613
5614
5615
5616
5617
5618
5619
5620
5621
5622
5623
5624
5625
5626
5627
5628
5629
5630
5631
5632
5633
5634
5635
5636
5637
5638
5639
5640
5641
5642
5643
5644
5645
5646
5647
5648
5649
5650
5651
5652
5653
5654
5655
5656
5657
5658
5659
5660
5661
5662
5663
5664
5665
5666
5667
5668
5669
5670
5671
5672
5673
5674
5675
5676
5677
5678
5679
5680
5681
5682
5683
5684
5685
5686
5687
5688
5689
5690
5691
5692
5693
5694
5695
5696
5697
5698
5699
5700
5701
5702
5703
5704
5705
5706
5707
5708
5709
5710
5711
5712
5713
5714
5715
5716
5717
5718
5719
5720
5721
5722
5723
5724
5725
5726
5727
5728
5729
5730
5731
5732
5733
5734
5735
5736
5737
5738
5739
5740
5741
5742
5743
5744
5745
5746
5747
5748
5749
5750
5751
5752
5753
5754
5755
5756
5757
5758
5759
5760
5761
5762
5763
5764
5765
5766
5767
5768
5769
5770
5771
5772
5773
5774
5775
5776
5777
5778
5779
5780
5781
5782
5783
5784
5785
5786
5787
5788
5789
5790
5791
5792
5793
5794
5795
5796
5797
5798
5799
5800
5801
5802
5803
5804
5805
5806
5807
5808
5809
5810
5811
5812
5813
5814
5815
5816
5817
5818
5819
5820
5821
5822
5823
5824
5825
5826
5827
5828
5829
5830
5831
5832
5833
5834
5835
5836
5837
5838
5839
5840
5841
5842
5843
5844
5845
5846
5847
5848
5849
5850
5851
5852
5853
5854
5855
5856
5857
5858
5859
5860
5861
5862
5863
5864
5865
5866
5867
5868
5869
5870
5871
5872
5873
5874
5875
5876
5877
5878
5879
5880
5881
5882
5883
5884
5885
5886
5887
5888
5889
5890
5891
5892
5893
5894
5895
5896
5897
5898
5899
5900
5901
5902
5903
5904
5905
5906
5907
5908
5909
5910
5911
5912
5913
5914
5915
5916
5917
5918
5919
5920
5921
5922
5923
5924
5925
5926
5927
5928
5929
5930
5931
5932
5933
5934
5935
5936
5937
5938
5939
5940
5941
5942
5943
5944
5945
5946
5947
5948
5949
5950
5951
5952
5953
5954
5955
5956
5957
5958
5959
5960
5961
5962
5963
5964
5965
5966
5967
5968
5969
5970
5971
5972
5973
5974
/*****************************************************************************

Copyright (c) 1994, 2014, Oracle and/or its affiliates. All Rights Reserved.
Copyright (c) 2008, Google Inc.
Copyright (c) 2012, Facebook Inc.
Copyright (c) 2015, MariaDB Corporation.

Portions of this file contain modifications contributed and copyrighted by
Google, Inc. Those modifications are gratefully acknowledged and are described
briefly in the InnoDB documentation. The contributions by Google are
incorporated with their permission, and subject to the conditions contained in
the file COPYING.Google.

This program is free software; you can redistribute it and/or modify it under
the terms of the GNU General Public License as published by the Free Software
Foundation; version 2 of the License.

This program is distributed in the hope that it will be useful, but WITHOUT
ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS
FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details.

You should have received a copy of the GNU General Public License along with
this program; if not, write to the Free Software Foundation, Inc.,
51 Franklin Street, Suite 500, Boston, MA 02110-1335 USA

*****************************************************************************/

/**************************************************//**
@file btr/btr0cur.cc
The index tree cursor

All changes that row operations make to a B-tree or the records
there must go through this module! Undo log records are written here
of every modify or insert of a clustered index record.

			NOTE!!!
To make sure we do not run out of disk space during a pessimistic
insert or update, we have to reserve 2 x the height of the index tree
many pages in the tablespace before we start the operation, because
if leaf splitting has been started, it is difficult to undo, except
by crashing the database and doing a roll-forward.

Created 10/16/1994 Heikki Tuuri
*******************************************************/

#include "btr0cur.h"

#ifdef UNIV_NONINL
#include "btr0cur.ic"
#endif

#include "row0upd.h"
#ifndef UNIV_HOTBACKUP
#include "mtr0log.h"
#include "page0page.h"
#include "page0zip.h"
#include "rem0rec.h"
#include "rem0cmp.h"
#include "buf0lru.h"
#include "btr0btr.h"
#include "btr0sea.h"
#include "row0log.h"
#include "row0purge.h"
#include "row0upd.h"
#include "trx0rec.h"
#include "trx0roll.h" /* trx_is_recv() */
#include "que0que.h"
#include "row0row.h"
#include "srv0srv.h"
#include "ibuf0ibuf.h"
#include "lock0lock.h"
#include "zlib.h"

/** Buffered B-tree operation types, introduced as part of delete buffering. */
enum btr_op_t {
	BTR_NO_OP = 0,			/*!< Not buffered */
	BTR_INSERT_OP,			/*!< Insert, do not ignore UNIQUE */
	BTR_INSERT_IGNORE_UNIQUE_OP,	/*!< Insert, ignoring UNIQUE */
	BTR_DELETE_OP,			/*!< Purge a delete-marked record */
	BTR_DELMARK_OP			/*!< Mark a record for deletion */
};

#ifdef UNIV_DEBUG
/** If the following is set to TRUE, this module prints a lot of
trace information of individual record operations */
UNIV_INTERN ibool	btr_cur_print_record_ops = FALSE;
#endif /* UNIV_DEBUG */

/** Number of searches down the B-tree in btr_cur_search_to_nth_level(). */
UNIV_INTERN ulint	btr_cur_n_non_sea	= 0;
/** Number of successful adaptive hash index lookups in
btr_cur_search_to_nth_level(). */
UNIV_INTERN ulint	btr_cur_n_sea		= 0;
/** Old value of btr_cur_n_non_sea.  Copied by
srv_refresh_innodb_monitor_stats().  Referenced by
srv_printf_innodb_monitor(). */
UNIV_INTERN ulint	btr_cur_n_non_sea_old	= 0;
/** Old value of btr_cur_n_sea.  Copied by
srv_refresh_innodb_monitor_stats().  Referenced by
srv_printf_innodb_monitor(). */
UNIV_INTERN ulint	btr_cur_n_sea_old	= 0;

#ifdef UNIV_DEBUG
/* Flag to limit optimistic insert records */
UNIV_INTERN uint	btr_cur_limit_optimistic_insert_debug = 0;
#endif /* UNIV_DEBUG */

/** In the optimistic insert, if the insert does not fit, but this much space
can be released by page reorganize, then it is reorganized */
#define BTR_CUR_PAGE_REORGANIZE_LIMIT	(UNIV_PAGE_SIZE / 32)

/** The structure of a BLOB part header */
/* @{ */
/*--------------------------------------*/
#define BTR_BLOB_HDR_PART_LEN		0	/*!< BLOB part len on this
						page */
#define BTR_BLOB_HDR_NEXT_PAGE_NO	4	/*!< next BLOB part page no,
						FIL_NULL if none */
/*--------------------------------------*/
#define BTR_BLOB_HDR_SIZE		8	/*!< Size of a BLOB
						part header, in bytes */

/** Estimated table level stats from sampled value.
@param value		sampled stats
@param index		index being sampled
@param sample		number of sampled rows
@param ext_size		external stored data size
@param not_empty	table not empty
@return estimated table wide stats from sampled value */
#define BTR_TABLE_STATS_FROM_SAMPLE(value, index, sample, ext_size, not_empty)\
	(((value) * (ib_int64_t) index->stat_n_leaf_pages		\
	  + (sample) - 1 + (ext_size) + (not_empty)) / ((sample) + (ext_size)))

/* @} */
#endif /* !UNIV_HOTBACKUP */

/** A BLOB field reference full of zero, for use in assertions and tests.
Initially, BLOB field references are set to zero, in
dtuple_convert_big_rec(). */
const byte field_ref_zero[BTR_EXTERN_FIELD_REF_SIZE] = {
	0, 0, 0, 0, 0,
	0, 0, 0, 0, 0,
	0, 0, 0, 0, 0,
	0, 0, 0, 0, 0,
};

#ifndef UNIV_HOTBACKUP
/*******************************************************************//**
Marks all extern fields in a record as owned by the record. This function
should be called if the delete mark of a record is removed: a not delete
marked record always owns all its extern fields. */
static
void
btr_cur_unmark_extern_fields(
/*=========================*/
	page_zip_des_t*	page_zip,/*!< in/out: compressed page whose uncompressed
				part will be updated, or NULL */
	rec_t*		rec,	/*!< in/out: record in a clustered index */
	dict_index_t*	index,	/*!< in: index of the page */
	const ulint*	offsets,/*!< in: array returned by rec_get_offsets() */
	mtr_t*		mtr);	/*!< in: mtr, or NULL if not logged */
/*******************************************************************//**
Adds path information to the cursor for the current page, for which
the binary search has been performed. */
static
void
btr_cur_add_path_info(
/*==================*/
	btr_cur_t*	cursor,		/*!< in: cursor positioned on a page */
	ulint		height,		/*!< in: height of the page in tree;
					0 means leaf node */
	ulint		root_height);	/*!< in: root node height in tree */
/***********************************************************//**
Frees the externally stored fields for a record, if the field is mentioned
in the update vector. */
static
void
btr_rec_free_updated_extern_fields(
/*===============================*/
	dict_index_t*	index,	/*!< in: index of rec; the index tree MUST be
				X-latched */
	rec_t*		rec,	/*!< in: record */
	page_zip_des_t*	page_zip,/*!< in: compressed page whose uncompressed
				part will be updated, or NULL */
	const ulint*	offsets,/*!< in: rec_get_offsets(rec, index) */
	const upd_t*	update,	/*!< in: update vector */
	enum trx_rb_ctx	rb_ctx,	/*!< in: rollback context */
	mtr_t*		mtr);	/*!< in: mini-transaction handle which contains
				an X-latch to record page and to the tree */
/***********************************************************//**
Frees the externally stored fields for a record. */
static
void
btr_rec_free_externally_stored_fields(
/*==================================*/
	dict_index_t*	index,	/*!< in: index of the data, the index
				tree MUST be X-latched */
	rec_t*		rec,	/*!< in: record */
	const ulint*	offsets,/*!< in: rec_get_offsets(rec, index) */
	page_zip_des_t*	page_zip,/*!< in: compressed page whose uncompressed
				part will be updated, or NULL */
	enum trx_rb_ctx	rb_ctx,	/*!< in: rollback context */
	mtr_t*		mtr);	/*!< in: mini-transaction handle which contains
				an X-latch to record page and to the index
				tree */
#endif /* !UNIV_HOTBACKUP */

/******************************************************//**
The following function is used to set the deleted bit of a record. */
UNIV_INLINE
void
btr_rec_set_deleted_flag(
/*=====================*/
	rec_t*		rec,	/*!< in/out: physical record */
	page_zip_des_t*	page_zip,/*!< in/out: compressed page (or NULL) */
	ulint		flag)	/*!< in: nonzero if delete marked */
{
	if (page_rec_is_comp(rec)) {
		rec_set_deleted_flag_new(rec, page_zip, flag);
	} else {
		ut_ad(!page_zip);
		rec_set_deleted_flag_old(rec, flag);
	}
}

#ifndef UNIV_HOTBACKUP
/*==================== B-TREE SEARCH =========================*/

/********************************************************************//**
Latches the leaf page or pages requested. */
static
void
btr_cur_latch_leaves(
/*=================*/
	page_t*		page,		/*!< in: leaf page where the search
					converged */
	ulint		space,		/*!< in: space id */
	ulint		zip_size,	/*!< in: compressed page size in bytes
					or 0 for uncompressed pages */
	ulint		page_no,	/*!< in: page number of the leaf */
	ulint		latch_mode,	/*!< in: BTR_SEARCH_LEAF, ... */
	btr_cur_t*	cursor,		/*!< in: cursor */
	mtr_t*		mtr)		/*!< in: mtr */
{
	ulint		mode;
	ulint		sibling_mode;
	ulint		left_page_no;
	ulint		right_page_no;
	buf_block_t*	get_block;

	ut_ad(page && mtr);

	switch (latch_mode) {
	case BTR_SEARCH_LEAF:
	case BTR_MODIFY_LEAF:
		mode = latch_mode == BTR_SEARCH_LEAF ? RW_S_LATCH : RW_X_LATCH;
		get_block = btr_block_get(
			space, zip_size, page_no, mode, cursor->index, mtr);

		SRV_CORRUPT_TABLE_CHECK(get_block, return;);

#ifdef UNIV_BTR_DEBUG
		ut_a(page_is_comp(get_block->frame) == page_is_comp(page));
#endif /* UNIV_BTR_DEBUG */
		get_block->check_index_page_at_flush = TRUE;
		return;
	case BTR_SEARCH_TREE:
	case BTR_MODIFY_TREE:
		if (UNIV_UNLIKELY(latch_mode == BTR_SEARCH_TREE)) {
			mode = RW_S_LATCH;
			sibling_mode = RW_NO_LATCH;
		} else {
			mode = sibling_mode = RW_X_LATCH;
		}
		/* Fetch and possibly latch also brothers from left to right */
		left_page_no = btr_page_get_prev(page, mtr);

		if (left_page_no != FIL_NULL) {
			get_block = btr_block_get(
				space, zip_size, left_page_no,
				sibling_mode, cursor->index, mtr);

			SRV_CORRUPT_TABLE_CHECK(get_block, return;);

#ifdef UNIV_BTR_DEBUG
			ut_a(page_is_comp(get_block->frame)
			     == page_is_comp(page));
			ut_a(btr_page_get_next(get_block->frame, mtr)
			     == page_get_page_no(page));
#endif /* UNIV_BTR_DEBUG */
			if (sibling_mode == RW_NO_LATCH) {
				/* btr_block_get() called with RW_NO_LATCH will
				fix the read block in the buffer.  This serves
				no purpose for the fake changes prefetching,
				thus we unfix the sibling blocks immediately.*/
				mtr_memo_release(mtr, get_block,
						 MTR_MEMO_BUF_FIX);
			} else {
				get_block->check_index_page_at_flush = TRUE;
			}
		}

		get_block = btr_block_get(
			space, zip_size, page_no,
			mode, cursor->index, mtr);

		SRV_CORRUPT_TABLE_CHECK(get_block, return;);

#ifdef UNIV_BTR_DEBUG
		ut_a(page_is_comp(get_block->frame) == page_is_comp(page));
#endif /* UNIV_BTR_DEBUG */
		get_block->check_index_page_at_flush = TRUE;

		right_page_no = btr_page_get_next(page, mtr);

		if (right_page_no != FIL_NULL) {
			get_block = btr_block_get(
				space, zip_size, right_page_no,
				sibling_mode, cursor->index, mtr);

			SRV_CORRUPT_TABLE_CHECK(get_block, return;);

#ifdef UNIV_BTR_DEBUG
			ut_a(page_is_comp(get_block->frame)
			     == page_is_comp(page));
			ut_a(btr_page_get_prev(get_block->frame, mtr)
			     == page_get_page_no(page));
#endif /* UNIV_BTR_DEBUG */
			if (sibling_mode == RW_NO_LATCH) {
				mtr_memo_release(mtr, get_block,
						 MTR_MEMO_BUF_FIX);
			} else {
				get_block->check_index_page_at_flush = TRUE;
			}
		}

		return;

	case BTR_SEARCH_PREV:
	case BTR_MODIFY_PREV:
		mode = latch_mode == BTR_SEARCH_PREV ? RW_S_LATCH : RW_X_LATCH;
		/* latch also left brother */
		left_page_no = btr_page_get_prev(page, mtr);

		if (left_page_no != FIL_NULL) {
			get_block = btr_block_get(
				space, zip_size,
				left_page_no, mode, cursor->index, mtr);
			cursor->left_block = get_block;

			SRV_CORRUPT_TABLE_CHECK(get_block, return;);

#ifdef UNIV_BTR_DEBUG
			ut_a(page_is_comp(get_block->frame)
			     == page_is_comp(page));
			ut_a(btr_page_get_next(get_block->frame, mtr)
			     == page_get_page_no(page));
#endif /* UNIV_BTR_DEBUG */
			get_block->check_index_page_at_flush = TRUE;
		}

		get_block = btr_block_get(
			space, zip_size, page_no, mode, cursor->index, mtr);

		SRV_CORRUPT_TABLE_CHECK(get_block, return;);

#ifdef UNIV_BTR_DEBUG
		ut_a(page_is_comp(get_block->frame) == page_is_comp(page));
#endif /* UNIV_BTR_DEBUG */
		get_block->check_index_page_at_flush = TRUE;
		return;
	}

	ut_error;
}

/********************************************************************//**
Searches an index tree and positions a tree cursor on a given level.
NOTE: n_fields_cmp in tuple must be set so that it cannot be compared
to node pointer page number fields on the upper levels of the tree!
Note that if mode is PAGE_CUR_LE, which is used in inserts, then
cursor->up_match and cursor->low_match both will have sensible values.
If mode is PAGE_CUR_GE, then up_match will a have a sensible value.

If mode is PAGE_CUR_LE , cursor is left at the place where an insert of the
search tuple should be performed in the B-tree. InnoDB does an insert
immediately after the cursor. Thus, the cursor may end up on a user record,
or on a page infimum record. */
UNIV_INTERN
dberr_t
btr_cur_search_to_nth_level(
/*========================*/
	dict_index_t*	index,	/*!< in: index */
	ulint		level,	/*!< in: the tree level of search */
	const dtuple_t*	tuple,	/*!< in: data tuple; NOTE: n_fields_cmp in
				tuple must be set so that it cannot get
				compared to the node ptr page number field! */
	ulint		mode,	/*!< in: PAGE_CUR_L, ...;
				Inserts should always be made using
				PAGE_CUR_LE to search the position! */
	ulint		latch_mode, /*!< in: BTR_SEARCH_LEAF, ..., ORed with
				at most one of BTR_INSERT, BTR_DELETE_MARK,
				BTR_DELETE, or BTR_ESTIMATE;
				cursor->left_block is used to store a pointer
				to the left neighbor page, in the cases
				BTR_SEARCH_PREV and BTR_MODIFY_PREV;
				NOTE that if has_search_latch
				is != 0, we maybe do not have a latch set
				on the cursor page, we assume
				the caller uses his search latch
				to protect the record! */
	btr_cur_t*	cursor, /*!< in/out: tree cursor; the cursor page is
				s- or x-latched, but see also above! */
	ulint		has_search_latch,/*!< in: info on the latch mode the
				caller currently has on btr_search_latch:
				RW_S_LATCH, or 0 */
	const char*	file,	/*!< in: file name */
	ulint		line,	/*!< in: line where called */
	mtr_t*		mtr)	/*!< in: mtr */
{
	page_t*		page;
	buf_block_t*	block;
	ulint		space;
	buf_block_t*	guess;
	ulint		height;
	ulint		page_no;
	ulint		up_match;
	ulint		up_bytes;
	ulint		low_match;
	ulint		low_bytes;
	ulint		savepoint;
	ulint		rw_latch;
	ulint		page_mode;
	ulint		buf_mode;
	ulint		estimate;
	ulint		zip_size;
	page_cur_t*	page_cursor;
	btr_op_t	btr_op;
	ulint		root_height = 0; /* remove warning */
	dberr_t		err = DB_SUCCESS;

#ifdef BTR_CUR_ADAPT
	btr_search_t*	info;
#endif
	mem_heap_t*	heap		= NULL;
	ulint		offsets_[REC_OFFS_NORMAL_SIZE];
	ulint*		offsets		= offsets_;
	rec_offs_init(offsets_);
	/* Currently, PAGE_CUR_LE is the only search mode used for searches
	ending to upper levels */

	ut_ad(level == 0 || mode == PAGE_CUR_LE);
	ut_ad(dict_index_check_search_tuple(index, tuple));
	ut_ad(!dict_index_is_ibuf(index) || ibuf_inside(mtr));
	ut_ad(dtuple_check_typed(tuple));
	ut_ad(!(index->type & DICT_FTS));
	ut_ad(index->page != FIL_NULL);

	UNIV_MEM_INVALID(&cursor->up_match, sizeof cursor->up_match);
	UNIV_MEM_INVALID(&cursor->up_bytes, sizeof cursor->up_bytes);
	UNIV_MEM_INVALID(&cursor->low_match, sizeof cursor->low_match);
	UNIV_MEM_INVALID(&cursor->low_bytes, sizeof cursor->low_bytes);
#ifdef UNIV_DEBUG
	cursor->up_match = ULINT_UNDEFINED;
	cursor->low_match = ULINT_UNDEFINED;
#endif

	ibool	s_latch_by_caller;

	s_latch_by_caller = latch_mode & BTR_ALREADY_S_LATCHED;

	ut_ad(!s_latch_by_caller
	      || mtr_memo_contains(mtr, dict_index_get_lock(index),
				   MTR_MEMO_S_LOCK));

	/* These flags are mutually exclusive, they are lumped together
	with the latch mode for historical reasons. It's possible for
	none of the flags to be set. */
	switch (UNIV_EXPECT(latch_mode
			    & (BTR_INSERT | BTR_DELETE | BTR_DELETE_MARK),
			    0)) {
	case 0:
		btr_op = BTR_NO_OP;
		break;
	case BTR_INSERT:
		btr_op = (latch_mode & BTR_IGNORE_SEC_UNIQUE)
			? BTR_INSERT_IGNORE_UNIQUE_OP
			: BTR_INSERT_OP;
		break;
	case BTR_DELETE:
		btr_op = BTR_DELETE_OP;
		ut_a(cursor->purge_node);
		break;
	case BTR_DELETE_MARK:
		btr_op = BTR_DELMARK_OP;
		break;
	default:
		/* only one of BTR_INSERT, BTR_DELETE, BTR_DELETE_MARK
		should be specified at a time */
		ut_error;
	}

	/* Operations on the insert buffer tree cannot be buffered. */
	ut_ad(btr_op == BTR_NO_OP || !dict_index_is_ibuf(index));
	/* Operations on the clustered index cannot be buffered. */
	ut_ad(btr_op == BTR_NO_OP || !dict_index_is_clust(index));

	estimate = latch_mode & BTR_ESTIMATE;

	/* Turn the flags unrelated to the latch mode off. */
	latch_mode = BTR_LATCH_MODE_WITHOUT_FLAGS(latch_mode);

	ut_ad(!s_latch_by_caller
	      || latch_mode == BTR_SEARCH_LEAF
	      || latch_mode == BTR_MODIFY_LEAF);

	cursor->flag = BTR_CUR_BINARY;
	cursor->index = index;

#ifndef BTR_CUR_ADAPT
	guess = NULL;
#else
	info = btr_search_get_info(index);

	guess = info->root_guess;

#ifdef BTR_CUR_HASH_ADAPT

# ifdef UNIV_SEARCH_PERF_STAT
	info->n_searches++;
# endif
	if (rw_lock_get_writer(btr_search_get_latch(cursor->index)) ==
	    RW_LOCK_NOT_LOCKED
	    && latch_mode <= BTR_MODIFY_LEAF
	    && info->last_hash_succ
	    && !estimate
# ifdef PAGE_CUR_LE_OR_EXTENDS
	    && mode != PAGE_CUR_LE_OR_EXTENDS
# endif /* PAGE_CUR_LE_OR_EXTENDS */
	    /* If !has_search_latch, we do a dirty read of
	    btr_search_enabled below, and btr_search_guess_on_hash()
	    will have to check it again. */
	    && UNIV_LIKELY(btr_search_enabled)
	    && btr_search_guess_on_hash(index, info, tuple, mode,
					latch_mode, cursor,
					has_search_latch, mtr)) {

		/* Search using the hash index succeeded */

		ut_ad(cursor->up_match != ULINT_UNDEFINED
		      || mode != PAGE_CUR_GE);
		ut_ad(cursor->up_match != ULINT_UNDEFINED
		      || mode != PAGE_CUR_LE);
		ut_ad(cursor->low_match != ULINT_UNDEFINED
		      || mode != PAGE_CUR_LE);
		btr_cur_n_sea++;

		return err;
	}
# endif /* BTR_CUR_HASH_ADAPT */
#endif /* BTR_CUR_ADAPT */
	btr_cur_n_non_sea++;

	/* If the hash search did not succeed, do binary search down the
	tree */

	if (has_search_latch) {
		/* Release possible search latch to obey latching order */
		rw_lock_s_unlock(btr_search_get_latch(cursor->index));
	}

	/* Store the position of the tree latch we push to mtr so that we
	know how to release it when we have latched leaf node(s) */

	savepoint = mtr_set_savepoint(mtr);

	switch (latch_mode) {
	case BTR_MODIFY_TREE:
		mtr_x_lock(dict_index_get_lock(index), mtr);
		break;
	case BTR_CONT_MODIFY_TREE:
		/* Do nothing */
		ut_ad(mtr_memo_contains(mtr, dict_index_get_lock(index),
					MTR_MEMO_X_LOCK));
		break;
	default:
		if (!s_latch_by_caller) {
			mtr_s_lock(dict_index_get_lock(index), mtr);
		}
	}

	page_cursor = btr_cur_get_page_cur(cursor);

	space = dict_index_get_space(index);
	page_no = dict_index_get_page(index);

	up_match = 0;
	up_bytes = 0;
	low_match = 0;
	low_bytes = 0;

	height = ULINT_UNDEFINED;

	/* We use these modified search modes on non-leaf levels of the
	B-tree. These let us end up in the right B-tree leaf. In that leaf
	we use the original search mode. */

	switch (mode) {
	case PAGE_CUR_GE:
		page_mode = PAGE_CUR_L;
		break;
	case PAGE_CUR_G:
		page_mode = PAGE_CUR_LE;
		break;
	default:
#ifdef PAGE_CUR_LE_OR_EXTENDS
		ut_ad(mode == PAGE_CUR_L || mode == PAGE_CUR_LE
		      || mode == PAGE_CUR_LE_OR_EXTENDS);
#else /* PAGE_CUR_LE_OR_EXTENDS */
		ut_ad(mode == PAGE_CUR_L || mode == PAGE_CUR_LE);
#endif /* PAGE_CUR_LE_OR_EXTENDS */
		page_mode = mode;
		break;
	}

	/* Loop and search until we arrive at the desired level */

search_loop:
	buf_mode = BUF_GET;
	rw_latch = RW_NO_LATCH;

	if (height != 0) {
		/* We are about to fetch the root or a non-leaf page. */
	} else if (latch_mode <= BTR_MODIFY_LEAF) {
		rw_latch = latch_mode;

		if (btr_op != BTR_NO_OP
		    && ibuf_should_try(index, btr_op != BTR_INSERT_OP)) {

			/* Try to buffer the operation if the leaf
			page is not in the buffer pool. */

			buf_mode = btr_op == BTR_DELETE_OP
				? BUF_GET_IF_IN_POOL_OR_WATCH
				: BUF_GET_IF_IN_POOL;
		}
	}

	zip_size = dict_table_zip_size(index->table);

retry_page_get:
	block = buf_page_get_gen(
		space, zip_size, page_no, rw_latch, guess, buf_mode,
		file, line, mtr, &err);

	if (err != DB_SUCCESS) {
		if (err == DB_DECRYPTION_FAILED) {
			ib_push_warning((void *)NULL,
				DB_DECRYPTION_FAILED,
				"Table %s is encrypted but encryption service or"
				" used key_id is not available. "
				" Can't continue reading table.",
				index->table->name);
			index->table->is_encrypted = true;
		}

		goto func_exit;
	}

	if (block == NULL) {
		SRV_CORRUPT_TABLE_CHECK(buf_mode == BUF_GET_IF_IN_POOL ||
					buf_mode == BUF_GET_IF_IN_POOL_OR_WATCH,
			{
				page_cursor->block = 0;
				page_cursor->rec = 0;
				if (estimate) {

					cursor->path_arr->nth_rec =
						ULINT_UNDEFINED;
				}

				goto func_exit;
			});

		/* This must be a search to perform an insert/delete
		mark/ delete; try using the insert/delete buffer */

		ut_ad(height == 0);
		ut_ad(cursor->thr);

		switch (btr_op) {
		case BTR_INSERT_OP:
		case BTR_INSERT_IGNORE_UNIQUE_OP:
			ut_ad(buf_mode == BUF_GET_IF_IN_POOL);

			if (ibuf_insert(IBUF_OP_INSERT, tuple, index,
					space, zip_size, page_no,
					cursor->thr)) {

				cursor->flag = BTR_CUR_INSERT_TO_IBUF;

				goto func_exit;
			}
			break;

		case BTR_DELMARK_OP:
			ut_ad(buf_mode == BUF_GET_IF_IN_POOL);

			if (ibuf_insert(IBUF_OP_DELETE_MARK, tuple,
					index, space, zip_size,
					page_no, cursor->thr)) {

				cursor->flag = BTR_CUR_DEL_MARK_IBUF;

				goto func_exit;
			}

			break;

		case BTR_DELETE_OP:
			ut_ad(buf_mode == BUF_GET_IF_IN_POOL_OR_WATCH);

			if (!row_purge_poss_sec(cursor->purge_node,
						index, tuple)) {

				/* The record cannot be purged yet. */
				cursor->flag = BTR_CUR_DELETE_REF;
			} else if (ibuf_insert(IBUF_OP_DELETE, tuple,
					       index, space, zip_size,
					       page_no,
					       cursor->thr)) {

				/* The purge was buffered. */
				cursor->flag = BTR_CUR_DELETE_IBUF;
			} else {
				/* The purge could not be buffered. */
				buf_pool_watch_unset(space, page_no);
				break;
			}

			buf_pool_watch_unset(space, page_no);
			goto func_exit;

		default:
			ut_error;
		}

		/* Insert to the insert/delete buffer did not succeed, we
		must read the page from disk. */

		buf_mode = BUF_GET;

		goto retry_page_get;
	}

	block->check_index_page_at_flush = TRUE;
	page = buf_block_get_frame(block);

	SRV_CORRUPT_TABLE_CHECK(page,
	{
		page_cursor->block = 0;
		page_cursor->rec = 0;

		if (estimate) {

			cursor->path_arr->nth_rec = ULINT_UNDEFINED;
		}

		goto func_exit;
	});

	if (rw_latch != RW_NO_LATCH) {
#ifdef UNIV_ZIP_DEBUG
		const page_zip_des_t*	page_zip
			= buf_block_get_page_zip(block);
		ut_a(!page_zip || page_zip_validate(page_zip, page, index));
#endif /* UNIV_ZIP_DEBUG */

		buf_block_dbg_add_level(
			block, dict_index_is_ibuf(index)
			? SYNC_IBUF_TREE_NODE : SYNC_TREE_NODE);
	}

	ut_ad(fil_page_get_type(page) == FIL_PAGE_INDEX);
	ut_ad(index->id == btr_page_get_index_id(page));

	if (UNIV_UNLIKELY(height == ULINT_UNDEFINED)) {
		/* We are in the root node */

		height = btr_page_get_level(page, mtr);
		root_height = height;
		cursor->tree_height = root_height + 1;

#ifdef BTR_CUR_ADAPT
		if (block != guess) {
			info->root_guess = block;
		}
#endif
	}

	if (height == 0) {
		if (rw_latch == RW_NO_LATCH) {

			btr_cur_latch_leaves(
				page, space, zip_size, page_no, latch_mode,
				cursor, mtr);
		}

		switch (latch_mode) {
		case BTR_MODIFY_TREE:
		case BTR_CONT_MODIFY_TREE:
			break;
		default:
			if (!s_latch_by_caller) {
				/* Release the tree s-latch */
				mtr_release_s_latch_at_savepoint(
					mtr, savepoint,
					dict_index_get_lock(index));
			}
		}

		page_mode = mode;
	}

	page_cur_search_with_match(
		block, index, tuple, page_mode, &up_match, &up_bytes,
		&low_match, &low_bytes, page_cursor);

	if (estimate) {
		btr_cur_add_path_info(cursor, height, root_height);
	}

	/* If this is the desired level, leave the loop */

	ut_ad(height == btr_page_get_level(page_cur_get_page(page_cursor),
					   mtr));

	if (level != height) {

		const rec_t*	node_ptr;
		ut_ad(height > 0);

		height--;
		guess = NULL;

		node_ptr = page_cur_get_rec(page_cursor);

		offsets = rec_get_offsets(
			node_ptr, index, offsets, ULINT_UNDEFINED, &heap);

		/* Go to the child node */
		page_no = btr_node_ptr_get_child_page_no(node_ptr, offsets);

		if (UNIV_UNLIKELY(height == 0 && dict_index_is_ibuf(index))) {
			/* We're doing a search on an ibuf tree and we're one
			level above the leaf page. */

			ut_ad(level == 0);

			buf_mode = BUF_GET;
			rw_latch = RW_NO_LATCH;
			goto retry_page_get;
		}

		goto search_loop;
	}

	if (level != 0) {
		/* x-latch the page */
		buf_block_t*	child_block = btr_block_get(
			space, zip_size, page_no, RW_X_LATCH, index, mtr);

		page = buf_block_get_frame(child_block);
		btr_assert_not_corrupted(child_block, index);
	} else {
		cursor->low_match = low_match;
		cursor->low_bytes = low_bytes;
		cursor->up_match = up_match;
		cursor->up_bytes = up_bytes;

#ifdef BTR_CUR_ADAPT
		/* We do a dirty read of btr_search_enabled here.  We
		will properly check btr_search_enabled again in
		btr_search_build_page_hash_index() before building a
		page hash index, while holding btr_search_latch. */
		if (btr_search_enabled) {
			btr_search_info_update(index, cursor);
		}
#endif
		ut_ad(cursor->up_match != ULINT_UNDEFINED
		      || mode != PAGE_CUR_GE);
		ut_ad(cursor->up_match != ULINT_UNDEFINED
		      || mode != PAGE_CUR_LE);
		ut_ad(cursor->low_match != ULINT_UNDEFINED
		      || mode != PAGE_CUR_LE);
	}

func_exit:

	if (UNIV_LIKELY_NULL(heap)) {
		mem_heap_free(heap);
	}

	if (has_search_latch) {

		rw_lock_s_lock(btr_search_get_latch(cursor->index));
	}

	return err;
}

/*****************************************************************//**
Opens a cursor at either end of an index. */
UNIV_INTERN
dberr_t
btr_cur_open_at_index_side_func(
/*============================*/
	bool		from_left,	/*!< in: true if open to the low end,
					false if to the high end */
	dict_index_t*	index,		/*!< in: index */
	ulint		latch_mode,	/*!< in: latch mode */
	btr_cur_t*	cursor,		/*!< in/out: cursor */
	ulint		level,		/*!< in: level to search for
					(0=leaf). */
	const char*	file,		/*!< in: file name */
	ulint		line,		/*!< in: line where called */
	mtr_t*		mtr)		/*!< in/out: mini-transaction */
{
	page_cur_t*	page_cursor;
	ulint		page_no;
	ulint		space;
	ulint		zip_size;
	ulint		height;
	ulint		root_height = 0; /* remove warning */
	rec_t*		node_ptr;
	ulint		estimate;
	ulint		savepoint;
	mem_heap_t*	heap		= NULL;
	ulint		offsets_[REC_OFFS_NORMAL_SIZE];
	ulint*		offsets		= offsets_;
	dberr_t		err = DB_SUCCESS;
	
	rec_offs_init(offsets_);

	estimate = latch_mode & BTR_ESTIMATE;
	latch_mode &= ~BTR_ESTIMATE;

	ut_ad(level != ULINT_UNDEFINED);

	/* Store the position of the tree latch we push to mtr so that we
	know how to release it when we have latched the leaf node */

	savepoint = mtr_set_savepoint(mtr);

	switch (latch_mode) {
	case BTR_CONT_MODIFY_TREE:
		break;
	case BTR_MODIFY_TREE:
		mtr_x_lock(dict_index_get_lock(index), mtr);
		break;
	case BTR_SEARCH_LEAF | BTR_ALREADY_S_LATCHED:
	case BTR_MODIFY_LEAF | BTR_ALREADY_S_LATCHED:
		ut_ad(mtr_memo_contains(mtr, dict_index_get_lock(index),
					MTR_MEMO_S_LOCK));
		break;
	default:
		mtr_s_lock(dict_index_get_lock(index), mtr);
	}

	page_cursor = btr_cur_get_page_cur(cursor);
	cursor->index = index;

	space = dict_index_get_space(index);
	zip_size = dict_table_zip_size(index->table);
	page_no = dict_index_get_page(index);

	height = ULINT_UNDEFINED;

	for (;;) {
		buf_block_t*	block=NULL;
		page_t*		page=NULL;

		block = buf_page_get_gen(space, zip_size, page_no,
					 RW_NO_LATCH, NULL, BUF_GET,
					 file, line, mtr, &err);
		if (err != DB_SUCCESS) {
			if (err == DB_DECRYPTION_FAILED) {
				ib_push_warning((void *)NULL,
					DB_DECRYPTION_FAILED,
					"Table %s is encrypted but encryption service or"
					" used key_id is not available. "
					" Can't continue reading table.",
					index->table->name);
				index->table->is_encrypted = true;
			}

			goto exit_loop;
		}

		page = buf_block_get_frame(block);

		SRV_CORRUPT_TABLE_CHECK(page,
		{
			page_cursor->block = 0;
			page_cursor->rec = 0;

			if (estimate) {

				cursor->path_arr->nth_rec =
					ULINT_UNDEFINED;
			}
			/* Can't use break with the macro */
			goto exit_loop;
		});

		ut_ad(fil_page_get_type(page) == FIL_PAGE_INDEX);

		ut_ad(index->id == btr_page_get_index_id(page));

		block->check_index_page_at_flush = TRUE;

		if (height == ULINT_UNDEFINED) {
			/* We are in the root node */

			height = btr_page_get_level(page, mtr);
			root_height = height;
			ut_a(height >= level);
		} else {
			/* TODO: flag the index corrupted if this fails */
			ut_ad(height == btr_page_get_level(page, mtr));
		}

		if (height == level) {
			btr_cur_latch_leaves(
				page, space, zip_size, page_no,
				latch_mode & ~BTR_ALREADY_S_LATCHED,
				cursor, mtr);

			if (height == 0) {
				/* In versions <= 3.23.52 we had
				forgotten to release the tree latch
				here. If in an index scan we had to
				scan far to find a record visible to
				the current transaction, that could
				starve others waiting for the tree
				latch. */

				switch (latch_mode) {
				case BTR_MODIFY_TREE:
				case BTR_CONT_MODIFY_TREE:
				case BTR_SEARCH_LEAF | BTR_ALREADY_S_LATCHED:
				case BTR_MODIFY_LEAF | BTR_ALREADY_S_LATCHED:
					break;
				default:
					/* Release the tree s-latch */

					mtr_release_s_latch_at_savepoint(
						mtr, savepoint,
						dict_index_get_lock(index));
				}
			}
		}

		if (from_left) {
			page_cur_set_before_first(block, page_cursor);
		} else {
			page_cur_set_after_last(block, page_cursor);
		}

		if (height == level) {
			if (estimate) {
				btr_cur_add_path_info(cursor, height,
						      root_height);
			}

			break;
		}

		ut_ad(height > 0);

		if (from_left) {
			page_cur_move_to_next(page_cursor);
		} else {
			page_cur_move_to_prev(page_cursor);
		}

		if (estimate) {
			btr_cur_add_path_info(cursor, height, root_height);
		}

		height--;

		node_ptr = page_cur_get_rec(page_cursor);
		offsets = rec_get_offsets(node_ptr, cursor->index, offsets,
					  ULINT_UNDEFINED, &heap);
		/* Go to the child node */
		page_no = btr_node_ptr_get_child_page_no(node_ptr, offsets);
	}

exit_loop:
	if (UNIV_LIKELY_NULL(heap)) {
		mem_heap_free(heap);
	}

	return err;
}

/**********************************************************************//**
Positions a cursor at a randomly chosen position within a B-tree. */
UNIV_INTERN
void
btr_cur_open_at_rnd_pos_func(
/*=========================*/
	dict_index_t*	index,		/*!< in: index */
	ulint		latch_mode,	/*!< in: BTR_SEARCH_LEAF, ... */
	btr_cur_t*	cursor,		/*!< in/out: B-tree cursor */
	const char*	file,		/*!< in: file name */
	ulint		line,		/*!< in: line where called */
	mtr_t*		mtr)		/*!< in: mtr */
{
	page_cur_t*	page_cursor;
	ulint		page_no;
	ulint		space;
	ulint		zip_size;
	ulint		height;
	rec_t*		node_ptr;
	mem_heap_t*	heap		= NULL;
	ulint		offsets_[REC_OFFS_NORMAL_SIZE];
	ulint*		offsets		= offsets_;
	rec_offs_init(offsets_);

	switch (latch_mode) {
	case BTR_MODIFY_TREE:
		mtr_x_lock(dict_index_get_lock(index), mtr);
		break;
	default:
		ut_ad(latch_mode != BTR_CONT_MODIFY_TREE);
		mtr_s_lock(dict_index_get_lock(index), mtr);
	}

	page_cursor = btr_cur_get_page_cur(cursor);
	cursor->index = index;

	space = dict_index_get_space(index);
	zip_size = dict_table_zip_size(index->table);
	page_no = dict_index_get_page(index);

	height = ULINT_UNDEFINED;

	for (;;) {
		buf_block_t*	block;
		page_t*		page;
		dberr_t		err=DB_SUCCESS;

		block = buf_page_get_gen(space, zip_size, page_no,
					 RW_NO_LATCH, NULL, BUF_GET,
					 file, line, mtr, &err);

		if (err != DB_SUCCESS) {
			if (err == DB_DECRYPTION_FAILED) {
				ib_push_warning((void *)NULL,
					DB_DECRYPTION_FAILED,
					"Table %s is encrypted but encryption service or"
					" used key_id is not available. "
					" Can't continue reading table.",
					index->table->name);
				index->table->is_encrypted = true;
			}
			goto exit_loop;
		}

		page = buf_block_get_frame(block);

		SRV_CORRUPT_TABLE_CHECK(page,
		{
			page_cursor->block = 0;
			page_cursor->rec = 0;

			goto exit_loop;
		});

		ut_ad(fil_page_get_type(page) == FIL_PAGE_INDEX);

		ut_ad(index->id == btr_page_get_index_id(page));

		if (height == ULINT_UNDEFINED) {
			/* We are in the root node */

			height = btr_page_get_level(page, mtr);
		}

		if (height == 0) {
			btr_cur_latch_leaves(page, space, zip_size, page_no,
					     latch_mode, cursor, mtr);
		}

		page_cur_open_on_rnd_user_rec(block, page_cursor);

		if (height == 0) {

			break;
		}

		ut_ad(height > 0);

		height--;

		node_ptr = page_cur_get_rec(page_cursor);
		offsets = rec_get_offsets(node_ptr, cursor->index, offsets,
					  ULINT_UNDEFINED, &heap);
		/* Go to the child node */
		page_no = btr_node_ptr_get_child_page_no(node_ptr, offsets);
	}

exit_loop:
	if (UNIV_LIKELY_NULL(heap)) {
		mem_heap_free(heap);
	}
}

/*==================== B-TREE INSERT =========================*/

/*************************************************************//**
Inserts a record if there is enough space, or if enough space can
be freed by reorganizing. Differs from btr_cur_optimistic_insert because
no heuristics is applied to whether it pays to use CPU time for
reorganizing the page or not.

IMPORTANT: The caller will have to update IBUF_BITMAP_FREE
if this is a compressed leaf page in a secondary index.
This has to be done either within the same mini-transaction,
or by invoking ibuf_reset_free_bits() before mtr_commit().

@return	pointer to inserted record if succeed, else NULL */
static __attribute__((nonnull, warn_unused_result))
rec_t*
btr_cur_insert_if_possible(
/*=======================*/
	btr_cur_t*	cursor,	/*!< in: cursor on page after which to insert;
				cursor stays valid */
	const dtuple_t*	tuple,	/*!< in: tuple to insert; the size info need not
				have been stored to tuple */
	ulint**		offsets,/*!< out: offsets on *rec */
	mem_heap_t**	heap,	/*!< in/out: pointer to memory heap, or NULL */
	ulint		n_ext,	/*!< in: number of externally stored columns */
	mtr_t*		mtr)	/*!< in/out: mini-transaction */
{
	page_cur_t*	page_cursor;
	rec_t*		rec;

	ut_ad(dtuple_check_typed(tuple));

	ut_ad(mtr_memo_contains(mtr, btr_cur_get_block(cursor),
				MTR_MEMO_PAGE_X_FIX));
	page_cursor = btr_cur_get_page_cur(cursor);

	/* Now, try the insert */
	rec = page_cur_tuple_insert(page_cursor, tuple, cursor->index,
				    offsets, heap, n_ext, mtr);

	/* If the record did not fit, reorganize.
	For compressed pages, page_cur_tuple_insert()
	attempted this already. */
	if (!rec && !page_cur_get_page_zip(page_cursor)
	    && btr_page_reorganize(page_cursor, cursor->index, mtr)) {
		rec = page_cur_tuple_insert(
			page_cursor, tuple, cursor->index,
			offsets, heap, n_ext, mtr);
	}

	ut_ad(!rec || rec_offs_validate(rec, cursor->index, *offsets));
	return(rec);
}

/*************************************************************//**
For an insert, checks the locks and does the undo logging if desired.
@return	DB_SUCCESS, DB_WAIT_LOCK, DB_FAIL, or error number */
UNIV_INLINE __attribute__((warn_unused_result, nonnull(2,3,5,6)))
dberr_t
btr_cur_ins_lock_and_undo(
/*======================*/
	ulint		flags,	/*!< in: undo logging and locking flags: if
				not zero, the parameters index and thr
				should be specified */
	btr_cur_t*	cursor,	/*!< in: cursor on page after which to insert */
	dtuple_t*	entry,	/*!< in/out: entry to insert */
	que_thr_t*	thr,	/*!< in: query thread or NULL */
	mtr_t*		mtr,	/*!< in/out: mini-transaction */
	ibool*		inherit)/*!< out: TRUE if the inserted new record maybe
				should inherit LOCK_GAP type locks from the
				successor record */
{
	dict_index_t*	index;
	dberr_t		err;
	rec_t*		rec;
	roll_ptr_t	roll_ptr;

	if (UNIV_UNLIKELY(thr && thr_get_trx(thr)->fake_changes)) {
		/* skip LOCK, UNDO */
		return(DB_SUCCESS);
	}

	/* Check if we have to wait for a lock: enqueue an explicit lock
	request if yes */

	rec = btr_cur_get_rec(cursor);
	index = cursor->index;

	ut_ad(!dict_index_is_online_ddl(index)
	      || dict_index_is_clust(index)
	      || (flags & BTR_CREATE_FLAG));

	err = lock_rec_insert_check_and_lock(flags, rec,
					     btr_cur_get_block(cursor),
					     index, thr, mtr, inherit);

	if (err != DB_SUCCESS
	    || !dict_index_is_clust(index) || dict_index_is_ibuf(index)) {

		return(err);
	}

	err = trx_undo_report_row_operation(flags, TRX_UNDO_INSERT_OP,
					    thr, index, entry,
					    NULL, 0, NULL, NULL,
					    &roll_ptr);
	if (err != DB_SUCCESS) {

		return(err);
	}

	/* Now we can fill in the roll ptr field in entry */

	if (!(flags & BTR_KEEP_SYS_FLAG)) {

		row_upd_index_entry_sys_field(entry, index,
					      DATA_ROLL_PTR, roll_ptr);
	}

	return(DB_SUCCESS);
}

#ifdef UNIV_DEBUG
/*************************************************************//**
Report information about a transaction. */
static
void
btr_cur_trx_report(
/*===============*/
	trx_id_t		trx_id,	/*!< in: transaction id */
	const dict_index_t*	index,	/*!< in: index */
	const char*		op)	/*!< in: operation */
{
	fprintf(stderr, "Trx with id " TRX_ID_FMT " going to ", trx_id);
	fputs(op, stderr);
	dict_index_name_print(stderr, NULL, index);
	putc('\n', stderr);
}
#endif /* UNIV_DEBUG */

/*************************************************************//**
Tries to perform an insert to a page in an index tree, next to cursor.
It is assumed that mtr holds an x-latch on the page. The operation does
not succeed if there is too little space on the page. If there is just
one record on the page, the insert will always succeed; this is to
prevent trying to split a page with just one record.
@return	DB_SUCCESS, DB_WAIT_LOCK, DB_FAIL, or error number */
UNIV_INTERN
dberr_t
btr_cur_optimistic_insert(
/*======================*/
	ulint		flags,	/*!< in: undo logging and locking flags: if not
				zero, the parameters index and thr should be
				specified */
	btr_cur_t*	cursor,	/*!< in: cursor on page after which to insert;
				cursor stays valid */
	ulint**		offsets,/*!< out: offsets on *rec */
	mem_heap_t**	heap,	/*!< in/out: pointer to memory heap, or NULL */
	dtuple_t*	entry,	/*!< in/out: entry to insert */
	rec_t**		rec,	/*!< out: pointer to inserted record if
				succeed */
	big_rec_t**	big_rec,/*!< out: big rec vector whose fields have to
				be stored externally by the caller, or
				NULL */
	ulint		n_ext,	/*!< in: number of externally stored columns */
	que_thr_t*	thr,	/*!< in: query thread or NULL */
	mtr_t*		mtr)	/*!< in/out: mini-transaction;
				if this function returns DB_SUCCESS on
				a leaf page of a secondary index in a
				compressed tablespace, the caller must
				mtr_commit(mtr) before latching
				any further pages */
{
	big_rec_t*	big_rec_vec	= NULL;
	dict_index_t*	index;
	page_cur_t*	page_cursor;
	buf_block_t*	block;
	page_t*		page;
	rec_t*		dummy;
	ibool		leaf;
	ibool		reorg;
	ibool		inherit = TRUE;
	ulint		zip_size;
	ulint		rec_size;
	dberr_t		err;

	*big_rec = NULL;

	block = btr_cur_get_block(cursor);

	SRV_CORRUPT_TABLE_CHECK(block, return(DB_CORRUPTION););

	page = buf_block_get_frame(block);
	index = cursor->index;

	ut_ad((thr && thr_get_trx(thr)->fake_changes)
	      || mtr_memo_contains(mtr, block, MTR_MEMO_PAGE_X_FIX));
	ut_ad(!dict_index_is_online_ddl(index)
	      || dict_index_is_clust(index)
	      || (flags & BTR_CREATE_FLAG));
	ut_ad(dtuple_check_typed(entry));

	zip_size = buf_block_get_zip_size(block);
#ifdef UNIV_DEBUG_VALGRIND
	if (zip_size) {
		UNIV_MEM_ASSERT_RW(page, UNIV_PAGE_SIZE);
		UNIV_MEM_ASSERT_RW(block->page.zip.data, zip_size);
	}
#endif /* UNIV_DEBUG_VALGRIND */

#ifdef UNIV_DEBUG
	if (btr_cur_print_record_ops && thr) {
		btr_cur_trx_report(thr_get_trx(thr)->id, index, "insert ");
		dtuple_print(stderr, entry);
	}
#endif /* UNIV_DEBUG */

	ut_ad((thr && thr_get_trx(thr)->fake_changes)
	      || mtr_memo_contains(mtr, block, MTR_MEMO_PAGE_X_FIX));

	leaf = page_is_leaf(page);

	/* Calculate the record size when entry is converted to a record */
	rec_size = rec_get_converted_size(index, entry, n_ext);

	if (page_zip_rec_needs_ext(rec_size, page_is_comp(page),
				   dtuple_get_n_fields(entry), zip_size)) {

		/* The record is so big that we have to store some fields
		externally on separate database pages */
		big_rec_vec = dtuple_convert_big_rec(index, entry, &n_ext);

		if (UNIV_UNLIKELY(big_rec_vec == NULL)) {

			return(DB_TOO_BIG_RECORD);
		}

		rec_size = rec_get_converted_size(index, entry, n_ext);
	}

	if (zip_size) {
		/* Estimate the free space of an empty compressed page.
		Subtract one byte for the encoded heap_no in the
		modification log. */
		ulint	free_space_zip = page_zip_empty_size(
			cursor->index->n_fields, zip_size);
		ulint	n_uniq = dict_index_get_n_unique_in_tree(index);

		ut_ad(dict_table_is_comp(index->table));

		if (free_space_zip == 0) {
too_big:
			if (big_rec_vec) {
				dtuple_convert_back_big_rec(
					index, entry, big_rec_vec);
			}

			return(DB_TOO_BIG_RECORD);
		}

		/* Subtract one byte for the encoded heap_no in the
		modification log. */
		free_space_zip--;

		/* There should be enough room for two node pointer
		records on an empty non-leaf page.  This prevents
		infinite page splits. */

		if (entry->n_fields >= n_uniq
		    && (REC_NODE_PTR_SIZE
			+ rec_get_converted_size_comp_prefix(
				index, entry->fields, n_uniq, NULL)
			/* On a compressed page, there is
			a two-byte entry in the dense
			page directory for every record.
			But there is no record header. */
			- (REC_N_NEW_EXTRA_BYTES - 2)
			> free_space_zip / 2)) {
			goto too_big;
		}
	}

	LIMIT_OPTIMISTIC_INSERT_DEBUG(page_get_n_recs(page),
				      goto fail);

	if (leaf && zip_size
	    && (page_get_data_size(page) + rec_size
		>= dict_index_zip_pad_optimal_page_size(index))) {
		/* If compression padding tells us that insertion will
		result in too packed up page i.e.: which is likely to
		cause compression failure then don't do an optimistic
		insertion. */
fail:
		err = DB_FAIL;
fail_err:

		if (big_rec_vec) {
			dtuple_convert_back_big_rec(index, entry, big_rec_vec);
		}

		return(err);
	}

	ulint	max_size = page_get_max_insert_size_after_reorganize(page, 1);

	if (page_has_garbage(page)) {
		if ((max_size < rec_size
		     || max_size < BTR_CUR_PAGE_REORGANIZE_LIMIT)
		    && page_get_n_recs(page) > 1
		    && page_get_max_insert_size(page, 1) < rec_size) {

			goto fail;
		}
	} else if (max_size < rec_size) {
		goto fail;
	}

	/* If there have been many consecutive inserts to the
	clustered index leaf page of an uncompressed table, check if
	we have to split the page to reserve enough free space for
	future updates of records. */

	if (leaf && !zip_size && dict_index_is_clust(index)
	    && page_get_n_recs(page) >= 2
	    && dict_index_get_space_reserve() + rec_size > max_size
	    && (btr_page_get_split_rec_to_right(cursor, &dummy)
		|| btr_page_get_split_rec_to_left(cursor, &dummy))) {
		goto fail;
	}

	/* Check locks and write to the undo log, if specified */
	err = btr_cur_ins_lock_and_undo(flags, cursor, entry,
					thr, mtr, &inherit);

	if (UNIV_UNLIKELY(err != DB_SUCCESS)) {

		goto fail_err;
	}

	if (UNIV_UNLIKELY(thr && thr_get_trx(thr)->fake_changes)) {
		/* skip CHANGE, LOG */
		*big_rec = big_rec_vec;
		return(err); /* == DB_SUCCESS */
	}

	page_cursor = btr_cur_get_page_cur(cursor);

	/* Now, try the insert */

	{
		const rec_t* page_cursor_rec = page_cur_get_rec(page_cursor);
		*rec = page_cur_tuple_insert(page_cursor, entry, index,
					     offsets, heap, n_ext, mtr);
		reorg = page_cursor_rec != page_cur_get_rec(page_cursor);
	}

	if (*rec) {
	} else if (zip_size) {
		/* Reset the IBUF_BITMAP_FREE bits, because
		page_cur_tuple_insert() will have attempted page
		reorganize before failing. */
		if (leaf && !dict_index_is_clust(index)) {
			ibuf_reset_free_bits(block);
		}

		goto fail;
	} else {
		ut_ad(!reorg);

		/* If the record did not fit, reorganize */
		if (!btr_page_reorganize(page_cursor, index, mtr)) {
			ut_ad(0);
			goto fail;
		}

		ut_ad(page_get_max_insert_size(page, 1) == max_size);

		reorg = TRUE;

		*rec = page_cur_tuple_insert(page_cursor, entry, index,
					     offsets, heap, n_ext, mtr);

		if (UNIV_UNLIKELY(!*rec)) {
			fputs("InnoDB: Error: cannot insert tuple ", stderr);
			dtuple_print(stderr, entry);
			fputs(" into ", stderr);
			dict_index_name_print(stderr, thr_get_trx(thr), index);
			fprintf(stderr, "\nInnoDB: max insert size %lu\n",
				(ulong) max_size);
			ut_error;
		}
	}

#ifdef BTR_CUR_HASH_ADAPT
	if (!reorg && leaf && (cursor->flag == BTR_CUR_HASH)) {
		btr_search_update_hash_node_on_insert(cursor);
	} else {
		btr_search_update_hash_on_insert(cursor);
	}
#endif

	if (!(flags & BTR_NO_LOCKING_FLAG) && inherit) {

		lock_update_insert(block, *rec);
	}

	if (leaf && !dict_index_is_clust(index)) {
		/* Update the free bits of the B-tree page in the
		insert buffer bitmap. */

		/* The free bits in the insert buffer bitmap must
		never exceed the free space on a page.  It is safe to
		decrement or reset the bits in the bitmap in a
		mini-transaction that is committed before the
		mini-transaction that affects the free space. */

		/* It is unsafe to increment the bits in a separately
		committed mini-transaction, because in crash recovery,
		the free bits could momentarily be set too high. */

		if (zip_size) {
			/* Update the bits in the same mini-transaction. */
			ibuf_update_free_bits_zip(block, mtr);
		} else {
			/* Decrement the bits in a separate
			mini-transaction. */
			ibuf_update_free_bits_if_full(
				block, max_size,
				rec_size + PAGE_DIR_SLOT_SIZE);
		}
	}

	*big_rec = big_rec_vec;

	return(DB_SUCCESS);
}

/*************************************************************//**
Performs an insert on a page of an index tree. It is assumed that mtr
holds an x-latch on the tree and on the cursor page. If the insert is
made on the leaf level, to avoid deadlocks, mtr must also own x-latches
to brothers of page, if those brothers exist.
@return	DB_SUCCESS or error number */
UNIV_INTERN
dberr_t
btr_cur_pessimistic_insert(
/*=======================*/
	ulint		flags,	/*!< in: undo logging and locking flags: if not
				zero, the parameter thr should be
				specified; if no undo logging is specified,
				then the caller must have reserved enough
				free extents in the file space so that the
				insertion will certainly succeed */
	btr_cur_t*	cursor,	/*!< in: cursor after which to insert;
				cursor stays valid */
	ulint**		offsets,/*!< out: offsets on *rec */
	mem_heap_t**	heap,	/*!< in/out: pointer to memory heap
				that can be emptied, or NULL */
	dtuple_t*	entry,	/*!< in/out: entry to insert */
	rec_t**		rec,	/*!< out: pointer to inserted record if
				succeed */
	big_rec_t**	big_rec,/*!< out: big rec vector whose fields have to
				be stored externally by the caller, or
				NULL */
	ulint		n_ext,	/*!< in: number of externally stored columns */
	que_thr_t*	thr,	/*!< in: query thread or NULL */
	mtr_t*		mtr)	/*!< in/out: mini-transaction */
{
	dict_index_t*	index		= cursor->index;
	ulint		zip_size	= dict_table_zip_size(index->table);
	big_rec_t*	big_rec_vec	= NULL;
	dberr_t		err;
	ibool		inherit = FALSE;
	ibool		success;
	ulint		n_reserved	= 0;

	ut_ad(dtuple_check_typed(entry));

	*big_rec = NULL;

	ut_ad((thr && thr_get_trx(thr)->fake_changes) || mtr_memo_contains(mtr,
				dict_index_get_lock(btr_cur_get_index(cursor)),
				MTR_MEMO_X_LOCK));
	ut_ad((thr && thr_get_trx(thr)->fake_changes) || mtr_memo_contains(mtr, btr_cur_get_block(cursor),
				MTR_MEMO_PAGE_X_FIX));
	ut_ad(!dict_index_is_online_ddl(index)
	      || dict_index_is_clust(index)
	      || (flags & BTR_CREATE_FLAG));

	cursor->flag = BTR_CUR_BINARY;

	/* Check locks and write to undo log, if specified */

	err = btr_cur_ins_lock_and_undo(flags, cursor, entry,
					thr, mtr, &inherit);

	if (err != DB_SUCCESS) {

		return(err);
	}

	if (!(flags & BTR_NO_UNDO_LOG_FLAG)) {

		ut_a(cursor->tree_height != ULINT_UNDEFINED);

		/* First reserve enough free space for the file segments
		of the index tree, so that the insert will not fail because
		of lack of space */

		ulint	n_extents = cursor->tree_height / 16 + 3;

		success = fsp_reserve_free_extents(&n_reserved, index->space,
						   n_extents, FSP_NORMAL, mtr);
		if (!success) {
			return(DB_OUT_OF_FILE_SPACE);
		}
	}

	if (page_zip_rec_needs_ext(rec_get_converted_size(index, entry, n_ext),
				   dict_table_is_comp(index->table),
				   dtuple_get_n_fields(entry),
				   zip_size)) {
		/* The record is so big that we have to store some fields
		externally on separate database pages */

		if (UNIV_LIKELY_NULL(big_rec_vec)) {
			/* This should never happen, but we handle
			the situation in a robust manner. */
			ut_ad(0);
			dtuple_convert_back_big_rec(index, entry, big_rec_vec);
		}

		big_rec_vec = dtuple_convert_big_rec(index, entry, &n_ext);

		if (big_rec_vec == NULL) {

			if (n_reserved > 0) {
				fil_space_release_free_extents(index->space,
							       n_reserved);
			}
			return(DB_TOO_BIG_RECORD);
		}
	}

	if (UNIV_UNLIKELY(thr && thr_get_trx(thr)->fake_changes)) {
		/* skip CHANGE, LOG */
		if (n_reserved > 0) {
			fil_space_release_free_extents(index->space,
						       n_reserved);
		}
		*big_rec = big_rec_vec;
		return(DB_SUCCESS);
	}

	if (dict_index_get_page(index)
	    == buf_block_get_page_no(btr_cur_get_block(cursor))) {

		/* The page is the root page */
		*rec = btr_root_raise_and_insert(
			flags, cursor, offsets, heap, entry, n_ext, mtr);
	} else {
		*rec = btr_page_split_and_insert(
			flags, cursor, offsets, heap, entry, n_ext, mtr);
	}

	ut_ad(page_rec_get_next(btr_cur_get_rec(cursor)) == *rec);

	if (!(flags & BTR_NO_LOCKING_FLAG)) {
		/* The cursor might be moved to the other page,
		and the max trx id field should be updated after
		the cursor was fixed. */
		if (!dict_index_is_clust(index)) {
			page_update_max_trx_id(
				btr_cur_get_block(cursor),
				btr_cur_get_page_zip(cursor),
				thr_get_trx(thr)->id, mtr);
		}
		if (!page_rec_is_infimum(btr_cur_get_rec(cursor))
		    || btr_page_get_prev(
			buf_block_get_frame(
				btr_cur_get_block(cursor)), mtr)
		       == FIL_NULL) {
			/* split and inserted need to call
			lock_update_insert() always. */
			inherit = TRUE;
		}
	}

#ifdef BTR_CUR_ADAPT
	btr_search_update_hash_on_insert(cursor);
#endif
	if (inherit && !(flags & BTR_NO_LOCKING_FLAG)) {

		lock_update_insert(btr_cur_get_block(cursor), *rec);
	}

	if (n_reserved > 0) {
		fil_space_release_free_extents(index->space, n_reserved);
	}

	*big_rec = big_rec_vec;

	return(DB_SUCCESS);
}

/*==================== B-TREE UPDATE =========================*/

/*************************************************************//**
For an update, checks the locks and does the undo logging.
@return	DB_SUCCESS, DB_WAIT_LOCK, or error number */
UNIV_INLINE __attribute__((warn_unused_result, nonnull(2,3,6,7)))
dberr_t
btr_cur_upd_lock_and_undo(
/*======================*/
	ulint		flags,	/*!< in: undo logging and locking flags */
	btr_cur_t*	cursor,	/*!< in: cursor on record to update */
	const ulint*	offsets,/*!< in: rec_get_offsets() on cursor */
	const upd_t*	update,	/*!< in: update vector */
	ulint		cmpl_info,/*!< in: compiler info on secondary index
				updates */
	que_thr_t*	thr,	/*!< in: query thread
				(can be NULL if BTR_NO_LOCKING_FLAG) */
	mtr_t*		mtr,	/*!< in/out: mini-transaction */
	roll_ptr_t*	roll_ptr)/*!< out: roll pointer */
{
	dict_index_t*	index;
	const rec_t*	rec;
	dberr_t		err;

	ut_ad(thr || (flags & BTR_NO_LOCKING_FLAG));

	if (UNIV_UNLIKELY(thr && thr_get_trx(thr)->fake_changes)) {
		/* skip LOCK, UNDO */
		return(DB_SUCCESS);
	}

	rec = btr_cur_get_rec(cursor);
	index = cursor->index;

	ut_ad(rec_offs_validate(rec, index, offsets));

	if (!dict_index_is_clust(index)) {
		ut_ad(dict_index_is_online_ddl(index)
		      == !!(flags & BTR_CREATE_FLAG));

		/* We do undo logging only when we update a clustered index
		record */
		return(lock_sec_rec_modify_check_and_lock(
			       flags, btr_cur_get_block(cursor), rec,
			       index, thr, mtr));
	}

	/* Check if we have to wait for a lock: enqueue an explicit lock
	request if yes */

	if (!(flags & BTR_NO_LOCKING_FLAG)) {
		err = lock_clust_rec_modify_check_and_lock(
			flags, btr_cur_get_block(cursor), rec, index,
			offsets, thr);
		if (err != DB_SUCCESS) {
			return(err);
		}
	}

	/* Append the info about the update in the undo log */

	return(trx_undo_report_row_operation(
		       flags, TRX_UNDO_MODIFY_OP, thr,
		       index, NULL, update,
		       cmpl_info, rec, offsets, roll_ptr));
}

/***********************************************************//**
Writes a redo log record of updating a record in-place. */
UNIV_INTERN
void
btr_cur_update_in_place_log(
/*========================*/
	ulint		flags,		/*!< in: flags */
	const rec_t*	rec,		/*!< in: record */
	dict_index_t*	index,		/*!< in: index of the record */
	const upd_t*	update,		/*!< in: update vector */
	trx_id_t	trx_id,		/*!< in: transaction id */
	roll_ptr_t	roll_ptr,	/*!< in: roll ptr */
	mtr_t*		mtr)		/*!< in: mtr */
{
	byte*		log_ptr;
	const page_t*	page	= page_align(rec);
	ut_ad(flags < 256);
	ut_ad(!!page_is_comp(page) == dict_table_is_comp(index->table));

	log_ptr = mlog_open_and_write_index(mtr, rec, index, page_is_comp(page)
					    ? MLOG_COMP_REC_UPDATE_IN_PLACE
					    : MLOG_REC_UPDATE_IN_PLACE,
					    1 + DATA_ROLL_PTR_LEN + 14 + 2
					    + MLOG_BUF_MARGIN);

	if (!log_ptr) {
		/* Logging in mtr is switched off during crash recovery */
		return;
	}

	/* For secondary indexes, we could skip writing the dummy system fields
	to the redo log but we have to change redo log parsing of
	MLOG_REC_UPDATE_IN_PLACE/MLOG_COMP_REC_UPDATE_IN_PLACE or we have to add
	new redo log record. For now, just write dummy sys fields to the redo
	log if we are updating a secondary index record.
	*/
	mach_write_to_1(log_ptr, flags);
	log_ptr++;

	if (dict_index_is_clust(index)) {
		log_ptr = row_upd_write_sys_vals_to_log(
				index, trx_id, roll_ptr, log_ptr, mtr);
	} else {
		/* Dummy system fields for a secondary index */
		/* TRX_ID Position */
		log_ptr += mach_write_compressed(log_ptr, 0);
		/* ROLL_PTR */
		trx_write_roll_ptr(log_ptr, 0);
		log_ptr += DATA_ROLL_PTR_LEN;
		/* TRX_ID */
		log_ptr += mach_ull_write_compressed(log_ptr, 0);
	}

	mach_write_to_2(log_ptr, page_offset(rec));
	log_ptr += 2;

	row_upd_index_write_log(update, log_ptr, mtr);
}
#endif /* UNIV_HOTBACKUP */

/***********************************************************//**
Parses a redo log record of updating a record in-place.
@return	end of log record or NULL */
UNIV_INTERN
byte*
btr_cur_parse_update_in_place(
/*==========================*/
	byte*		ptr,	/*!< in: buffer */
	byte*		end_ptr,/*!< in: buffer end */
	page_t*		page,	/*!< in/out: page or NULL */
	page_zip_des_t*	page_zip,/*!< in/out: compressed page, or NULL */
	dict_index_t*	index)	/*!< in: index corresponding to page */
{
	ulint		flags;
	rec_t*		rec;
	upd_t*		update;
	ulint		pos;
	trx_id_t	trx_id;
	roll_ptr_t	roll_ptr;
	ulint		rec_offset;
	mem_heap_t*	heap;
	ulint*		offsets;

	if (end_ptr < ptr + 1) {

		return(NULL);
	}

	flags = mach_read_from_1(ptr);
	ptr++;

	ptr = row_upd_parse_sys_vals(ptr, end_ptr, &pos, &trx_id, &roll_ptr);

	if (ptr == NULL) {

		return(NULL);
	}

	if (end_ptr < ptr + 2) {

		return(NULL);
	}

	rec_offset = mach_read_from_2(ptr);
	ptr += 2;

	ut_a(rec_offset <= UNIV_PAGE_SIZE);

	heap = mem_heap_create(256);

	ptr = row_upd_index_parse(ptr, end_ptr, heap, &update);

	if (!ptr || !page) {

		goto func_exit;
	}

	ut_a((ibool)!!page_is_comp(page) == dict_table_is_comp(index->table));
	rec = page + rec_offset;

	/* We do not need to reserve btr_search_latch, as the page is only
	being recovered, and there cannot be a hash index to it. */

	offsets = rec_get_offsets(rec, index, NULL, ULINT_UNDEFINED, &heap);

	if (!(flags & BTR_KEEP_SYS_FLAG)) {
		row_upd_rec_sys_fields_in_recovery(rec, page_zip, offsets,
						   pos, trx_id, roll_ptr);
	}

	row_upd_rec_in_place(rec, index, offsets, update, page_zip);

func_exit:
	mem_heap_free(heap);

	return(ptr);
}

#ifndef UNIV_HOTBACKUP
/*************************************************************//**
See if there is enough place in the page modification log to log
an update-in-place.

@retval false if out of space; IBUF_BITMAP_FREE will be reset
outside mtr if the page was recompressed
@retval	true if enough place;

IMPORTANT: The caller will have to update IBUF_BITMAP_FREE if this is
a secondary index leaf page. This has to be done either within the
same mini-transaction, or by invoking ibuf_reset_free_bits() before
mtr_commit(mtr). */
UNIV_INTERN
bool
btr_cur_update_alloc_zip_func(
/*==========================*/
	page_zip_des_t*	page_zip,/*!< in/out: compressed page */
	page_cur_t*	cursor,	/*!< in/out: B-tree page cursor */
	dict_index_t*	index,	/*!< in: the index corresponding to cursor */
#ifdef UNIV_DEBUG
	ulint*		offsets,/*!< in/out: offsets of the cursor record */
#endif /* UNIV_DEBUG */
	ulint		length,	/*!< in: size needed */
	bool		create,	/*!< in: true=delete-and-insert,
				false=update-in-place */
	mtr_t*		mtr,	/*!< in/out: mini-transaction */
	trx_t*		trx)	/*!< in: NULL or transaction */
{
	const page_t*	page = page_cur_get_page(cursor);

	ut_ad(page_zip == page_cur_get_page_zip(cursor));
	ut_ad(page_zip);
	ut_ad(!dict_index_is_ibuf(index));
	ut_ad(rec_offs_validate(page_cur_get_rec(cursor), index, offsets));

	if (page_zip_available(page_zip, dict_index_is_clust(index),
			       length, create)) {
		return(true);
	}

	if (!page_zip->m_nonempty && !page_has_garbage(page)) {
		/* The page has been freshly compressed, so
		reorganizing it will not help. */
		return(false);
	}

	if (create && page_is_leaf(page)
	    && (length + page_get_data_size(page)
		>= dict_index_zip_pad_optimal_page_size(index))) {
		return(false);
	}

	if (UNIV_UNLIKELY(trx && trx->fake_changes)) {
		/* Don't call page_zip_compress_write_log_no_data as that has
		assert which would fail. Assume there won't be a compression
		failure. */

		return(true);
	}

	if (!btr_page_reorganize(cursor, index, mtr)) {
		goto out_of_space;
	}

	rec_offs_make_valid(page_cur_get_rec(cursor), index, offsets);

	/* After recompressing a page, we must make sure that the free
	bits in the insert buffer bitmap will not exceed the free
	space on the page.  Because this function will not attempt
	recompression unless page_zip_available() fails above, it is
	safe to reset the free bits if page_zip_available() fails
	again, below.  The free bits can safely be reset in a separate
	mini-transaction.  If page_zip_available() succeeds below, we
	can be sure that the btr_page_reorganize() above did not reduce
	the free space available on the page. */

	if (page_zip_available(page_zip, dict_index_is_clust(index),
			       length, create)) {
		return(true);
	}

out_of_space:
	ut_ad(rec_offs_validate(page_cur_get_rec(cursor), index, offsets));

	/* Out of space: reset the free bits. */
	if (!dict_index_is_clust(index) && page_is_leaf(page)) {
		ibuf_reset_free_bits(page_cur_get_block(cursor));
	}

	return(false);
}

/*************************************************************//**
Updates a record when the update causes no size changes in its fields.
We assume here that the ordering fields of the record do not change.
@return locking or undo log related error code, or
@retval DB_SUCCESS on success
@retval DB_ZIP_OVERFLOW if there is not enough space left
on the compressed page (IBUF_BITMAP_FREE was reset outside mtr) */
UNIV_INTERN
dberr_t
btr_cur_update_in_place(
/*====================*/
	ulint		flags,	/*!< in: undo logging and locking flags */
	btr_cur_t*	cursor,	/*!< in: cursor on the record to update;
				cursor stays valid and positioned on the
				same record */
	ulint*		offsets,/*!< in/out: offsets on cursor->page_cur.rec */
	const upd_t*	update,	/*!< in: update vector */
	ulint		cmpl_info,/*!< in: compiler info on secondary index
				updates */
	que_thr_t*	thr,	/*!< in: query thread */
	trx_id_t	trx_id,	/*!< in: transaction id */
	mtr_t*		mtr)	/*!< in/out: mini-transaction; if this
				is a secondary index, the caller must
				mtr_commit(mtr) before latching any
				further pages */
{
	dict_index_t*	index;
	buf_block_t*	block;
	page_zip_des_t*	page_zip;
	dberr_t		err;
	rec_t*		rec;
	roll_ptr_t	roll_ptr	= 0;
	ulint		was_delete_marked;
	ibool		is_hashed;
	trx_t*		trx;

	rec = btr_cur_get_rec(cursor);
	index = cursor->index;
	ut_ad(rec_offs_validate(rec, index, offsets));
	ut_ad(!!page_rec_is_comp(rec) == dict_table_is_comp(index->table));
	/* The insert buffer tree should never be updated in place. */
	ut_ad(!dict_index_is_ibuf(index));
	ut_ad(dict_index_is_online_ddl(index) == !!(flags & BTR_CREATE_FLAG)
	      || dict_index_is_clust(index));
	ut_ad(thr_get_trx(thr)->id == trx_id
	      || (flags & ~(BTR_KEEP_POS_FLAG | BTR_KEEP_IBUF_BITMAP))
	      == (BTR_NO_UNDO_LOG_FLAG | BTR_NO_LOCKING_FLAG
		  | BTR_CREATE_FLAG | BTR_KEEP_SYS_FLAG));
	ut_ad(fil_page_get_type(btr_cur_get_page(cursor)) == FIL_PAGE_INDEX);
	ut_ad(btr_page_get_index_id(btr_cur_get_page(cursor)) == index->id);

#ifdef UNIV_DEBUG
	if (btr_cur_print_record_ops) {
		btr_cur_trx_report(trx_id, index, "update ");
		rec_print_new(stderr, rec, offsets);
	}
#endif /* UNIV_DEBUG */

	block = btr_cur_get_block(cursor);
	page_zip = buf_block_get_page_zip(block);
	trx = thr_get_trx(thr);

	/* Check that enough space is available on the compressed page. */
	if (page_zip) {
		if (!btr_cur_update_alloc_zip(
			    page_zip, btr_cur_get_page_cur(cursor),
			    index, offsets, rec_offs_size(offsets),
			    false, mtr, trx)) {
			return(DB_ZIP_OVERFLOW);
		}

		rec = btr_cur_get_rec(cursor);
	}

	/* Do lock checking and undo logging */
	err = btr_cur_upd_lock_and_undo(flags, cursor, offsets,
					update, cmpl_info,
					thr, mtr, &roll_ptr);
	if (UNIV_UNLIKELY(err != DB_SUCCESS)) {
		/* We may need to update the IBUF_BITMAP_FREE
		bits after a reorganize that was done in
		btr_cur_update_alloc_zip(). */
		goto func_exit;
	}

	if (UNIV_UNLIKELY(trx->fake_changes)) {
		/* skip CHANGE, LOG */
		return(err); /* == DB_SUCCESS */
	}

	if (!(flags & BTR_KEEP_SYS_FLAG)) {
		row_upd_rec_sys_fields(rec, NULL, index, offsets,
				       thr_get_trx(thr), roll_ptr);
	}

	was_delete_marked = rec_get_deleted_flag(
		rec, page_is_comp(buf_block_get_frame(block)));

	is_hashed = (block->index != NULL);

	if (is_hashed) {
		/* TO DO: Can we skip this if none of the fields
		index->search_info->curr_n_fields
		are being updated? */

		/* The function row_upd_changes_ord_field_binary works only
		if the update vector was built for a clustered index, we must
		NOT call it if index is secondary */

		if (!dict_index_is_clust(index)
		    || row_upd_changes_ord_field_binary(index, update, thr,
							NULL, NULL)) {

			/* Remove possible hash index pointer to this record */
			btr_search_update_hash_on_delete(cursor);
		}

		rw_lock_x_lock(btr_search_get_latch(cursor->index));
	}

	row_upd_rec_in_place(rec, index, offsets, update, page_zip);

	if (is_hashed) {
		rw_lock_x_unlock(btr_search_get_latch(cursor->index));
	}

	btr_cur_update_in_place_log(flags, rec, index, update,
				    trx_id, roll_ptr, mtr);

	if (was_delete_marked
	    && !rec_get_deleted_flag(
		    rec, page_is_comp(buf_block_get_frame(block)))) {
		/* The new updated record owns its possible externally
		stored fields */

		btr_cur_unmark_extern_fields(page_zip,
					     rec, index, offsets, mtr);
	}

	ut_ad(err == DB_SUCCESS);

func_exit:
	if (page_zip
	    && !(flags & BTR_KEEP_IBUF_BITMAP)
	    && !dict_index_is_clust(index)
	    && page_is_leaf(buf_block_get_frame(block))) {
		/* Update the free bits in the insert buffer. */
		ibuf_update_free_bits_zip(block, mtr);
	}

	return(err);
}

/*************************************************************//**
Tries to update a record on a page in an index tree. It is assumed that mtr
holds an x-latch on the page. The operation does not succeed if there is too
little space on the page or if the update would result in too empty a page,
so that tree compression is recommended. We assume here that the ordering
fields of the record do not change.
@return error code, including
@retval DB_SUCCESS on success
@retval DB_OVERFLOW if the updated record does not fit
@retval DB_UNDERFLOW if the page would become too empty
@retval DB_ZIP_OVERFLOW if there is not enough space left
on the compressed page (IBUF_BITMAP_FREE was reset outside mtr) */
UNIV_INTERN
dberr_t
btr_cur_optimistic_update(
/*======================*/
	ulint		flags,	/*!< in: undo logging and locking flags */
	btr_cur_t*	cursor,	/*!< in: cursor on the record to update;
				cursor stays valid and positioned on the
				same record */
	ulint**		offsets,/*!< out: offsets on cursor->page_cur.rec */
	mem_heap_t**	heap,	/*!< in/out: pointer to NULL or memory heap */
	const upd_t*	update,	/*!< in: update vector; this must also
				contain trx id and roll ptr fields */
	ulint		cmpl_info,/*!< in: compiler info on secondary index
				updates */
	que_thr_t*	thr,	/*!< in: query thread */
	trx_id_t	trx_id,	/*!< in: transaction id */
	mtr_t*		mtr)	/*!< in/out: mini-transaction; if this
				is a secondary index, the caller must
				mtr_commit(mtr) before latching any
				further pages */
{
	dict_index_t*	index;
	page_cur_t*	page_cursor;
	dberr_t		err;
	buf_block_t*	block;
	page_t*		page;
	page_zip_des_t*	page_zip;
	rec_t*		rec;
	ulint		max_size;
	ulint		new_rec_size;
	ulint		old_rec_size;
	dtuple_t*	new_entry;
	roll_ptr_t	roll_ptr;
	ulint		i;
	ulint		n_ext;

	block = btr_cur_get_block(cursor);
	page = buf_block_get_frame(block);
	rec = btr_cur_get_rec(cursor);
	index = cursor->index;
	ut_ad(!!page_rec_is_comp(rec) == dict_table_is_comp(index->table));
	ut_ad(thr_get_trx(thr)->fake_changes
	      || mtr_memo_contains(mtr, block, MTR_MEMO_PAGE_X_FIX));
	/* The insert buffer tree should never be updated in place. */
	ut_ad(!dict_index_is_ibuf(index));
	ut_ad(dict_index_is_online_ddl(index) == !!(flags & BTR_CREATE_FLAG)
	      || dict_index_is_clust(index));
	ut_ad(thr_get_trx(thr)->id == trx_id
	      || (flags & ~(BTR_KEEP_POS_FLAG | BTR_KEEP_IBUF_BITMAP))
	      == (BTR_NO_UNDO_LOG_FLAG | BTR_NO_LOCKING_FLAG
		  | BTR_CREATE_FLAG | BTR_KEEP_SYS_FLAG));
	ut_ad(fil_page_get_type(page) == FIL_PAGE_INDEX);
	ut_ad(btr_page_get_index_id(page) == index->id);

	*offsets = rec_get_offsets(rec, index, *offsets,
				   ULINT_UNDEFINED, heap);
#if defined UNIV_DEBUG || defined UNIV_BLOB_LIGHT_DEBUG
	ut_a(!rec_offs_any_null_extern(rec, *offsets)
	     || trx_is_recv(thr_get_trx(thr)));
#endif /* UNIV_DEBUG || UNIV_BLOB_LIGHT_DEBUG */

#ifdef UNIV_DEBUG
	if (btr_cur_print_record_ops) {
		btr_cur_trx_report(trx_id, index, "update ");
		rec_print_new(stderr, rec, *offsets);
	}
#endif /* UNIV_DEBUG */

	if (!row_upd_changes_field_size_or_external(index, *offsets, update)) {

		/* The simplest and the most common case: the update does not
		change the size of any field and none of the updated fields is
		externally stored in rec or update, and there is enough space
		on the compressed page to log the update. */

		return(btr_cur_update_in_place(
			       flags, cursor, *offsets, update,
			       cmpl_info, thr, trx_id, mtr));
	}

	if (rec_offs_any_extern(*offsets)) {
any_extern:
		/* Externally stored fields are treated in pessimistic
		update */

		return(DB_OVERFLOW);
	}

	for (i = 0; i < upd_get_n_fields(update); i++) {
		if (dfield_is_ext(&upd_get_nth_field(update, i)->new_val)) {

			goto any_extern;
		}
	}

	page_cursor = btr_cur_get_page_cur(cursor);

	if (!*heap) {
		*heap = mem_heap_create(
			rec_offs_size(*offsets)
			+ DTUPLE_EST_ALLOC(rec_offs_n_fields(*offsets)));
	}

	new_entry = row_rec_to_index_entry(rec, index, *offsets,
					   &n_ext, *heap);
	/* We checked above that there are no externally stored fields. */
	ut_a(!n_ext);

	/* The page containing the clustered index record
	corresponding to new_entry is latched in mtr.
	Thus the following call is safe. */
	row_upd_index_replace_new_col_vals_index_pos(new_entry, index, update,
						     FALSE, *heap);
	old_rec_size = rec_offs_size(*offsets);
	new_rec_size = rec_get_converted_size(index, new_entry, 0);

	page_zip = buf_block_get_page_zip(block);
#ifdef UNIV_ZIP_DEBUG
	ut_a(!page_zip || page_zip_validate(page_zip, page, index));
#endif /* UNIV_ZIP_DEBUG */

	if (page_zip) {
		if (!btr_cur_update_alloc_zip(
			    page_zip, page_cursor, index, *offsets,
			    new_rec_size, true, mtr, thr_get_trx(thr))) {
			return(DB_ZIP_OVERFLOW);
		}

		rec = page_cur_get_rec(page_cursor);
	}

	if (UNIV_UNLIKELY(new_rec_size
			  >= (page_get_free_space_of_empty(page_is_comp(page))
			      / 2))) {
		/* We may need to update the IBUF_BITMAP_FREE
		bits after a reorganize that was done in
		btr_cur_update_alloc_zip(). */
		err = DB_OVERFLOW;
		goto func_exit;
	}

	if (UNIV_UNLIKELY(page_get_data_size(page)
			  - old_rec_size + new_rec_size
			  < BTR_CUR_PAGE_COMPRESS_LIMIT)) {
		/* We may need to update the IBUF_BITMAP_FREE
		bits after a reorganize that was done in
		btr_cur_update_alloc_zip(). */

		/* The page would become too empty */
		err = DB_UNDERFLOW;
		goto func_exit;
	}

	/* We do not attempt to reorganize if the page is compressed.
	This is because the page may fail to compress after reorganization. */
	max_size = page_zip
		? page_get_max_insert_size(page, 1)
		: (old_rec_size
		   + page_get_max_insert_size_after_reorganize(page, 1));

	if (!(((max_size >= BTR_CUR_PAGE_REORGANIZE_LIMIT)
	       && (max_size >= new_rec_size))
	      || (page_get_n_recs(page) <= 1))) {

		/* We may need to update the IBUF_BITMAP_FREE
		bits after a reorganize that was done in
		btr_cur_update_alloc_zip(). */

		/* There was not enough space, or it did not pay to
		reorganize: for simplicity, we decide what to do assuming a
		reorganization is needed, though it might not be necessary */

		err = DB_OVERFLOW;
		goto func_exit;
	}

	/* Do lock checking and undo logging */
	err = btr_cur_upd_lock_and_undo(flags, cursor, *offsets,
					update, cmpl_info,
					thr, mtr, &roll_ptr);
	if (err != DB_SUCCESS) {
		/* We may need to update the IBUF_BITMAP_FREE
		bits after a reorganize that was done in
		btr_cur_update_alloc_zip(). */
		goto func_exit;
	}

	if (UNIV_UNLIKELY(thr_get_trx(thr)->fake_changes)) {
		/* skip CHANGE, LOG */
		ut_ad(err == DB_SUCCESS);
		return(DB_SUCCESS);
	}

	/* Ok, we may do the replacement. Store on the page infimum the
	explicit locks on rec, before deleting rec (see the comment in
	btr_cur_pessimistic_update). */

	lock_rec_store_on_page_infimum(block, rec);

	btr_search_update_hash_on_delete(cursor);

	page_cur_delete_rec(page_cursor, index, *offsets, mtr);

	page_cur_move_to_prev(page_cursor);

	if (!(flags & BTR_KEEP_SYS_FLAG)) {
		row_upd_index_entry_sys_field(new_entry, index, DATA_ROLL_PTR,
					      roll_ptr);
		row_upd_index_entry_sys_field(new_entry, index, DATA_TRX_ID,
					      trx_id);
	}

	/* There are no externally stored columns in new_entry */
	rec = btr_cur_insert_if_possible(
		cursor, new_entry, offsets, heap, 0/*n_ext*/, mtr);
	ut_a(rec); /* <- We calculated above the insert would fit */

	/* Restore the old explicit lock state on the record */

	lock_rec_restore_from_page_infimum(block, rec, block);

	page_cur_move_to_next(page_cursor);
	ut_ad(err == DB_SUCCESS);

func_exit:
	if (page_zip
	    && !(flags & BTR_KEEP_IBUF_BITMAP)
	    && !dict_index_is_clust(index)
	    && page_is_leaf(page)) {
		/* Update the free bits in the insert buffer. */
		ibuf_update_free_bits_zip(block, mtr);
	}

	return(err);
}

/*************************************************************//**
If, in a split, a new supremum record was created as the predecessor of the
updated record, the supremum record must inherit exactly the locks on the
updated record. In the split it may have inherited locks from the successor
of the updated record, which is not correct. This function restores the
right locks for the new supremum. */
static
void
btr_cur_pess_upd_restore_supremum(
/*==============================*/
	buf_block_t*	block,	/*!< in: buffer block of rec */
	const rec_t*	rec,	/*!< in: updated record */
	mtr_t*		mtr)	/*!< in: mtr */
{
	page_t*		page;
	buf_block_t*	prev_block;
	ulint		space;
	ulint		zip_size;
	ulint		prev_page_no;

	page = buf_block_get_frame(block);

	if (page_rec_get_next(page_get_infimum_rec(page)) != rec) {
		/* Updated record is not the first user record on its page */

		return;
	}

	space = buf_block_get_space(block);
	zip_size = buf_block_get_zip_size(block);
	prev_page_no = btr_page_get_prev(page, mtr);

	ut_ad(prev_page_no != FIL_NULL);
	prev_block = buf_page_get_with_no_latch(space, zip_size,
						prev_page_no, mtr);
#ifdef UNIV_BTR_DEBUG
	ut_a(btr_page_get_next(prev_block->frame, mtr)
	     == page_get_page_no(page));
#endif /* UNIV_BTR_DEBUG */

	/* We must already have an x-latch on prev_block! */
	ut_ad(mtr_memo_contains(mtr, prev_block, MTR_MEMO_PAGE_X_FIX));

	lock_rec_reset_and_inherit_gap_locks(prev_block, block,
					     PAGE_HEAP_NO_SUPREMUM,
					     page_rec_get_heap_no(rec));
}

/*************************************************************//**
Check if the total length of the modified blob for the row is within 10%
of the total redo log size.  This constraint on the blob length is to
avoid overwriting the redo logs beyond the last checkpoint lsn.
@return	DB_SUCCESS or DB_TOO_BIG_FOR_REDO. */
static
dberr_t
btr_check_blob_limit(const big_rec_t*	big_rec_vec)
{
	const	ib_uint64_t redo_size = srv_n_log_files * srv_log_file_size
		* UNIV_PAGE_SIZE;
	const	ib_uint64_t redo_10p = redo_size / 10;
	ib_uint64_t	total_blob_len = 0;
	dberr_t	err = DB_SUCCESS;

	/* Calculate the total number of bytes for blob data */
	for (ulint i = 0; i < big_rec_vec->n_fields; i++) {
		total_blob_len += big_rec_vec->fields[i].len;
	}

	if (total_blob_len > redo_10p) {
		ib_logf(IB_LOG_LEVEL_ERROR, "The total blob data"
			" length (" UINT64PF ") is greater than"
			" 10%% of the total redo log size (" UINT64PF
			"). Please increase total redo log size.",
			total_blob_len, redo_size);
		err = DB_TOO_BIG_FOR_REDO;
	}

	return(err);
}

/*************************************************************//**
Performs an update of a record on a page of a tree. It is assumed
that mtr holds an x-latch on the tree and on the cursor page. If the
update is made on the leaf level, to avoid deadlocks, mtr must also
own x-latches to brothers of page, if those brothers exist. We assume
here that the ordering fields of the record do not change.
@return	DB_SUCCESS or error code */
UNIV_INTERN
dberr_t
btr_cur_pessimistic_update(
/*=======================*/
	ulint		flags,	/*!< in: undo logging, locking, and rollback
				flags */
	btr_cur_t*	cursor,	/*!< in/out: cursor on the record to update;
				cursor may become invalid if *big_rec == NULL
				|| !(flags & BTR_KEEP_POS_FLAG) */
	ulint**		offsets,/*!< out: offsets on cursor->page_cur.rec */
	mem_heap_t**	offsets_heap,
				/*!< in/out: pointer to memory heap
				that can be emptied, or NULL */
	mem_heap_t*	entry_heap,
				/*!< in/out: memory heap for allocating
				big_rec and the index tuple */
	big_rec_t**	big_rec,/*!< out: big rec vector whose fields have to
				be stored externally by the caller, or NULL */
	const upd_t*	update,	/*!< in: update vector; this is allowed also
				contain trx id and roll ptr fields, but
				the values in update vector have no effect */
	ulint		cmpl_info,/*!< in: compiler info on secondary index
				updates */
	que_thr_t*	thr,	/*!< in: query thread */
	trx_id_t	trx_id,	/*!< in: transaction id */
	mtr_t*		mtr)	/*!< in/out: mini-transaction; must be
				committed before latching any further pages */
{
	big_rec_t*	big_rec_vec	= NULL;
	big_rec_t*	dummy_big_rec;
	dict_index_t*	index;
	buf_block_t*	block;
	page_t*		page;
	page_zip_des_t*	page_zip;
	rec_t*		rec;
	page_cur_t*	page_cursor;
	dberr_t		err;
	dberr_t		optim_err;
	roll_ptr_t	roll_ptr;
	ibool		was_first;
	ulint		n_reserved	= 0;
	ulint		n_ext;
	trx_t*		trx;

	*offsets = NULL;
	*big_rec = NULL;

	block = btr_cur_get_block(cursor);
	page = buf_block_get_frame(block);
	page_zip = buf_block_get_page_zip(block);
	index = cursor->index;

	ut_ad(thr_get_trx(thr)->fake_changes
	      || mtr_memo_contains(mtr, dict_index_get_lock(index),
				   MTR_MEMO_X_LOCK));
	ut_ad(thr_get_trx(thr)->fake_changes
	      || mtr_memo_contains(mtr, block, MTR_MEMO_PAGE_X_FIX));
#ifdef UNIV_ZIP_DEBUG
	ut_a(!page_zip || page_zip_validate(page_zip, page, index));
#endif /* UNIV_ZIP_DEBUG */
	/* The insert buffer tree should never be updated in place. */
	ut_ad(!dict_index_is_ibuf(index));
	ut_ad(dict_index_is_online_ddl(index) == !!(flags & BTR_CREATE_FLAG)
	      || dict_index_is_clust(index));
	ut_ad(thr_get_trx(thr)->id == trx_id
	      || (flags & ~BTR_KEEP_POS_FLAG)
	      == (BTR_NO_UNDO_LOG_FLAG | BTR_NO_LOCKING_FLAG
		  | BTR_CREATE_FLAG | BTR_KEEP_SYS_FLAG));

	err = optim_err = btr_cur_optimistic_update(
		flags | BTR_KEEP_IBUF_BITMAP,
		cursor, offsets, offsets_heap, update,
		cmpl_info, thr, trx_id, mtr);

	switch (err) {
	case DB_ZIP_OVERFLOW:
	case DB_UNDERFLOW:
	case DB_OVERFLOW:
		break;
	default:
	err_exit:
		/* We suppressed this with BTR_KEEP_IBUF_BITMAP.
		For DB_ZIP_OVERFLOW, the IBUF_BITMAP_FREE bits were
		already reset by btr_cur_update_alloc_zip() if the
		page was recompressed. */
		if (page_zip
		    && optim_err != DB_ZIP_OVERFLOW
		    && !dict_index_is_clust(index)
		    && page_is_leaf(page)) {
			ibuf_update_free_bits_zip(block, mtr);
		}

		return(err);
	}

	/* Do lock checking and undo logging */
	err = btr_cur_upd_lock_and_undo(flags, cursor, *offsets,
					update, cmpl_info,
					thr, mtr, &roll_ptr);
	if (err != DB_SUCCESS) {
		goto err_exit;
	}

	if (optim_err == DB_OVERFLOW) {
		ulint	reserve_flag;
		ulint	n_extents;

		/* First reserve enough free space for the file segments
		of the index tree, so that the update will not fail because
		of lack of space */
		if (UNIV_UNLIKELY(cursor->tree_height == ULINT_UNDEFINED)) {
			/* When the tree height is uninitialized due to fake
			changes, reserve some hardcoded number of extents.  */
			ut_a(thr_get_trx(thr)->fake_changes);
			n_extents = 3;
		}
		else {
			n_extents = cursor->tree_height / 16 + 3;
		}

		if (flags & BTR_NO_UNDO_LOG_FLAG) {
			reserve_flag = FSP_CLEANING;
		} else {
			reserve_flag = FSP_NORMAL;
		}

		if (!fsp_reserve_free_extents(&n_reserved, index->space,
					      n_extents, reserve_flag, mtr)) {
			err = DB_OUT_OF_FILE_SPACE;
			goto err_exit;
		}
	}

	rec = btr_cur_get_rec(cursor);

	*offsets = rec_get_offsets(
		rec, index, *offsets, ULINT_UNDEFINED, offsets_heap);

	dtuple_t*	new_entry = row_rec_to_index_entry(
		rec, index, *offsets, &n_ext, entry_heap);

	/* The page containing the clustered index record
	corresponding to new_entry is latched in mtr.  If the
	clustered index record is delete-marked, then its externally
	stored fields cannot have been purged yet, because then the
	purge would also have removed the clustered index record
	itself.  Thus the following call is safe. */
	row_upd_index_replace_new_col_vals_index_pos(new_entry, index, update,
						     FALSE, entry_heap);

	trx = thr_get_trx(thr);

	if (!(flags & BTR_KEEP_SYS_FLAG) && UNIV_LIKELY(!trx->fake_changes)) {
		row_upd_index_entry_sys_field(new_entry, index, DATA_ROLL_PTR,
					      roll_ptr);
		row_upd_index_entry_sys_field(new_entry, index, DATA_TRX_ID,
					      trx_id);
	}

	if ((flags & BTR_NO_UNDO_LOG_FLAG) && rec_offs_any_extern(*offsets)) {
		/* We are in a transaction rollback undoing a row
		update: we must free possible externally stored fields
		which got new values in the update, if they are not
		inherited values. They can be inherited if we have
		updated the primary key to another value, and then
		update it back again. */

		ut_ad(big_rec_vec == NULL);

		/* fake_changes should not cause undo. so never reaches here */
		ut_ad(!(trx->fake_changes));

		btr_rec_free_updated_extern_fields(
			index, rec, page_zip, *offsets, update,
			trx_is_recv(thr_get_trx(thr))
			? RB_RECOVERY : RB_NORMAL, mtr);
	}

	/* We have to set appropriate extern storage bits in the new
	record to be inserted: we have to remember which fields were such */

	ut_ad(!page_is_comp(page) || !rec_get_node_ptr_flag(rec));
	ut_ad(rec_offs_validate(rec, index, *offsets));
	n_ext += btr_push_update_extern_fields(new_entry, update, entry_heap);

	if (page_zip) {
		ut_ad(page_is_comp(page));
		if (page_zip_rec_needs_ext(
			    rec_get_converted_size(index, new_entry, n_ext),
			    TRUE,
			    dict_index_get_n_fields(index),
			    page_zip_get_size(page_zip))) {

			goto make_external;
		}
	} else if (page_zip_rec_needs_ext(
			   rec_get_converted_size(index, new_entry, n_ext),
			   page_is_comp(page), 0, 0)) {
make_external:
		big_rec_vec = dtuple_convert_big_rec(index, new_entry, &n_ext);
		if (UNIV_UNLIKELY(big_rec_vec == NULL)) {

			/* We cannot goto return_after_reservations,
			because we may need to update the
			IBUF_BITMAP_FREE bits, which was suppressed by
			BTR_KEEP_IBUF_BITMAP. */
#ifdef UNIV_ZIP_DEBUG
			ut_a(!page_zip
			     || page_zip_validate(page_zip, page, index));
#endif /* UNIV_ZIP_DEBUG */
			if (n_reserved > 0) {
				fil_space_release_free_extents(
					index->space, n_reserved);
			}

			err = DB_TOO_BIG_RECORD;
			goto err_exit;
		}

		ut_ad(page_is_leaf(page));
		ut_ad(dict_index_is_clust(index));
		ut_ad(flags & BTR_KEEP_POS_FLAG);
	}

	if (UNIV_UNLIKELY(trx->fake_changes)) {
		/* skip CHANGE, LOG */
		err = DB_SUCCESS;
		goto return_after_reservations;
	}

	if (big_rec_vec) {

		err = btr_check_blob_limit(big_rec_vec);

		if (err != DB_SUCCESS) {
			if (n_reserved > 0) {
				fil_space_release_free_extents(
					index->space, n_reserved);
			}
			goto err_exit;
		}
	}

	/* Store state of explicit locks on rec on the page infimum record,
	before deleting rec. The page infimum acts as a dummy carrier of the
	locks, taking care also of lock releases, before we can move the locks
	back on the actual record. There is a special case: if we are
	inserting on the root page and the insert causes a call of
	btr_root_raise_and_insert. Therefore we cannot in the lock system
	delete the lock structs set on the root page even if the root
	page carries just node pointers. */

	lock_rec_store_on_page_infimum(block, rec);

	btr_search_update_hash_on_delete(cursor);

#ifdef UNIV_ZIP_DEBUG
	ut_a(!page_zip || page_zip_validate(page_zip, page, index));
#endif /* UNIV_ZIP_DEBUG */
	page_cursor = btr_cur_get_page_cur(cursor);

	page_cur_delete_rec(page_cursor, index, *offsets, mtr);

	page_cur_move_to_prev(page_cursor);

	rec = btr_cur_insert_if_possible(cursor, new_entry,
					 offsets, offsets_heap, n_ext, mtr);

	if (rec) {
		page_cursor->rec = rec;

		lock_rec_restore_from_page_infimum(btr_cur_get_block(cursor),
						   rec, block);

		if (!rec_get_deleted_flag(rec, rec_offs_comp(*offsets))) {
			/* The new inserted record owns its possible externally
			stored fields */
			btr_cur_unmark_extern_fields(
				page_zip, rec, index, *offsets, mtr);
		}

		bool adjust = big_rec_vec && (flags & BTR_KEEP_POS_FLAG);

		if (btr_cur_compress_if_useful(cursor, adjust, mtr)) {
			if (adjust) {
				rec_offs_make_valid(
					page_cursor->rec, index, *offsets);
			}
		} else if (page_zip &&
			   !dict_index_is_clust(index)
			   && page_is_leaf(page)) {
			/* Update the free bits in the insert buffer.
			This is the same block which was skipped by
			BTR_KEEP_IBUF_BITMAP. */
			ibuf_update_free_bits_zip(block, mtr);
		}

		err = DB_SUCCESS;
		goto return_after_reservations;
	} else {
		/* If the page is compressed and it initially
		compresses very well, and there is a subsequent insert
		of a badly-compressing record, it is possible for
		btr_cur_optimistic_update() to return DB_UNDERFLOW and
		btr_cur_insert_if_possible() to return FALSE. */
		ut_a(page_zip || optim_err != DB_UNDERFLOW);

		/* Out of space: reset the free bits.
		This is the same block which was skipped by
		BTR_KEEP_IBUF_BITMAP. */
		if (!dict_index_is_clust(index) && page_is_leaf(page)) {
			ibuf_reset_free_bits(block);
		}
	}

	if (big_rec_vec) {
		ut_ad(page_is_leaf(page));
		ut_ad(dict_index_is_clust(index));
		ut_ad(flags & BTR_KEEP_POS_FLAG);

		/* btr_page_split_and_insert() in
		btr_cur_pessimistic_insert() invokes
		mtr_memo_release(mtr, index->lock, MTR_MEMO_X_LOCK).
		We must keep the index->lock when we created a
		big_rec, so that row_upd_clust_rec() can store the
		big_rec in the same mini-transaction. */

		mtr_x_lock(dict_index_get_lock(index), mtr);
	}

	/* Was the record to be updated positioned as the first user
	record on its page? */
	was_first = page_cur_is_before_first(page_cursor);

	/* Lock checks and undo logging were already performed by
	btr_cur_upd_lock_and_undo(). We do not try
	btr_cur_optimistic_insert() because
	btr_cur_insert_if_possible() already failed above. */

	err = btr_cur_pessimistic_insert(BTR_NO_UNDO_LOG_FLAG
					 | BTR_NO_LOCKING_FLAG
					 | BTR_KEEP_SYS_FLAG,
					 cursor, offsets, offsets_heap,
					 new_entry, &rec,
					 &dummy_big_rec, n_ext, NULL, mtr);
	ut_a(rec);
	ut_a(err == DB_SUCCESS);
	ut_a(dummy_big_rec == NULL);
	ut_ad(rec_offs_validate(rec, cursor->index, *offsets));
	page_cursor->rec = rec;

	if (dict_index_is_sec_or_ibuf(index)) {
		/* Update PAGE_MAX_TRX_ID in the index page header.
		It was not updated by btr_cur_pessimistic_insert()
		because of BTR_NO_LOCKING_FLAG. */
		buf_block_t*	rec_block;

		rec_block = btr_cur_get_block(cursor);

		page_update_max_trx_id(rec_block,
				       buf_block_get_page_zip(rec_block),
				       trx_id, mtr);
	}

	if (!rec_get_deleted_flag(rec, rec_offs_comp(*offsets))) {
		/* The new inserted record owns its possible externally
		stored fields */
		buf_block_t*	rec_block = btr_cur_get_block(cursor);

#ifdef UNIV_ZIP_DEBUG
		ut_a(!page_zip || page_zip_validate(page_zip, page, index));
		page = buf_block_get_frame(rec_block);
#endif /* UNIV_ZIP_DEBUG */
		page_zip = buf_block_get_page_zip(rec_block);

		btr_cur_unmark_extern_fields(page_zip,
					     rec, index, *offsets, mtr);
	}

	lock_rec_restore_from_page_infimum(btr_cur_get_block(cursor),
					   rec, block);

	/* If necessary, restore also the correct lock state for a new,
	preceding supremum record created in a page split. While the old
	record was nonexistent, the supremum might have inherited its locks
	from a wrong record. */

	if (!was_first) {
		btr_cur_pess_upd_restore_supremum(btr_cur_get_block(cursor),
						  rec, mtr);
	}

return_after_reservations:
#ifdef UNIV_ZIP_DEBUG
	ut_a(!page_zip || page_zip_validate(page_zip, page, index));
#endif /* UNIV_ZIP_DEBUG */

	if (n_reserved > 0) {
		fil_space_release_free_extents(index->space, n_reserved);
	}

	*big_rec = big_rec_vec;

	return(err);
}

/*==================== B-TREE DELETE MARK AND UNMARK ===============*/

/****************************************************************//**
Writes the redo log record for delete marking or unmarking of an index
record. */
UNIV_INLINE
void
btr_cur_del_mark_set_clust_rec_log(
/*===============================*/
	rec_t*		rec,	/*!< in: record */
	dict_index_t*	index,	/*!< in: index of the record */
	trx_id_t	trx_id,	/*!< in: transaction id */
	roll_ptr_t	roll_ptr,/*!< in: roll ptr to the undo log record */
	mtr_t*		mtr)	/*!< in: mtr */
{
	byte*	log_ptr;

	ut_ad(!!page_rec_is_comp(rec) == dict_table_is_comp(index->table));

	log_ptr = mlog_open_and_write_index(mtr, rec, index,
					    page_rec_is_comp(rec)
					    ? MLOG_COMP_REC_CLUST_DELETE_MARK
					    : MLOG_REC_CLUST_DELETE_MARK,
					    1 + 1 + DATA_ROLL_PTR_LEN
					    + 14 + 2);

	if (!log_ptr) {
		/* Logging in mtr is switched off during crash recovery */
		return;
	}

	*log_ptr++ = 0;
	*log_ptr++ = 1;

	log_ptr = row_upd_write_sys_vals_to_log(
		index, trx_id, roll_ptr, log_ptr, mtr);
	mach_write_to_2(log_ptr, page_offset(rec));
	log_ptr += 2;

	mlog_close(mtr, log_ptr);
}
#endif /* !UNIV_HOTBACKUP */

/****************************************************************//**
Parses the redo log record for delete marking or unmarking of a clustered
index record.
@return	end of log record or NULL */
UNIV_INTERN
byte*
btr_cur_parse_del_mark_set_clust_rec(
/*=================================*/
	byte*		ptr,	/*!< in: buffer */
	byte*		end_ptr,/*!< in: buffer end */
	page_t*		page,	/*!< in/out: page or NULL */
	page_zip_des_t*	page_zip,/*!< in/out: compressed page, or NULL */
	dict_index_t*	index)	/*!< in: index corresponding to page */
{
	ulint		flags;
	ulint		val;
	ulint		pos;
	trx_id_t	trx_id;
	roll_ptr_t	roll_ptr;
	ulint		offset;
	rec_t*		rec;

	ut_ad(!page
	      || !!page_is_comp(page) == dict_table_is_comp(index->table));

	if (end_ptr < ptr + 2) {

		return(NULL);
	}

	flags = mach_read_from_1(ptr);
	ptr++;
	val = mach_read_from_1(ptr);
	ptr++;

	ptr = row_upd_parse_sys_vals(ptr, end_ptr, &pos, &trx_id, &roll_ptr);

	if (ptr == NULL) {

		return(NULL);
	}

	if (end_ptr < ptr + 2) {

		return(NULL);
	}

	offset = mach_read_from_2(ptr);
	ptr += 2;

	ut_a(offset <= UNIV_PAGE_SIZE);

	if (page) {
		rec = page + offset;

		/* We do not need to reserve btr_search_latch, as the page
		is only being recovered, and there cannot be a hash index to
		it. Besides, these fields are being updated in place
		and the adaptive hash index does not depend on them. */

		btr_rec_set_deleted_flag(rec, page_zip, val);

		if (!(flags & BTR_KEEP_SYS_FLAG)) {
			mem_heap_t*	heap		= NULL;
			ulint		offsets_[REC_OFFS_NORMAL_SIZE];
			rec_offs_init(offsets_);

			row_upd_rec_sys_fields_in_recovery(
				rec, page_zip,
				rec_get_offsets(rec, index, offsets_,
						ULINT_UNDEFINED, &heap),
				pos, trx_id, roll_ptr);
			if (UNIV_LIKELY_NULL(heap)) {
				mem_heap_free(heap);
			}
		}
	}

	return(ptr);
}

#ifndef UNIV_HOTBACKUP
/***********************************************************//**
Marks a clustered index record deleted. Writes an undo log record to
undo log on this delete marking. Writes in the trx id field the id
of the deleting transaction, and in the roll ptr field pointer to the
undo log record created.
@return	DB_SUCCESS, DB_LOCK_WAIT, or error number */
UNIV_INTERN
dberr_t
btr_cur_del_mark_set_clust_rec(
/*===========================*/
	buf_block_t*	block,	/*!< in/out: buffer block of the record */
	rec_t*		rec,	/*!< in/out: record */
	dict_index_t*	index,	/*!< in: clustered index of the record */
	const ulint*	offsets,/*!< in: rec_get_offsets(rec) */
	que_thr_t*	thr,	/*!< in: query thread */
	mtr_t*		mtr)	/*!< in/out: mini-transaction */
{
	roll_ptr_t	roll_ptr;
	dberr_t		err;
	page_zip_des_t*	page_zip;
	trx_t*		trx;

	ut_ad(dict_index_is_clust(index));
	ut_ad(rec_offs_validate(rec, index, offsets));
	ut_ad(!!page_rec_is_comp(rec) == dict_table_is_comp(index->table));
	ut_ad(buf_block_get_frame(block) == page_align(rec));
	ut_ad(page_is_leaf(page_align(rec)));

#ifdef UNIV_DEBUG
	if (btr_cur_print_record_ops && thr) {
		btr_cur_trx_report(thr_get_trx(thr)->id, index, "del mark ");
		rec_print_new(stderr, rec, offsets);
	}
#endif /* UNIV_DEBUG */

	ut_ad(dict_index_is_clust(index));
	ut_ad(!rec_get_deleted_flag(rec, rec_offs_comp(offsets)));

	if (UNIV_UNLIKELY(thr_get_trx(thr)->fake_changes)) {
		/* skip LOCK, UNDO, CHANGE, LOG */
		return(DB_SUCCESS);
	}

	err = lock_clust_rec_modify_check_and_lock(BTR_NO_LOCKING_FLAG, block,
						   rec, index, offsets, thr);

	if (err != DB_SUCCESS) {

		return(err);
	}

	err = trx_undo_report_row_operation(0, TRX_UNDO_MODIFY_OP, thr,
					    index, NULL, NULL, 0, rec, offsets,
					    &roll_ptr);
	if (err != DB_SUCCESS) {

		return(err);
	}

	/* The btr_search_latch is not needed here, because
	the adaptive hash index does not depend on the delete-mark
	and the delete-mark is being updated in place. */

	page_zip = buf_block_get_page_zip(block);

	btr_blob_dbg_set_deleted_flag(rec, index, offsets, TRUE);
	btr_rec_set_deleted_flag(rec, page_zip, TRUE);

	trx = thr_get_trx(thr);

	if (dict_index_is_online_ddl(index)) {
		row_log_table_delete(rec, index, offsets, NULL);
	}

	row_upd_rec_sys_fields(rec, page_zip, index, offsets, trx, roll_ptr);

	btr_cur_del_mark_set_clust_rec_log(rec, index, trx->id,
					   roll_ptr, mtr);

	return(err);
}

/****************************************************************//**
Writes the redo log record for a delete mark setting of a secondary
index record. */
UNIV_INLINE
void
btr_cur_del_mark_set_sec_rec_log(
/*=============================*/
	rec_t*		rec,	/*!< in: record */
	ibool		val,	/*!< in: value to set */
	mtr_t*		mtr)	/*!< in: mtr */
{
	byte*	log_ptr;
	ut_ad(val <= 1);

	log_ptr = mlog_open(mtr, 11 + 1 + 2);

	if (!log_ptr) {
		/* Logging in mtr is switched off during crash recovery:
		in that case mlog_open returns NULL */
		return;
	}

	log_ptr = mlog_write_initial_log_record_fast(
		rec, MLOG_REC_SEC_DELETE_MARK, log_ptr, mtr);
	mach_write_to_1(log_ptr, val);
	log_ptr++;

	mach_write_to_2(log_ptr, page_offset(rec));
	log_ptr += 2;

	mlog_close(mtr, log_ptr);
}
#endif /* !UNIV_HOTBACKUP */

/****************************************************************//**
Parses the redo log record for delete marking or unmarking of a secondary
index record.
@return	end of log record or NULL */
UNIV_INTERN
byte*
btr_cur_parse_del_mark_set_sec_rec(
/*===============================*/
	byte*		ptr,	/*!< in: buffer */
	byte*		end_ptr,/*!< in: buffer end */
	page_t*		page,	/*!< in/out: page or NULL */
	page_zip_des_t*	page_zip)/*!< in/out: compressed page, or NULL */
{
	ulint	val;
	ulint	offset;
	rec_t*	rec;

	if (end_ptr < ptr + 3) {

		return(NULL);
	}

	val = mach_read_from_1(ptr);
	ptr++;

	offset = mach_read_from_2(ptr);
	ptr += 2;

	ut_a(offset <= UNIV_PAGE_SIZE);

	if (page) {
		rec = page + offset;

		/* We do not need to reserve btr_search_latch, as the page
		is only being recovered, and there cannot be a hash index to
		it. Besides, the delete-mark flag is being updated in place
		and the adaptive hash index does not depend on it. */

		btr_rec_set_deleted_flag(rec, page_zip, val);
	}

	return(ptr);
}

#ifndef UNIV_HOTBACKUP
/***********************************************************//**
Sets a secondary index record delete mark to TRUE or FALSE.
@return	DB_SUCCESS, DB_LOCK_WAIT, or error number */
UNIV_INTERN
dberr_t
btr_cur_del_mark_set_sec_rec(
/*=========================*/
	ulint		flags,	/*!< in: locking flag */
	btr_cur_t*	cursor,	/*!< in: cursor */
	ibool		val,	/*!< in: value to set */
	que_thr_t*	thr,	/*!< in: query thread */
	mtr_t*		mtr)	/*!< in/out: mini-transaction */
{
	buf_block_t*	block;
	rec_t*		rec;
	dberr_t		err;

	if (UNIV_UNLIKELY(thr_get_trx(thr)->fake_changes)) {
		/* skip LOCK, CHANGE, LOG */
		return(DB_SUCCESS);
	}

	block = btr_cur_get_block(cursor);
	rec = btr_cur_get_rec(cursor);

#ifdef UNIV_DEBUG
	if (btr_cur_print_record_ops && thr) {
		btr_cur_trx_report(thr_get_trx(thr)->id, cursor->index,
				   "del mark ");
		rec_print(stderr, rec, cursor->index);
	}
#endif /* UNIV_DEBUG */

	err = lock_sec_rec_modify_check_and_lock(flags,
						 btr_cur_get_block(cursor),
						 rec, cursor->index, thr, mtr);
	if (err != DB_SUCCESS) {

		return(err);
	}

	ut_ad(!!page_rec_is_comp(rec)
	      == dict_table_is_comp(cursor->index->table));

	/* We do not need to reserve btr_search_latch, as the
	delete-mark flag is being updated in place and the adaptive
	hash index does not depend on it. */
	btr_rec_set_deleted_flag(rec, buf_block_get_page_zip(block), val);

	btr_cur_del_mark_set_sec_rec_log(rec, val, mtr);

	return(DB_SUCCESS);
}

/***********************************************************//**
Sets a secondary index record's delete mark to the given value. This
function is only used by the insert buffer merge mechanism. */
UNIV_INTERN
void
btr_cur_set_deleted_flag_for_ibuf(
/*==============================*/
	rec_t*		rec,		/*!< in/out: record */
	page_zip_des_t*	page_zip,	/*!< in/out: compressed page
					corresponding to rec, or NULL
					when the tablespace is
					uncompressed */
	ibool		val,		/*!< in: value to set */
	mtr_t*		mtr)		/*!< in/out: mini-transaction */
{
	/* We do not need to reserve btr_search_latch, as the page
	has just been read to the buffer pool and there cannot be
	a hash index to it.  Besides, the delete-mark flag is being
	updated in place and the adaptive hash index does not depend
	on it. */

	btr_rec_set_deleted_flag(rec, page_zip, val);

	btr_cur_del_mark_set_sec_rec_log(rec, val, mtr);
}

/*==================== B-TREE RECORD REMOVE =========================*/

/*************************************************************//**
Tries to compress a page of the tree if it seems useful. It is assumed
that mtr holds an x-latch on the tree and on the cursor page. To avoid
deadlocks, mtr must also own x-latches to brothers of page, if those
brothers exist. NOTE: it is assumed that the caller has reserved enough
free extents so that the compression will always succeed if done!
@return	TRUE if compression occurred */
UNIV_INTERN
ibool
btr_cur_compress_if_useful(
/*=======================*/
	btr_cur_t*	cursor,	/*!< in/out: cursor on the page to compress;
				cursor does not stay valid if !adjust and
				compression occurs */
	ibool		adjust,	/*!< in: TRUE if should adjust the
				cursor position even if compression occurs */
	mtr_t*		mtr)	/*!< in/out: mini-transaction */
{
	ut_ad(mtr_memo_contains(mtr,
				dict_index_get_lock(btr_cur_get_index(cursor)),
				MTR_MEMO_X_LOCK));
	ut_ad(mtr_memo_contains(mtr, btr_cur_get_block(cursor),
				MTR_MEMO_PAGE_X_FIX));

	return(btr_cur_compress_recommendation(cursor, mtr)
	       && btr_compress(cursor, adjust, mtr));
}

/*******************************************************//**
Removes the record on which the tree cursor is positioned on a leaf page.
It is assumed that the mtr has an x-latch on the page where the cursor is
positioned, but no latch on the whole tree.
@return	TRUE if success, i.e., the page did not become too empty */
UNIV_INTERN
ibool
btr_cur_optimistic_delete_func(
/*===========================*/
	btr_cur_t*	cursor,	/*!< in: cursor on leaf page, on the record to
				delete; cursor stays valid: if deletion
				succeeds, on function exit it points to the
				successor of the deleted record */
#ifdef UNIV_DEBUG
	ulint		flags,	/*!< in: BTR_CREATE_FLAG or 0 */
#endif /* UNIV_DEBUG */
	mtr_t*		mtr)	/*!< in: mtr; if this function returns
				TRUE on a leaf page of a secondary
				index, the mtr must be committed
				before latching any further pages */
{
	buf_block_t*	block;
	rec_t*		rec;
	mem_heap_t*	heap		= NULL;
	ulint		offsets_[REC_OFFS_NORMAL_SIZE];
	ulint*		offsets		= offsets_;
	ibool		no_compress_needed;
	rec_offs_init(offsets_);

	ut_ad(flags == 0 || flags == BTR_CREATE_FLAG);
	ut_ad(mtr_memo_contains(mtr, btr_cur_get_block(cursor),
				MTR_MEMO_PAGE_X_FIX));
	/* This is intended only for leaf page deletions */

	block = btr_cur_get_block(cursor);

	SRV_CORRUPT_TABLE_CHECK(block, return(DB_CORRUPTION););

	ut_ad(page_is_leaf(buf_block_get_frame(block)));
	ut_ad(!dict_index_is_online_ddl(cursor->index)
	      || dict_index_is_clust(cursor->index)
	      || (flags & BTR_CREATE_FLAG));

	rec = btr_cur_get_rec(cursor);
	offsets = rec_get_offsets(rec, cursor->index, offsets,
				  ULINT_UNDEFINED, &heap);

	no_compress_needed = !rec_offs_any_extern(offsets)
		&& btr_cur_can_delete_without_compress(
			cursor, rec_offs_size(offsets), mtr);

	if (no_compress_needed) {

		page_t*		page	= buf_block_get_frame(block);
		page_zip_des_t*	page_zip= buf_block_get_page_zip(block);

		lock_update_delete(block, rec);

		btr_search_update_hash_on_delete(cursor);

		if (page_zip) {
#ifdef UNIV_ZIP_DEBUG
			ut_a(page_zip_validate(page_zip, page, cursor->index));
#endif /* UNIV_ZIP_DEBUG */
			page_cur_delete_rec(btr_cur_get_page_cur(cursor),
					    cursor->index, offsets, mtr);
#ifdef UNIV_ZIP_DEBUG
			ut_a(page_zip_validate(page_zip, page, cursor->index));
#endif /* UNIV_ZIP_DEBUG */

			/* On compressed pages, the IBUF_BITMAP_FREE
			space is not affected by deleting (purging)
			records, because it is defined as the minimum
			of space available *without* reorganize, and
			space available in the modification log. */
		} else {
			const ulint	max_ins
				= page_get_max_insert_size_after_reorganize(
					page, 1);

			page_cur_delete_rec(btr_cur_get_page_cur(cursor),
					    cursor->index, offsets, mtr);

			/* The change buffer does not handle inserts
			into non-leaf pages, into clustered indexes,
			or into the change buffer. */
			if (page_is_leaf(page)
			    && !dict_index_is_clust(cursor->index)
			    && !dict_index_is_ibuf(cursor->index)) {
				ibuf_update_free_bits_low(block, max_ins, mtr);
			}
		}
	}

	if (UNIV_LIKELY_NULL(heap)) {
		mem_heap_free(heap);
	}

	return(no_compress_needed);
}

/*************************************************************//**
Removes the record on which the tree cursor is positioned. Tries
to compress the page if its fillfactor drops below a threshold
or if it is the only page on the level. It is assumed that mtr holds
an x-latch on the tree and on the cursor page. To avoid deadlocks,
mtr must also own x-latches to brothers of page, if those brothers
exist.
@return	TRUE if compression occurred */
UNIV_INTERN
ibool
btr_cur_pessimistic_delete(
/*=======================*/
	dberr_t*	err,	/*!< out: DB_SUCCESS or DB_OUT_OF_FILE_SPACE;
				the latter may occur because we may have
				to update node pointers on upper levels,
				and in the case of variable length keys
				these may actually grow in size */
	ibool		has_reserved_extents, /*!< in: TRUE if the
				caller has already reserved enough free
				extents so that he knows that the operation
				will succeed */
	btr_cur_t*	cursor,	/*!< in: cursor on the record to delete;
				if compression does not occur, the cursor
				stays valid: it points to successor of
				deleted record on function exit */
	ulint		flags,	/*!< in: BTR_CREATE_FLAG or 0 */
	enum trx_rb_ctx	rb_ctx,	/*!< in: rollback context */
	mtr_t*		mtr)	/*!< in: mtr */
{
	buf_block_t*	block;
	page_t*		page;
	page_zip_des_t*	page_zip;
	dict_index_t*	index;
	rec_t*		rec;
	ulint		n_reserved	= 0;
	ibool		success;
	ibool		ret		= FALSE;
	ulint		level;
	mem_heap_t*	heap;
	ulint*		offsets;

	block = btr_cur_get_block(cursor);
	page = buf_block_get_frame(block);
	index = btr_cur_get_index(cursor);

	ut_ad(flags == 0 || flags == BTR_CREATE_FLAG);
	ut_ad(!dict_index_is_online_ddl(index)
	      || dict_index_is_clust(index)
	      || (flags & BTR_CREATE_FLAG));
	ut_ad(mtr_memo_contains(mtr, dict_index_get_lock(index),
				MTR_MEMO_X_LOCK));
	ut_ad(mtr_memo_contains(mtr, block, MTR_MEMO_PAGE_X_FIX));
	if (!has_reserved_extents) {
		/* First reserve enough free space for the file segments
		of the index tree, so that the node pointer updates will
		not fail because of lack of space */

		ut_a(cursor->tree_height != ULINT_UNDEFINED);

		ulint	n_extents = cursor->tree_height / 32 + 1;

		success = fsp_reserve_free_extents(&n_reserved,
						   index->space,
						   n_extents,
						   FSP_CLEANING, mtr);
		if (!success) {
			*err = DB_OUT_OF_FILE_SPACE;

			return(FALSE);
		}
	}

	heap = mem_heap_create(1024);
	rec = btr_cur_get_rec(cursor);
	page_zip = buf_block_get_page_zip(block);
#ifdef UNIV_ZIP_DEBUG
	ut_a(!page_zip || page_zip_validate(page_zip, page, index));
#endif /* UNIV_ZIP_DEBUG */

	offsets = rec_get_offsets(rec, index, NULL, ULINT_UNDEFINED, &heap);

	if (rec_offs_any_extern(offsets)) {
		btr_rec_free_externally_stored_fields(index,
						      rec, offsets, page_zip,
						      rb_ctx, mtr);
#ifdef UNIV_ZIP_DEBUG
		ut_a(!page_zip || page_zip_validate(page_zip, page, index));
#endif /* UNIV_ZIP_DEBUG */
	}

	if (UNIV_UNLIKELY(page_get_n_recs(page) < 2)
	    && UNIV_UNLIKELY(dict_index_get_page(index)
			     != buf_block_get_page_no(block))) {

		/* If there is only one record, drop the whole page in
		btr_discard_page, if this is not the root page */

		btr_discard_page(cursor, mtr);

		ret = TRUE;

		goto return_after_reservations;
	}

	if (flags == 0) {
		lock_update_delete(block, rec);
	}

	level = btr_page_get_level(page, mtr);

	if (level > 0
	    && UNIV_UNLIKELY(rec == page_rec_get_next(
				     page_get_infimum_rec(page)))) {

		rec_t*	next_rec = page_rec_get_next(rec);

		if (btr_page_get_prev(page, mtr) == FIL_NULL) {

			/* If we delete the leftmost node pointer on a
			non-leaf level, we must mark the new leftmost node
			pointer as the predefined minimum record */

			/* This will make page_zip_validate() fail until
			page_cur_delete_rec() completes.  This is harmless,
			because everything will take place within a single
			mini-transaction and because writing to the redo log
			is an atomic operation (performed by mtr_commit()). */
			btr_set_min_rec_mark(next_rec, mtr);
		} else {
			/* Otherwise, if we delete the leftmost node pointer
			on a page, we have to change the father node pointer
			so that it is equal to the new leftmost node pointer
			on the page */

			btr_node_ptr_delete(index, block, mtr);

			dtuple_t*	node_ptr = dict_index_build_node_ptr(
				index, next_rec, buf_block_get_page_no(block),
				heap, level);

			btr_insert_on_non_leaf_level(
				flags, index, level + 1, node_ptr, mtr);
		}
	}

	btr_search_update_hash_on_delete(cursor);

	page_cur_delete_rec(btr_cur_get_page_cur(cursor), index, offsets, mtr);
#ifdef UNIV_ZIP_DEBUG
	ut_a(!page_zip || page_zip_validate(page_zip, page, index));
#endif /* UNIV_ZIP_DEBUG */

	ut_ad(btr_check_node_ptr(index, block, mtr));

return_after_reservations:
	*err = DB_SUCCESS;

	mem_heap_free(heap);

	if (ret == FALSE) {
		ret = btr_cur_compress_if_useful(cursor, FALSE, mtr);
	}

	if (n_reserved > 0) {
		fil_space_release_free_extents(index->space, n_reserved);
	}

	return(ret);
}

/*******************************************************************//**
Adds path information to the cursor for the current page, for which
the binary search has been performed. */
static
void
btr_cur_add_path_info(
/*==================*/
	btr_cur_t*	cursor,		/*!< in: cursor positioned on a page */
	ulint		height,		/*!< in: height of the page in tree;
					0 means leaf node */
	ulint		root_height)	/*!< in: root node height in tree */
{
	btr_path_t*	slot;
	const rec_t*	rec;
	const page_t*	page;

	ut_a(cursor->path_arr);

	if (root_height >= BTR_PATH_ARRAY_N_SLOTS - 1) {
		/* Do nothing; return empty path */

		slot = cursor->path_arr;
		slot->nth_rec = ULINT_UNDEFINED;

		return;
	}

	if (height == 0) {
		/* Mark end of slots for path */
		slot = cursor->path_arr + root_height + 1;
		slot->nth_rec = ULINT_UNDEFINED;
	}

	rec = btr_cur_get_rec(cursor);

	slot = cursor->path_arr + (root_height - height);

	page = page_align(rec);

	slot->nth_rec = page_rec_get_n_recs_before(rec);
	slot->n_recs = page_get_n_recs(page);
	slot->page_no = page_get_page_no(page);
	slot->page_level = btr_page_get_level_low(page);
}

/*******************************************************************//**
Estimate the number of rows between slot1 and slot2 for any level on a
B-tree. This function starts from slot1->page and reads a few pages to
the right, counting their records. If we reach slot2->page quickly then
we know exactly how many records there are between slot1 and slot2 and
we set is_n_rows_exact to TRUE. If we cannot reach slot2->page quickly
then we calculate the average number of records in the pages scanned
so far and assume that all pages that we did not scan up to slot2->page
contain the same number of records, then we multiply that average to
the number of pages between slot1->page and slot2->page (which is
n_rows_on_prev_level). In this case we set is_n_rows_exact to FALSE.
@return	number of rows (exact or estimated) */
static
ib_int64_t
btr_estimate_n_rows_in_range_on_level(
/*==================================*/
	dict_index_t*	index,			/*!< in: index */
	btr_path_t*	slot1,			/*!< in: left border */
	btr_path_t*	slot2,			/*!< in: right border */
	ib_int64_t	n_rows_on_prev_level,	/*!< in: number of rows
						on the previous level for the
						same descend paths; used to
						determine the numbe of pages
						on this level */
	ibool*		is_n_rows_exact)	/*!< out: TRUE if the returned
						value is exact i.e. not an
						estimation */
{
	ulint		space;
	ib_int64_t	n_rows;
	ulint		n_pages_read;
	ulint		page_no;
	ulint		zip_size;
	ulint		level;

	space = dict_index_get_space(index);

	n_rows = 0;
	n_pages_read = 0;

	/* Assume by default that we will scan all pages between
	slot1->page_no and slot2->page_no */
	*is_n_rows_exact = TRUE;

	/* add records from slot1->page_no which are to the right of
	the record which serves as a left border of the range, if any */
	if (slot1->nth_rec < slot1->n_recs) {
		n_rows += slot1->n_recs - slot1->nth_rec;
	}

	/* add records from slot2->page_no which are to the left of
	the record which servers as a right border of the range, if any */
	if (slot2->nth_rec > 1) {
		n_rows += slot2->nth_rec - 1;
	}

	/* count the records in the pages between slot1->page_no and
	slot2->page_no (non inclusive), if any */

	zip_size = fil_space_get_zip_size(space);

	/* Do not read more than this number of pages in order not to hurt
	performance with this code which is just an estimation. If we read
	this many pages before reaching slot2->page_no then we estimate the
	average from the pages scanned so far */
#	define N_PAGES_READ_LIMIT	10

	page_no = slot1->page_no;
	level = slot1->page_level;

	do {
		mtr_t		mtr;
		page_t*		page;
		buf_block_t*	block;
		dberr_t		err=DB_SUCCESS;

		mtr_start(&mtr);

		/* Fetch the page. Because we are not holding the
		index->lock, the tree may have changed and we may be
		attempting to read a page that is no longer part of
		the B-tree. We pass BUF_GET_POSSIBLY_FREED in order to
		silence a debug assertion about this. */
		block = buf_page_get_gen(space, zip_size, page_no, RW_S_LATCH,
					 NULL, BUF_GET_POSSIBLY_FREED,
					 __FILE__, __LINE__, &mtr, &err);

		if (err != DB_SUCCESS) {
			if (err == DB_DECRYPTION_FAILED) {
				ib_push_warning((void *)NULL,
					DB_DECRYPTION_FAILED,
					"Table %s is encrypted but encryption service or"
					" used key_id is not available. "
					" Can't continue reading table.",
					index->table->name);
				index->table->is_encrypted = true;
			}

			mtr_commit(&mtr);
			goto inexact;
		}


		page = buf_block_get_frame(block);

		/* It is possible that the tree has been reorganized in the
		meantime and this is a different page. If this happens the
		calculated estimate will be bogus, which is not fatal as
		this is only an estimate. We are sure that a page with
		page_no exists because InnoDB never frees pages, only
		reuses them. */
		if (fil_page_get_type(page) != FIL_PAGE_INDEX
		    || btr_page_get_index_id(page) != index->id
		    || btr_page_get_level_low(page) != level) {

			/* The page got reused for something else */
			mtr_commit(&mtr);
			goto inexact;
		}

		/* It is possible but highly unlikely that the page was
		originally written by an old version of InnoDB that did
		not initialize FIL_PAGE_TYPE on other than B-tree pages.
		For example, this could be an almost-empty BLOB page
		that happens to contain the magic values in the fields
		that we checked above. */

		n_pages_read++;

		if (page_no != slot1->page_no) {
			/* Do not count the records on slot1->page_no,
			we already counted them before this loop. */
			n_rows += page_get_n_recs(page);
		}

		page_no = btr_page_get_next(page, &mtr);

		mtr_commit(&mtr);

		if (n_pages_read == N_PAGES_READ_LIMIT
		    || page_no == FIL_NULL) {
			/* Either we read too many pages or
			we reached the end of the level without passing
			through slot2->page_no, the tree must have changed
			in the meantime */
			goto inexact;
		}

	} while (page_no != slot2->page_no);

	return(n_rows);

inexact:

	*is_n_rows_exact = FALSE;

	/* We did interrupt before reaching slot2->page */

	if (n_pages_read > 0) {
		/* The number of pages on this level is
		n_rows_on_prev_level, multiply it by the
		average number of recs per page so far */
		n_rows = n_rows_on_prev_level
			* n_rows / n_pages_read;
	} else {
		/* The tree changed before we could even
		start with slot1->page_no */
		n_rows = 10;
	}

	return(n_rows);
}

/*******************************************************************//**
Estimates the number of rows in a given index range.
@return	estimated number of rows */
UNIV_INTERN
ib_int64_t
btr_estimate_n_rows_in_range(
/*=========================*/
	dict_index_t*	index,	/*!< in: index */
	const dtuple_t*	tuple1,	/*!< in: range start, may also be empty tuple */
	ulint		mode1,	/*!< in: search mode for range start */
	const dtuple_t*	tuple2,	/*!< in: range end, may also be empty tuple */
	ulint		mode2,	/*!< in: search mode for range end */
	trx_t*		trx)	/*!< in: trx */
{
	btr_path_t	path1[BTR_PATH_ARRAY_N_SLOTS];
	btr_path_t	path2[BTR_PATH_ARRAY_N_SLOTS];
	btr_cur_t	cursor;
	btr_path_t*	slot1;
	btr_path_t*	slot2;
	ibool		diverged;
	ibool		diverged_lot;
	ulint		divergence_level;
	ib_int64_t	n_rows;
	ibool		is_n_rows_exact;
	ulint		i;
	mtr_t		mtr;
	ib_int64_t	table_n_rows;

	table_n_rows = dict_table_get_n_rows(index->table);

	mtr_start_trx(&mtr, trx);

	cursor.path_arr = path1;

	if (dtuple_get_n_fields(tuple1) > 0) {

		btr_cur_search_to_nth_level(index, 0, tuple1, mode1,
					    BTR_SEARCH_LEAF | BTR_ESTIMATE,
					    &cursor, 0,
					    __FILE__, __LINE__, &mtr);
	} else {
		btr_cur_open_at_index_side(true, index,
					   BTR_SEARCH_LEAF | BTR_ESTIMATE,
					   &cursor, 0, &mtr);
	}

	mtr_commit(&mtr);

	mtr_start_trx(&mtr, trx);

	cursor.path_arr = path2;

	if (dtuple_get_n_fields(tuple2) > 0) {

		btr_cur_search_to_nth_level(index, 0, tuple2, mode2,
					    BTR_SEARCH_LEAF | BTR_ESTIMATE,
					    &cursor, 0,
					    __FILE__, __LINE__, &mtr);
	} else {
		btr_cur_open_at_index_side(false, index,
					   BTR_SEARCH_LEAF | BTR_ESTIMATE,
					   &cursor, 0, &mtr);
	}

	mtr_commit(&mtr);

	/* We have the path information for the range in path1 and path2 */

	n_rows = 1;
	is_n_rows_exact = TRUE;
	diverged = FALSE;	    /* This becomes true when the path is not
				    the same any more */
	diverged_lot = FALSE;	    /* This becomes true when the paths are
				    not the same or adjacent any more */
	divergence_level = 1000000; /* This is the level where paths diverged
				    a lot */
	for (i = 0; ; i++) {
		ut_ad(i < BTR_PATH_ARRAY_N_SLOTS);

		slot1 = path1 + i;
		slot2 = path2 + i;

		if (slot1->nth_rec == ULINT_UNDEFINED
		    || slot2->nth_rec == ULINT_UNDEFINED) {

			if (i > divergence_level + 1 && !is_n_rows_exact) {
				/* In trees whose height is > 1 our algorithm
				tends to underestimate: multiply the estimate
				by 2: */

				n_rows = n_rows * 2;
			}

			DBUG_EXECUTE_IF("bug14007649", return(n_rows););

			/* Do not estimate the number of rows in the range
			to over 1 / 2 of the estimated rows in the whole
			table */

			if (n_rows > table_n_rows / 2 && !is_n_rows_exact) {

				n_rows = table_n_rows / 2;

				/* If there are just 0 or 1 rows in the table,
				then we estimate all rows are in the range */

				if (n_rows == 0) {
					n_rows = table_n_rows;
				}
			}

			return(n_rows);
		}

		if (!diverged && slot1->nth_rec != slot2->nth_rec) {

			diverged = TRUE;

			if (slot1->nth_rec < slot2->nth_rec) {
				n_rows = slot2->nth_rec - slot1->nth_rec;

				if (n_rows > 1) {
					diverged_lot = TRUE;
					divergence_level = i;
				}
			} else {
				/* It is possible that
				slot1->nth_rec >= slot2->nth_rec
				if, for example, we have a single page
				tree which contains (inf, 5, 6, supr)
				and we select where x > 20 and x < 30;
				in this case slot1->nth_rec will point
				to the supr record and slot2->nth_rec
				will point to 6 */
				n_rows = 0;
			}

		} else if (diverged && !diverged_lot) {

			if (slot1->nth_rec < slot1->n_recs
			    || slot2->nth_rec > 1) {

				diverged_lot = TRUE;
				divergence_level = i;

				n_rows = 0;

				if (slot1->nth_rec < slot1->n_recs) {
					n_rows += slot1->n_recs
						- slot1->nth_rec;
				}

				if (slot2->nth_rec > 1) {
					n_rows += slot2->nth_rec - 1;
				}
			}
		} else if (diverged_lot) {

			n_rows = btr_estimate_n_rows_in_range_on_level(
				index, slot1, slot2, n_rows,
				&is_n_rows_exact);
		}
	}
}

/*******************************************************************//**
Record the number of non_null key values in a given index for
each n-column prefix of the index where 1 <= n <= dict_index_get_n_unique(index).
The estimates are eventually stored in the array:
index->stat_n_non_null_key_vals[], which is indexed from 0 to n-1. */
static
void
btr_record_not_null_field_in_rec(
/*=============================*/
	ulint		n_unique,	/*!< in: dict_index_get_n_unique(index),
					number of columns uniquely determine
					an index entry */
	const ulint*	offsets,	/*!< in: rec_get_offsets(rec, index),
					its size could be for all fields or
					that of "n_unique" */
	ib_uint64_t*	n_not_null)	/*!< in/out: array to record number of
					not null rows for n-column prefix */
{
	ulint	i;

	ut_ad(rec_offs_n_fields(offsets) >= n_unique);

	if (n_not_null == NULL) {
		return;
	}

	for (i = 0; i < n_unique; i++) {
		if (rec_offs_nth_sql_null(offsets, i)) {
			break;
		}

		n_not_null[i]++;
	}
}

/*******************************************************************//**
Estimates the number of different key values in a given index, for
each n-column prefix of the index where 1 <= n <= dict_index_get_n_unique(index).
The estimates are stored in the array index->stat_n_diff_key_vals[] (indexed
0..n_uniq-1) and the number of pages that were sampled is saved in
index->stat_n_sample_sizes[].
If innodb_stats_method is nulls_ignored, we also record the number of
non-null values for each prefix and stored the estimates in
array index->stat_n_non_null_key_vals. */
UNIV_INTERN
void
btr_estimate_number_of_different_key_vals(
/*======================================*/
	dict_index_t*	index)	/*!< in: index */
{
	btr_cur_t	cursor;
	page_t*		page;
	rec_t*		rec;
	ulint		n_cols;
	ulint		matched_fields;
	ulint		matched_bytes;
	ib_uint64_t*	n_diff;
	ib_uint64_t*	n_not_null;
	ibool		stats_null_not_equal;
	ullint		n_sample_pages=1; /* number of pages to sample */
	ulint		not_empty_flag	= 0;
	ulint		total_external_size = 0;
	ulint		i;
	ulint		j;
	ullint		add_on;
	mtr_t		mtr;
	mem_heap_t*	heap		= NULL;
	ulint*		offsets_rec	= NULL;
	ulint*		offsets_next_rec = NULL;

	n_cols = dict_index_get_n_unique(index);

	heap = mem_heap_create((sizeof *n_diff + sizeof *n_not_null)
			       * n_cols
			       + dict_index_get_n_fields(index)
			       * (sizeof *offsets_rec
				  + sizeof *offsets_next_rec));

	n_diff = (ib_uint64_t*) mem_heap_zalloc(
		heap, n_cols * sizeof(ib_int64_t));

	n_not_null = NULL;

	/* Check srv_innodb_stats_method setting, and decide whether we
	need to record non-null value and also decide if NULL is
	considered equal (by setting stats_null_not_equal value) */
	switch (srv_innodb_stats_method) {
	case SRV_STATS_NULLS_IGNORED:
		n_not_null = (ib_uint64_t*) mem_heap_zalloc(
			heap, n_cols * sizeof *n_not_null);
		/* fall through */

	case SRV_STATS_NULLS_UNEQUAL:
		/* for both SRV_STATS_NULLS_IGNORED and SRV_STATS_NULLS_UNEQUAL
		case, we will treat NULLs as unequal value */
		stats_null_not_equal = TRUE;
		break;

	case SRV_STATS_NULLS_EQUAL:
		stats_null_not_equal = FALSE;
		break;

	default:
		ut_error;
        }

	if (srv_stats_sample_traditional) {
		/* It makes no sense to test more pages than are contained
		in the index, thus we lower the number if it is too high */
		if (srv_stats_transient_sample_pages > index->stat_index_size) {
			if (index->stat_index_size > 0) {
				n_sample_pages = index->stat_index_size;
			}
		} else {
			n_sample_pages = srv_stats_transient_sample_pages;
		}
	} else {
		/* New logaritmic number of pages that are estimated.
		Number of pages estimated should be between 1 and
		index->stat_index_size.

		If we have only 0 or 1 index pages then we can only take 1
		sample. We have already initialized n_sample_pages to 1.

		So taking index size as I and sample as S and log(I)*S as L

		requirement 1) we want the out limit of the expression to not exceed I;
		requirement 2) we want the ideal pages to be at least S;
		so the current expression is min(I, max( min(S,I), L)

		looking for simplifications:

		case 1: assume S < I
		min(I, max( min(S,I), L) -> min(I , max( S, L))

		but since L=LOG2(I)*S and log2(I) >=1   L>S always so max(S,L) = L.

		so we have: min(I , L)

		case 2: assume I < S
		    min(I, max( min(S,I), L) -> min(I, max( I, L))

		case 2a: L > I
		    min(I, max( I, L)) -> min(I, L) -> I

		case 2b: when L < I
		    min(I, max( I, L))  ->  min(I, I ) -> I

		so taking all case2 paths is I, our expression is:
		n_pages = S < I? min(I,L) : I
                */
		if (index->stat_index_size > 1) {
			n_sample_pages = (srv_stats_transient_sample_pages < index->stat_index_size) ?
				ut_min(index->stat_index_size,
				       log2(index->stat_index_size)*srv_stats_transient_sample_pages)
				: index->stat_index_size;

		}
	}

	/* Sanity check */
	ut_ad(n_sample_pages > 0 && n_sample_pages <= (index->stat_index_size < 1 ? 1 : index->stat_index_size));

	/* We sample some pages in the index to get an estimate */

	for (i = 0; i < n_sample_pages; i++) {
		mtr_start(&mtr);

		btr_cur_open_at_rnd_pos(index, BTR_SEARCH_LEAF, &cursor, &mtr);

		/* Count the number of different key values for each prefix of
		the key on this index page. If the prefix does not determine
		the index record uniquely in the B-tree, then we subtract one
		because otherwise our algorithm would give a wrong estimate
		for an index where there is just one key value. */

		page = btr_cur_get_page(&cursor);

		SRV_CORRUPT_TABLE_CHECK(page, goto exit_loop;);

		rec = page_rec_get_next(page_get_infimum_rec(page));

		if (!page_rec_is_supremum(rec)) {
			not_empty_flag = 1;
			offsets_rec = rec_get_offsets(rec, index, offsets_rec,
						      ULINT_UNDEFINED, &heap);

			if (n_not_null != NULL) {
				btr_record_not_null_field_in_rec(
					n_cols, offsets_rec, n_not_null);
			}
		}

		while (!page_rec_is_supremum(rec)) {
			rec_t*	next_rec = page_rec_get_next(rec);
			if (page_rec_is_supremum(next_rec)) {
				total_external_size +=
					btr_rec_get_externally_stored_len(
						rec, offsets_rec);
				break;
			}

			matched_fields = 0;
			matched_bytes = 0;
			offsets_next_rec = rec_get_offsets(next_rec, index,
							   offsets_next_rec,
							   ULINT_UNDEFINED,
							   &heap);

			cmp_rec_rec_with_match(rec, next_rec,
					       offsets_rec, offsets_next_rec,
					       index, stats_null_not_equal,
					       &matched_fields,
					       &matched_bytes);

			for (j = matched_fields; j < n_cols; j++) {
				/* We add one if this index record has
				a different prefix from the previous */

				n_diff[j]++;
			}

			if (n_not_null != NULL) {
				btr_record_not_null_field_in_rec(
					n_cols, offsets_next_rec, n_not_null);
			}

			total_external_size
				+= btr_rec_get_externally_stored_len(
					rec, offsets_rec);

			rec = next_rec;
			/* Initialize offsets_rec for the next round
			and assign the old offsets_rec buffer to
			offsets_next_rec. */
			{
				ulint*	offsets_tmp = offsets_rec;
				offsets_rec = offsets_next_rec;
				offsets_next_rec = offsets_tmp;
			}
		}


		if (n_cols == dict_index_get_n_unique_in_tree(index)) {

			/* If there is more than one leaf page in the tree,
			we add one because we know that the first record
			on the page certainly had a different prefix than the
			last record on the previous index page in the
			alphabetical order. Before this fix, if there was
			just one big record on each clustered index page, the
			algorithm grossly underestimated the number of rows
			in the table. */

			if (btr_page_get_prev(page, &mtr) != FIL_NULL
			    || btr_page_get_next(page, &mtr) != FIL_NULL) {

				n_diff[n_cols - 1]++;
			}
		}

		mtr_commit(&mtr);
	}

exit_loop:
	/* If we saw k borders between different key values on
	n_sample_pages leaf pages, we can estimate how many
	there will be in index->stat_n_leaf_pages */

	/* We must take into account that our sample actually represents
	also the pages used for external storage of fields (those pages are
	included in index->stat_n_leaf_pages) */

	for (j = 0; j < n_cols; j++) {
		index->stat_n_diff_key_vals[j]
			= BTR_TABLE_STATS_FROM_SAMPLE(
				n_diff[j], index, n_sample_pages,
				total_external_size, not_empty_flag);

		/* If the tree is small, smaller than
		10 * n_sample_pages + total_external_size, then
		the above estimate is ok. For bigger trees it is common that we
		do not see any borders between key values in the few pages
		we pick. But still there may be n_sample_pages
		different key values, or even more. Let us try to approximate
		that: */

		add_on = index->stat_n_leaf_pages
			/ (10 * (n_sample_pages
				 + total_external_size));

		if (add_on > n_sample_pages) {
			add_on = n_sample_pages;
		}

		index->stat_n_diff_key_vals[j] += add_on;

		index->stat_n_sample_sizes[j] = n_sample_pages;

		/* Update the stat_n_non_null_key_vals[] with our
		sampled result. stat_n_non_null_key_vals[] is created
		and initialized to zero in dict_index_add_to_cache(),
		along with stat_n_diff_key_vals[] array */
		if (n_not_null != NULL) {
			index->stat_n_non_null_key_vals[j] =
				 BTR_TABLE_STATS_FROM_SAMPLE(
					n_not_null[j], index, n_sample_pages,
					total_external_size, not_empty_flag);
		}
	}

	mem_heap_free(heap);
}

/*================== EXTERNAL STORAGE OF BIG FIELDS ===================*/

/***********************************************************//**
Gets the offset of the pointer to the externally stored part of a field.
@return	offset of the pointer to the externally stored part */
static
ulint
btr_rec_get_field_ref_offs(
/*=======================*/
	const ulint*	offsets,/*!< in: array returned by rec_get_offsets() */
	ulint		n)	/*!< in: index of the external field */
{
	ulint	field_ref_offs;
	ulint	local_len;

	ut_a(rec_offs_nth_extern(offsets, n));
	field_ref_offs = rec_get_nth_field_offs(offsets, n, &local_len);
	ut_a(local_len != UNIV_SQL_NULL);
	ut_a(local_len >= BTR_EXTERN_FIELD_REF_SIZE);

	return(field_ref_offs + local_len - BTR_EXTERN_FIELD_REF_SIZE);
}

/** Gets a pointer to the externally stored part of a field.
@param rec	record
@param offsets	rec_get_offsets(rec)
@param n	index of the externally stored field
@return pointer to the externally stored part */
#define btr_rec_get_field_ref(rec, offsets, n)			\
	((rec) + btr_rec_get_field_ref_offs(offsets, n))

/** Gets the externally stored size of a record, in units of a database page.
@param[in]	rec	record
@param[in]	offsets	array returned by rec_get_offsets()
@return	externally stored part, in units of a database page */

ulint
btr_rec_get_externally_stored_len(
	const rec_t*	rec,
	const ulint*	offsets)
{
	ulint	n_fields;
	ulint	total_extern_len = 0;
	ulint	i;

	ut_ad(!rec_offs_comp(offsets) || !rec_get_node_ptr_flag(rec));

	if (!rec_offs_any_extern(offsets)) {
		return(0);
	}

	n_fields = rec_offs_n_fields(offsets);

	for (i = 0; i < n_fields; i++) {
		if (rec_offs_nth_extern(offsets, i)) {

			ulint	extern_len = mach_read_from_4(
				btr_rec_get_field_ref(rec, offsets, i)
				+ BTR_EXTERN_LEN + 4);

			total_extern_len += ut_calc_align(extern_len,
							  UNIV_PAGE_SIZE);
		}
	}

	return(total_extern_len / UNIV_PAGE_SIZE);
}

/*******************************************************************//**
Sets the ownership bit of an externally stored field in a record. */
static
void
btr_cur_set_ownership_of_extern_field(
/*==================================*/
	page_zip_des_t*	page_zip,/*!< in/out: compressed page whose uncompressed
				part will be updated, or NULL */
	rec_t*		rec,	/*!< in/out: clustered index record */
	dict_index_t*	index,	/*!< in: index of the page */
	const ulint*	offsets,/*!< in: array returned by rec_get_offsets() */
	ulint		i,	/*!< in: field number */
	ibool		val,	/*!< in: value to set */
	mtr_t*		mtr)	/*!< in: mtr, or NULL if not logged */
{
	byte*	data;
	ulint	local_len;
	ulint	byte_val;

	data = rec_get_nth_field(rec, offsets, i, &local_len);
	ut_ad(rec_offs_nth_extern(offsets, i));
	ut_a(local_len >= BTR_EXTERN_FIELD_REF_SIZE);

	local_len -= BTR_EXTERN_FIELD_REF_SIZE;

	byte_val = mach_read_from_1(data + local_len + BTR_EXTERN_LEN);

	if (val) {
		byte_val = byte_val & (~BTR_EXTERN_OWNER_FLAG);
	} else {
#if defined UNIV_DEBUG || defined UNIV_BLOB_LIGHT_DEBUG
		ut_a(!(byte_val & BTR_EXTERN_OWNER_FLAG));
#endif /* UNIV_DEBUG || UNIV_BLOB_LIGHT_DEBUG */
		byte_val = byte_val | BTR_EXTERN_OWNER_FLAG;
	}

	if (page_zip) {
		mach_write_to_1(data + local_len + BTR_EXTERN_LEN, byte_val);
		page_zip_write_blob_ptr(page_zip, rec, index, offsets, i, mtr);
	} else if (mtr != NULL) {

		mlog_write_ulint(data + local_len + BTR_EXTERN_LEN, byte_val,
				 MLOG_1BYTE, mtr);
	} else {
		mach_write_to_1(data + local_len + BTR_EXTERN_LEN, byte_val);
	}

	btr_blob_dbg_owner(rec, index, offsets, i, val);
}

/*******************************************************************//**
Marks non-updated off-page fields as disowned by this record. The ownership
must be transferred to the updated record which is inserted elsewhere in the
index tree. In purge only the owner of externally stored field is allowed
to free the field. */
UNIV_INTERN
void
btr_cur_disown_inherited_fields(
/*============================*/
	page_zip_des_t*	page_zip,/*!< in/out: compressed page whose uncompressed
				part will be updated, or NULL */
	rec_t*		rec,	/*!< in/out: record in a clustered index */
	dict_index_t*	index,	/*!< in: index of the page */
	const ulint*	offsets,/*!< in: array returned by rec_get_offsets() */
	const upd_t*	update,	/*!< in: update vector */
	mtr_t*		mtr)	/*!< in/out: mini-transaction */
{
	ulint	i;

	ut_ad(rec_offs_validate(rec, index, offsets));
	ut_ad(!rec_offs_comp(offsets) || !rec_get_node_ptr_flag(rec));
	ut_ad(rec_offs_any_extern(offsets));
	ut_ad(mtr);

	for (i = 0; i < rec_offs_n_fields(offsets); i++) {
		if (rec_offs_nth_extern(offsets, i)
		    && !upd_get_field_by_field_no(update, i)) {
			btr_cur_set_ownership_of_extern_field(
				page_zip, rec, index, offsets, i, FALSE, mtr);
		}
	}
}

/*******************************************************************//**
Marks all extern fields in a record as owned by the record. This function
should be called if the delete mark of a record is removed: a not delete
marked record always owns all its extern fields. */
static
void
btr_cur_unmark_extern_fields(
/*=========================*/
	page_zip_des_t*	page_zip,/*!< in/out: compressed page whose uncompressed
				part will be updated, or NULL */
	rec_t*		rec,	/*!< in/out: record in a clustered index */
	dict_index_t*	index,	/*!< in: index of the page */
	const ulint*	offsets,/*!< in: array returned by rec_get_offsets() */
	mtr_t*		mtr)	/*!< in: mtr, or NULL if not logged */
{
	ulint	n;
	ulint	i;

	ut_ad(!rec_offs_comp(offsets) || !rec_get_node_ptr_flag(rec));
	n = rec_offs_n_fields(offsets);

	if (!rec_offs_any_extern(offsets)) {

		return;
	}

	for (i = 0; i < n; i++) {
		if (rec_offs_nth_extern(offsets, i)) {

			btr_cur_set_ownership_of_extern_field(
				page_zip, rec, index, offsets, i, TRUE, mtr);
		}
	}
}

/*******************************************************************//**
Flags the data tuple fields that are marked as extern storage in the
update vector.  We use this function to remember which fields we must
mark as extern storage in a record inserted for an update.
@return	number of flagged external columns */
UNIV_INTERN
ulint
btr_push_update_extern_fields(
/*==========================*/
	dtuple_t*	tuple,	/*!< in/out: data tuple */
	const upd_t*	update,	/*!< in: update vector */
	mem_heap_t*	heap)	/*!< in: memory heap */
{
	ulint			n_pushed	= 0;
	ulint			n;
	const upd_field_t*	uf;

	ut_ad(tuple);
	ut_ad(update);

	uf = update->fields;
	n = upd_get_n_fields(update);

	for (; n--; uf++) {
		if (dfield_is_ext(&uf->new_val)) {
			dfield_t*	field
				= dtuple_get_nth_field(tuple, uf->field_no);

			if (!dfield_is_ext(field)) {
				dfield_set_ext(field);
				n_pushed++;
			}

			switch (uf->orig_len) {
				byte*	data;
				ulint	len;
				byte*	buf;
			case 0:
				break;
			case BTR_EXTERN_FIELD_REF_SIZE:
				/* Restore the original locally stored
				part of the column.  In the undo log,
				InnoDB writes a longer prefix of externally
				stored columns, so that column prefixes
				in secondary indexes can be reconstructed. */
				dfield_set_data(field, (byte*) dfield_get_data(field)
						+ dfield_get_len(field)
						- BTR_EXTERN_FIELD_REF_SIZE,
						BTR_EXTERN_FIELD_REF_SIZE);
				dfield_set_ext(field);
				break;
			default:
				/* Reconstruct the original locally
				stored part of the column.  The data
				will have to be copied. */
				ut_a(uf->orig_len > BTR_EXTERN_FIELD_REF_SIZE);

				data = (byte*) dfield_get_data(field);
				len = dfield_get_len(field);

				buf = (byte*) mem_heap_alloc(heap,
							     uf->orig_len);
				/* Copy the locally stored prefix. */
				memcpy(buf, data,
				       uf->orig_len
				       - BTR_EXTERN_FIELD_REF_SIZE);
				/* Copy the BLOB pointer. */
				memcpy(buf + uf->orig_len
				       - BTR_EXTERN_FIELD_REF_SIZE,
				       data + len - BTR_EXTERN_FIELD_REF_SIZE,
				       BTR_EXTERN_FIELD_REF_SIZE);

				dfield_set_data(field, buf, uf->orig_len);
				dfield_set_ext(field);
			}
		}
	}

	return(n_pushed);
}

/*******************************************************************//**
Returns the length of a BLOB part stored on the header page.
@return	part length */
static
ulint
btr_blob_get_part_len(
/*==================*/
	const byte*	blob_header)	/*!< in: blob header */
{
	return(mach_read_from_4(blob_header + BTR_BLOB_HDR_PART_LEN));
}

/*******************************************************************//**
Returns the page number where the next BLOB part is stored.
@return	page number or FIL_NULL if no more pages */
static
ulint
btr_blob_get_next_page_no(
/*======================*/
	const byte*	blob_header)	/*!< in: blob header */
{
	return(mach_read_from_4(blob_header + BTR_BLOB_HDR_NEXT_PAGE_NO));
}

/*******************************************************************//**
Deallocate a buffer block that was reserved for a BLOB part. */
static
void
btr_blob_free(
/*==========*/
	buf_block_t*	block,	/*!< in: buffer block */
	ibool		all,	/*!< in: TRUE=remove also the compressed page
				if there is one */
	mtr_t*		mtr)	/*!< in: mini-transaction to commit */
{
	buf_pool_t*	buf_pool = buf_pool_from_block(block);
	ulint		space	= buf_block_get_space(block);
	ulint		page_no	= buf_block_get_page_no(block);
	bool		freed	= false;

	ut_ad(mtr_memo_contains(mtr, block, MTR_MEMO_PAGE_X_FIX));

	mtr_commit(mtr);

	mutex_enter(&buf_pool->LRU_list_mutex);
	mutex_enter(&block->mutex);

	/* Only free the block if it is still allocated to
	the same file page. */

	if (buf_block_get_state(block)
	    == BUF_BLOCK_FILE_PAGE
	    && buf_block_get_space(block) == space
	    && buf_block_get_page_no(block) == page_no) {

		freed = buf_LRU_free_page(&block->page, all);

		if (!freed && all && block->page.zip.data
		    /* Now, buf_LRU_free_page() may release mutexes
		    temporarily */
		    && buf_block_get_state(block) == BUF_BLOCK_FILE_PAGE
		    && buf_block_get_space(block) == space
		    && buf_block_get_page_no(block) == page_no) {

			/* Attempt to deallocate the uncompressed page
			if the whole block cannot be deallocted. */
			freed = buf_LRU_free_page(&block->page, false);
		}
	}

	if (!freed) {
		mutex_exit(&buf_pool->LRU_list_mutex);
	}

	mutex_exit(&block->mutex);
}

/*******************************************************************//**
Stores the fields in big_rec_vec to the tablespace and puts pointers to
them in rec.  The extern flags in rec will have to be set beforehand.
The fields are stored on pages allocated from leaf node
file segment of the index tree.
@return	DB_SUCCESS or DB_OUT_OF_FILE_SPACE or DB_TOO_BIG_FOR_REDO */
UNIV_INTERN
dberr_t
btr_store_big_rec_extern_fields(
/*============================*/
	dict_index_t*	index,		/*!< in: index of rec; the index tree
					MUST be X-latched */
	buf_block_t*	rec_block,	/*!< in/out: block containing rec */
	rec_t*		rec,		/*!< in/out: record */
	const ulint*	offsets,	/*!< in: rec_get_offsets(rec, index);
					the "external storage" flags in offsets
					will not correspond to rec when
					this function returns */
	const big_rec_t*big_rec_vec,	/*!< in: vector containing fields
					to be stored externally */
	mtr_t*		btr_mtr,	/*!< in: mtr containing the
					latches to the clustered index */
	enum blob_op	op)		/*! in: operation code */
{
	ulint		rec_page_no;
	byte*		field_ref;
	ulint		extern_len;
	ulint		store_len;
	ulint		page_no;
	ulint		space_id;
	ulint		zip_size;
	ulint		prev_page_no;
	ulint		hint_page_no;
	ulint		i;
	mtr_t		mtr;
	mtr_t*		alloc_mtr;
	mem_heap_t*	heap = NULL;
	page_zip_des_t*	page_zip;
	z_stream	c_stream;
	buf_block_t**	freed_pages	= NULL;
	ulint		n_freed_pages	= 0;
	dberr_t		error		= DB_SUCCESS;

	ut_ad(rec_offs_validate(rec, index, offsets));
	ut_ad(rec_offs_any_extern(offsets));
	ut_ad(btr_mtr);
	ut_ad(mtr_memo_contains(btr_mtr, dict_index_get_lock(index),
				MTR_MEMO_X_LOCK));
	ut_ad(mtr_memo_contains(btr_mtr, rec_block, MTR_MEMO_PAGE_X_FIX));
	ut_ad(buf_block_get_frame(rec_block) == page_align(rec));
	ut_a(dict_index_is_clust(index));

	page_zip = buf_block_get_page_zip(rec_block);
	ut_a(dict_table_zip_size(index->table)
	     == buf_block_get_zip_size(rec_block));

	space_id = buf_block_get_space(rec_block);
	zip_size = buf_block_get_zip_size(rec_block);
	rec_page_no = buf_block_get_page_no(rec_block);
	ut_a(fil_page_get_type(page_align(rec)) == FIL_PAGE_INDEX);

	error = btr_check_blob_limit(big_rec_vec);

	if (error != DB_SUCCESS) {
		ut_ad(op == BTR_STORE_INSERT);
		return(error);
	}

	if (page_zip) {
		int	err;

		/* Zlib deflate needs 128 kilobytes for the default
		window size, plus 512 << memLevel, plus a few
		kilobytes for small objects.  We use reduced memLevel
		to limit the memory consumption, and preallocate the
		heap, hoping to avoid memory fragmentation. */
		heap = mem_heap_create(250000);
		page_zip_set_alloc(&c_stream, heap);

		err = deflateInit2(&c_stream, page_zip_level,
				   Z_DEFLATED, 15, 7, Z_DEFAULT_STRATEGY);
		ut_a(err == Z_OK);
	}

	if (btr_blob_op_is_update(op)) {
		/* Avoid reusing pages that have been previously freed
		in btr_mtr. */
		if (btr_mtr->n_freed_pages) {
			if (heap == NULL) {
				heap = mem_heap_create(
					btr_mtr->n_freed_pages
					* sizeof *freed_pages);
			}

			freed_pages = static_cast<buf_block_t**>(
				mem_heap_alloc(
					heap,
					btr_mtr->n_freed_pages
					* sizeof *freed_pages));
			n_freed_pages = 0;
		}

		/* Because btr_mtr will be committed after mtr, it is
		possible that the tablespace has been extended when
		the B-tree record was updated or inserted, or it will
		be extended while allocating pages for big_rec.

		TODO: In mtr (not btr_mtr), write a redo log record
		about extending the tablespace to its current size,
		and remember the current size. Whenever the tablespace
		grows as pages are allocated, write further redo log
		records to mtr. (Currently tablespace extension is not
		covered by the redo log. If it were, the record would
		only be written to btr_mtr, which is committed after
		mtr.) */
		alloc_mtr = btr_mtr;
	} else {
		/* Use the local mtr for allocations. */
		alloc_mtr = &mtr;
	}

#if defined UNIV_DEBUG || defined UNIV_BLOB_LIGHT_DEBUG
	/* All pointers to externally stored columns in the record
	must either be zero or they must be pointers to inherited
	columns, owned by this record or an earlier record version. */
	for (i = 0; i < rec_offs_n_fields(offsets); i++) {
		if (!rec_offs_nth_extern(offsets, i)) {
			continue;
		}
		field_ref = btr_rec_get_field_ref(rec, offsets, i);

		ut_a(!(field_ref[BTR_EXTERN_LEN] & BTR_EXTERN_OWNER_FLAG));
		/* Either this must be an update in place,
		or the BLOB must be inherited, or the BLOB pointer
		must be zero (will be written in this function). */
		ut_a(op == BTR_STORE_UPDATE
		     || (field_ref[BTR_EXTERN_LEN] & BTR_EXTERN_INHERITED_FLAG)
		     || !memcmp(field_ref, field_ref_zero,
				BTR_EXTERN_FIELD_REF_SIZE));
	}
#endif /* UNIV_DEBUG || UNIV_BLOB_LIGHT_DEBUG */
	/* We have to create a file segment to the tablespace
	for each field and put the pointer to the field in rec */

	for (i = 0; i < big_rec_vec->n_fields; i++) {
		field_ref = btr_rec_get_field_ref(
			rec, offsets, big_rec_vec->fields[i].field_no);
#if defined UNIV_DEBUG || defined UNIV_BLOB_LIGHT_DEBUG
		/* A zero BLOB pointer should have been initially inserted. */
		ut_a(!memcmp(field_ref, field_ref_zero,
			     BTR_EXTERN_FIELD_REF_SIZE));
#endif /* UNIV_DEBUG || UNIV_BLOB_LIGHT_DEBUG */
		extern_len = big_rec_vec->fields[i].len;
		UNIV_MEM_ASSERT_RW(big_rec_vec->fields[i].data,
				   extern_len);

		ut_a(extern_len > 0);

		prev_page_no = FIL_NULL;

		if (page_zip) {
			int	err = deflateReset(&c_stream);
			ut_a(err == Z_OK);

			c_stream.next_in = (Bytef*)
				big_rec_vec->fields[i].data;
			c_stream.avail_in = static_cast<uInt>(extern_len);
		}

		for (;;) {
			buf_block_t*	block;
			page_t*		page;

			mtr_start(&mtr);

			if (prev_page_no == FIL_NULL) {
				hint_page_no = 1 + rec_page_no;
			} else {
				hint_page_no = prev_page_no + 1;
			}

alloc_another:
			block = btr_page_alloc(index, hint_page_no,
					       FSP_NO_DIR, 0, alloc_mtr, &mtr);
			if (UNIV_UNLIKELY(block == NULL)) {
				mtr_commit(&mtr);
				error = DB_OUT_OF_FILE_SPACE;
				goto func_exit;
			}

			if (rw_lock_get_x_lock_count(&block->lock) > 1) {
				/* This page must have been freed in
				btr_mtr previously. Put it aside, and
				allocate another page for the BLOB data. */
				ut_ad(alloc_mtr == btr_mtr);
				ut_ad(btr_blob_op_is_update(op));
				ut_ad(n_freed_pages < btr_mtr->n_freed_pages);
				freed_pages[n_freed_pages++] = block;
				goto alloc_another;
			}

			page_no = buf_block_get_page_no(block);
			page = buf_block_get_frame(block);

			if (prev_page_no != FIL_NULL) {
				buf_block_t*	prev_block;
				page_t*		prev_page;

				prev_block = buf_page_get(space_id, zip_size,
							  prev_page_no,
							  RW_X_LATCH, &mtr);
				buf_block_dbg_add_level(prev_block,
							SYNC_EXTERN_STORAGE);
				prev_page = buf_block_get_frame(prev_block);

				if (page_zip) {
					mlog_write_ulint(
						prev_page + FIL_PAGE_NEXT,
						page_no, MLOG_4BYTES, &mtr);
					memcpy(buf_block_get_page_zip(
						       prev_block)
					       ->data + FIL_PAGE_NEXT,
					       prev_page + FIL_PAGE_NEXT, 4);
				} else {
					mlog_write_ulint(
						prev_page + FIL_PAGE_DATA
						+ BTR_BLOB_HDR_NEXT_PAGE_NO,
						page_no, MLOG_4BYTES, &mtr);
				}

			} else if (dict_index_is_online_ddl(index)) {
				row_log_table_blob_alloc(index, page_no);
			}

			if (page_zip) {
				int		err;
				page_zip_des_t*	blob_page_zip;

				/* Write FIL_PAGE_TYPE to the redo log
				separately, before logging any other
				changes to the page, so that the debug
				assertions in
				recv_parse_or_apply_log_rec_body() can
				be made simpler.  Before InnoDB Plugin
				1.0.4, the initialization of
				FIL_PAGE_TYPE was logged as part of
				the mlog_log_string() below. */

				mlog_write_ulint(page + FIL_PAGE_TYPE,
						 prev_page_no == FIL_NULL
						 ? FIL_PAGE_TYPE_ZBLOB
						 : FIL_PAGE_TYPE_ZBLOB2,
						 MLOG_2BYTES, &mtr);

				c_stream.next_out = page
					+ FIL_PAGE_DATA;
				c_stream.avail_out
					= static_cast<uInt>(page_zip_get_size(page_zip))
					- FIL_PAGE_DATA;

				err = deflate(&c_stream, Z_FINISH);
				ut_a(err == Z_OK || err == Z_STREAM_END);
				ut_a(err == Z_STREAM_END
				     || c_stream.avail_out == 0);

				/* Write the "next BLOB page" pointer */
				mlog_write_ulint(page + FIL_PAGE_NEXT,
						 FIL_NULL, MLOG_4BYTES, &mtr);
				/* Initialize the unused "prev page" pointer */
				mlog_write_ulint(page + FIL_PAGE_PREV,
						 FIL_NULL, MLOG_4BYTES, &mtr);
				/* Write a back pointer to the record
				into the otherwise unused area.  This
				information could be useful in
				debugging.  Later, we might want to
				implement the possibility to relocate
				BLOB pages.  Then, we would need to be
				able to adjust the BLOB pointer in the
				record.  We do not store the heap
				number of the record, because it can
				change in page_zip_reorganize() or
				btr_page_reorganize().  However, also
				the page number of the record may
				change when B-tree nodes are split or
				merged. */
				mlog_write_ulint(page
						 + FIL_PAGE_FILE_FLUSH_LSN_OR_KEY_VERSION,
						 space_id,
						 MLOG_4BYTES, &mtr);
				mlog_write_ulint(page
						 + FIL_PAGE_FILE_FLUSH_LSN_OR_KEY_VERSION + 4,
						 rec_page_no,
						 MLOG_4BYTES, &mtr);

				/* Zero out the unused part of the page. */
				memset(page + page_zip_get_size(page_zip)
				       - c_stream.avail_out,
				       0, c_stream.avail_out);
				mlog_log_string(page
						+ FIL_PAGE_FILE_FLUSH_LSN_OR_KEY_VERSION,
						page_zip_get_size(page_zip)
						- FIL_PAGE_FILE_FLUSH_LSN_OR_KEY_VERSION,
						&mtr);
				/* Copy the page to compressed storage,
				because it will be flushed to disk
				from there. */
				blob_page_zip = buf_block_get_page_zip(block);
				ut_ad(blob_page_zip);
				ut_ad(page_zip_get_size(blob_page_zip)
				      == page_zip_get_size(page_zip));
				memcpy(blob_page_zip->data, page,
				       page_zip_get_size(page_zip));

				if (err == Z_OK && prev_page_no != FIL_NULL) {

					goto next_zip_page;
				}

				if (alloc_mtr == &mtr) {
					rec_block = buf_page_get(
						space_id, zip_size,
						rec_page_no,
						RW_X_LATCH, &mtr);
					buf_block_dbg_add_level(
						rec_block,
						SYNC_NO_ORDER_CHECK);
				}

				if (err == Z_STREAM_END) {
					mach_write_to_4(field_ref
							+ BTR_EXTERN_LEN, 0);
					mach_write_to_4(field_ref
							+ BTR_EXTERN_LEN + 4,
							c_stream.total_in);
				} else {
					memset(field_ref + BTR_EXTERN_LEN,
					       0, 8);
				}

				if (prev_page_no == FIL_NULL) {
					btr_blob_dbg_add_blob(
						rec, big_rec_vec->fields[i]
						.field_no, page_no, index,
						"store");

					mach_write_to_4(field_ref
							+ BTR_EXTERN_SPACE_ID,
							space_id);

					mach_write_to_4(field_ref
							+ BTR_EXTERN_PAGE_NO,
							page_no);

					mach_write_to_4(field_ref
							+ BTR_EXTERN_OFFSET,
							FIL_PAGE_NEXT);
				}

				page_zip_write_blob_ptr(
					page_zip, rec, index, offsets,
					big_rec_vec->fields[i].field_no,
					alloc_mtr);

next_zip_page:
				prev_page_no = page_no;

				/* Commit mtr and release the
				uncompressed page frame to save memory. */
				btr_blob_free(block, FALSE, &mtr);

				if (err == Z_STREAM_END) {
					break;
				}
			} else {
				mlog_write_ulint(page + FIL_PAGE_TYPE,
						 FIL_PAGE_TYPE_BLOB,
						 MLOG_2BYTES, &mtr);

				if (extern_len > (UNIV_PAGE_SIZE
						  - FIL_PAGE_DATA
						  - BTR_BLOB_HDR_SIZE
						  - FIL_PAGE_DATA_END)) {
					store_len = UNIV_PAGE_SIZE
						- FIL_PAGE_DATA
						- BTR_BLOB_HDR_SIZE
						- FIL_PAGE_DATA_END;
				} else {
					store_len = extern_len;
				}

				mlog_write_string(page + FIL_PAGE_DATA
						  + BTR_BLOB_HDR_SIZE,
						  (const byte*)
						  big_rec_vec->fields[i].data
						  + big_rec_vec->fields[i].len
						  - extern_len,
						  store_len, &mtr);
				mlog_write_ulint(page + FIL_PAGE_DATA
						 + BTR_BLOB_HDR_PART_LEN,
						 store_len, MLOG_4BYTES, &mtr);
				mlog_write_ulint(page + FIL_PAGE_DATA
						 + BTR_BLOB_HDR_NEXT_PAGE_NO,
						 FIL_NULL, MLOG_4BYTES, &mtr);

				extern_len -= store_len;

				if (alloc_mtr == &mtr) {
					rec_block = buf_page_get(
						space_id, zip_size,
						rec_page_no,
						RW_X_LATCH, &mtr);
					buf_block_dbg_add_level(
						rec_block,
						SYNC_NO_ORDER_CHECK);
				}

				mlog_write_ulint(field_ref + BTR_EXTERN_LEN, 0,
						 MLOG_4BYTES, alloc_mtr);
				mlog_write_ulint(field_ref
						 + BTR_EXTERN_LEN + 4,
						 big_rec_vec->fields[i].len
						 - extern_len,
						 MLOG_4BYTES, alloc_mtr);

				if (prev_page_no == FIL_NULL) {
					btr_blob_dbg_add_blob(
						rec, big_rec_vec->fields[i]
						.field_no, page_no, index,
						"store");

					mlog_write_ulint(field_ref
							 + BTR_EXTERN_SPACE_ID,
							 space_id, MLOG_4BYTES,
							 alloc_mtr);

					mlog_write_ulint(field_ref
							 + BTR_EXTERN_PAGE_NO,
							 page_no, MLOG_4BYTES,
							 alloc_mtr);

					mlog_write_ulint(field_ref
							 + BTR_EXTERN_OFFSET,
							 FIL_PAGE_DATA,
							 MLOG_4BYTES,
							 alloc_mtr);
				}

				prev_page_no = page_no;

				mtr_commit(&mtr);

				if (extern_len == 0) {
					break;
				}
			}
		}

		DBUG_EXECUTE_IF("btr_store_big_rec_extern",
				error = DB_OUT_OF_FILE_SPACE;
				goto func_exit;);
	}

func_exit:
	if (page_zip) {
		deflateEnd(&c_stream);
	}

	if (n_freed_pages) {
		ulint	i;

		ut_ad(alloc_mtr == btr_mtr);
		ut_ad(btr_blob_op_is_update(op));

		for (i = 0; i < n_freed_pages; i++) {
			btr_page_free_low(index, freed_pages[i], 0, true, alloc_mtr);
		}
	}

	if (heap != NULL) {
		mem_heap_free(heap);
	}

#if defined UNIV_DEBUG || defined UNIV_BLOB_LIGHT_DEBUG
	/* All pointers to externally stored columns in the record
	must be valid. */
	for (i = 0; i < rec_offs_n_fields(offsets); i++) {
		if (!rec_offs_nth_extern(offsets, i)) {
			continue;
		}

		field_ref = btr_rec_get_field_ref(rec, offsets, i);

		/* The pointer must not be zero if the operation
		succeeded. */
		ut_a(0 != memcmp(field_ref, field_ref_zero,
				 BTR_EXTERN_FIELD_REF_SIZE)
		     || error != DB_SUCCESS);
		/* The column must not be disowned by this record. */
		ut_a(!(field_ref[BTR_EXTERN_LEN] & BTR_EXTERN_OWNER_FLAG));
	}
#endif /* UNIV_DEBUG || UNIV_BLOB_LIGHT_DEBUG */
	return(error);
}

/*******************************************************************//**
Check the FIL_PAGE_TYPE on an uncompressed BLOB page. */
static
void
btr_check_blob_fil_page_type(
/*=========================*/
	ulint		space_id,	/*!< in: space id */
	ulint		page_no,	/*!< in: page number */
	const page_t*	page,		/*!< in: page */
	ibool		read)		/*!< in: TRUE=read, FALSE=purge */
{
	ulint	type = fil_page_get_type(page);

	ut_a(space_id == page_get_space_id(page));
	ut_a(page_no == page_get_page_no(page));

	if (UNIV_UNLIKELY(type != FIL_PAGE_TYPE_BLOB)) {
		ulint	flags = fil_space_get_flags(space_id);

#ifndef UNIV_DEBUG /* Improve debug test coverage */
		if (dict_tf_get_format(flags) == UNIV_FORMAT_A) {
			/* Old versions of InnoDB did not initialize
			FIL_PAGE_TYPE on BLOB pages.  Do not print
			anything about the type mismatch when reading
			a BLOB page that is in Antelope format.*/
			return;
		}
#endif /* !UNIV_DEBUG */

		ut_print_timestamp(stderr);
		fprintf(stderr,
			"  InnoDB: FIL_PAGE_TYPE=%lu"
			" on BLOB %s space %lu page %lu flags %lx\n",
			(ulong) type, read ? "read" : "purge",
			(ulong) space_id, (ulong) page_no, (ulong) flags);
		ut_error;
	}
}

/*******************************************************************//**
Frees the space in an externally stored field to the file space
management if the field in data is owned by the externally stored field,
in a rollback we may have the additional condition that the field must
not be inherited. */
UNIV_INTERN
void
btr_free_externally_stored_field(
/*=============================*/
	dict_index_t*	index,		/*!< in: index of the data, the index
					tree MUST be X-latched; if the tree
					height is 1, then also the root page
					must be X-latched! (this is relevant
					in the case this function is called
					from purge where 'data' is located on
					an undo log page, not an index
					page) */
	byte*		field_ref,	/*!< in/out: field reference */
	const rec_t*	rec,		/*!< in: record containing field_ref, for
					page_zip_write_blob_ptr(), or NULL */
	const ulint*	offsets,	/*!< in: rec_get_offsets(rec, index),
					or NULL */
	page_zip_des_t*	page_zip,	/*!< in: compressed page corresponding
					to rec, or NULL if rec == NULL */
	ulint		i,		/*!< in: field number of field_ref;
					ignored if rec == NULL */
	enum trx_rb_ctx	rb_ctx,		/*!< in: rollback context */
	mtr_t*		local_mtr __attribute__((unused))) /*!< in: mtr
					containing the latch to data an an
					X-latch to the index tree */
{
	page_t*		page;
	const ulint	space_id	= mach_read_from_4(
		field_ref + BTR_EXTERN_SPACE_ID);
	const ulint	start_page	= mach_read_from_4(
		field_ref + BTR_EXTERN_PAGE_NO);
	ulint		rec_zip_size = dict_table_zip_size(index->table);
	ulint		ext_zip_size;
	ulint		page_no;
	ulint		next_page_no;
	mtr_t		mtr;

	ut_ad(dict_index_is_clust(index));
	ut_ad(mtr_memo_contains(local_mtr, dict_index_get_lock(index),
				MTR_MEMO_X_LOCK));
	ut_ad(mtr_memo_contains_page(local_mtr, field_ref,
				     MTR_MEMO_PAGE_X_FIX));
	ut_ad(!rec || rec_offs_validate(rec, index, offsets));
	ut_ad(!rec || field_ref == btr_rec_get_field_ref(rec, offsets, i));

	if (UNIV_UNLIKELY(!memcmp(field_ref, field_ref_zero,
				  BTR_EXTERN_FIELD_REF_SIZE))) {
		/* In the rollback, we may encounter a clustered index
		record with some unwritten off-page columns. There is
		nothing to free then. */
		ut_a(rb_ctx != RB_NONE);
		return;
	}

	ut_ad(space_id == index->space);

	if (UNIV_UNLIKELY(space_id != dict_index_get_space(index))) {
		ext_zip_size = fil_space_get_zip_size(space_id);
		/* This must be an undo log record in the system tablespace,
		that is, in row_purge_upd_exist_or_extern().
		Currently, externally stored records are stored in the
		same tablespace as the referring records. */
		ut_ad(!page_get_space_id(page_align(field_ref)));
		ut_ad(!rec);
		ut_ad(!page_zip);
	} else {
		ext_zip_size = rec_zip_size;
	}

	if (!rec) {
		/* This is a call from row_purge_upd_exist_or_extern(). */
		ut_ad(!page_zip);
		rec_zip_size = 0;
	}

#ifdef UNIV_BLOB_DEBUG
	if (!(field_ref[BTR_EXTERN_LEN] & BTR_EXTERN_OWNER_FLAG)
	    && !((field_ref[BTR_EXTERN_LEN] & BTR_EXTERN_INHERITED_FLAG)
		 && (rb_ctx == RB_NORMAL || rb_ctx == RB_RECOVERY))) {
		/* This off-page column will be freed.
		Check that no references remain. */

		btr_blob_dbg_t	b;

		b.blob_page_no = start_page;

		if (rec) {
			/* Remove the reference from the record to the
			BLOB. If the BLOB were not freed, the
			reference would be removed when the record is
			removed. Freeing the BLOB will overwrite the
			BTR_EXTERN_PAGE_NO in the field_ref of the
			record with FIL_NULL, which would make the
			btr_blob_dbg information inconsistent with the
			record. */
			b.ref_page_no = page_get_page_no(page_align(rec));
			b.ref_heap_no = page_rec_get_heap_no(rec);
			b.ref_field_no = i;
			btr_blob_dbg_rbt_delete(index, &b, "free");
		}

		btr_blob_dbg_assert_empty(index, b.blob_page_no);
	}
#endif /* UNIV_BLOB_DEBUG */

	for (;;) {
#ifdef UNIV_SYNC_DEBUG
		buf_block_t*	rec_block;
#endif /* UNIV_SYNC_DEBUG */
		buf_block_t*	ext_block;

		mtr_start(&mtr);

#ifdef UNIV_SYNC_DEBUG
		rec_block =
#endif /* UNIV_SYNC_DEBUG */
		buf_page_get(page_get_space_id(page_align(field_ref)),
			     rec_zip_size,
			     page_get_page_no(page_align(field_ref)),
			     RW_X_LATCH, &mtr);
		buf_block_dbg_add_level(rec_block, SYNC_NO_ORDER_CHECK);
		page_no = mach_read_from_4(field_ref + BTR_EXTERN_PAGE_NO);

		if (/* There is no external storage data */
		    page_no == FIL_NULL
		    /* This field does not own the externally stored field */
		    || (mach_read_from_1(field_ref + BTR_EXTERN_LEN)
			& BTR_EXTERN_OWNER_FLAG)
		    /* Rollback and inherited field */
		    || ((rb_ctx == RB_NORMAL || rb_ctx == RB_RECOVERY)
			&& (mach_read_from_1(field_ref + BTR_EXTERN_LEN)
			    & BTR_EXTERN_INHERITED_FLAG))) {

			/* Do not free */
			mtr_commit(&mtr);

			return;
		}

		if (page_no == start_page && dict_index_is_online_ddl(index)) {
			row_log_table_blob_free(index, start_page);
		}

		ext_block = buf_page_get(space_id, ext_zip_size, page_no,
					 RW_X_LATCH, &mtr);
		buf_block_dbg_add_level(ext_block, SYNC_EXTERN_STORAGE);
		page = buf_block_get_frame(ext_block);

		if (ext_zip_size) {
			/* Note that page_zip will be NULL
			in row_purge_upd_exist_or_extern(). */
			switch (fil_page_get_type(page)) {
			case FIL_PAGE_TYPE_ZBLOB:
			case FIL_PAGE_TYPE_ZBLOB2:
				break;
			default:
				ut_error;
			}
			next_page_no = mach_read_from_4(page + FIL_PAGE_NEXT);

			btr_page_free_low(index, ext_block, 0, true, &mtr);

			if (page_zip != NULL) {
				mach_write_to_4(field_ref + BTR_EXTERN_PAGE_NO,
						next_page_no);
				mach_write_to_4(field_ref + BTR_EXTERN_LEN + 4,
						0);
				page_zip_write_blob_ptr(page_zip, rec, index,
							offsets, i, &mtr);
			} else {
				mlog_write_ulint(field_ref
						 + BTR_EXTERN_PAGE_NO,
						 next_page_no,
						 MLOG_4BYTES, &mtr);
				mlog_write_ulint(field_ref
						 + BTR_EXTERN_LEN + 4, 0,
						 MLOG_4BYTES, &mtr);
			}
		} else {
			ut_a(!page_zip);
			btr_check_blob_fil_page_type(space_id, page_no, page,
						     FALSE);

			next_page_no = mach_read_from_4(
				page + FIL_PAGE_DATA
				+ BTR_BLOB_HDR_NEXT_PAGE_NO);

			/* We must supply the page level (= 0) as an argument
			because we did not store it on the page (we save the
			space overhead from an index page header. */

			btr_page_free_low(index, ext_block, 0, true, &mtr);

			mlog_write_ulint(field_ref + BTR_EXTERN_PAGE_NO,
					 next_page_no,
					 MLOG_4BYTES, &mtr);
			/* Zero out the BLOB length.  If the server
			crashes during the execution of this function,
			trx_rollback_or_clean_all_recovered() could
			dereference the half-deleted BLOB, fetching a
			wrong prefix for the BLOB. */
			mlog_write_ulint(field_ref + BTR_EXTERN_LEN + 4,
					 0,
					 MLOG_4BYTES, &mtr);
		}

		/* Commit mtr and release the BLOB block to save memory. */
		btr_blob_free(ext_block, TRUE, &mtr);
	}
}

/***********************************************************//**
Frees the externally stored fields for a record. */
static
void
btr_rec_free_externally_stored_fields(
/*==================================*/
	dict_index_t*	index,	/*!< in: index of the data, the index
				tree MUST be X-latched */
	rec_t*		rec,	/*!< in/out: record */
	const ulint*	offsets,/*!< in: rec_get_offsets(rec, index) */
	page_zip_des_t*	page_zip,/*!< in: compressed page whose uncompressed
				part will be updated, or NULL */
	enum trx_rb_ctx	rb_ctx,	/*!< in: rollback context */
	mtr_t*		mtr)	/*!< in: mini-transaction handle which contains
				an X-latch to record page and to the index
				tree */
{
	ulint	n_fields;
	ulint	i;

	ut_ad(rec_offs_validate(rec, index, offsets));
	ut_ad(mtr_memo_contains_page(mtr, rec, MTR_MEMO_PAGE_X_FIX));
	/* Free possible externally stored fields in the record */

	ut_ad(dict_table_is_comp(index->table) == !!rec_offs_comp(offsets));
	n_fields = rec_offs_n_fields(offsets);

	for (i = 0; i < n_fields; i++) {
		if (rec_offs_nth_extern(offsets, i)) {
			btr_free_externally_stored_field(
				index, btr_rec_get_field_ref(rec, offsets, i),
				rec, offsets, page_zip, i, rb_ctx, mtr);
		}
	}
}

/***********************************************************//**
Frees the externally stored fields for a record, if the field is mentioned
in the update vector. */
static
void
btr_rec_free_updated_extern_fields(
/*===============================*/
	dict_index_t*	index,	/*!< in: index of rec; the index tree MUST be
				X-latched */
	rec_t*		rec,	/*!< in/out: record */
	page_zip_des_t*	page_zip,/*!< in: compressed page whose uncompressed
				part will be updated, or NULL */
	const ulint*	offsets,/*!< in: rec_get_offsets(rec, index) */
	const upd_t*	update,	/*!< in: update vector */
	enum trx_rb_ctx	rb_ctx,	/*!< in: rollback context */
	mtr_t*		mtr)	/*!< in: mini-transaction handle which contains
				an X-latch to record page and to the tree */
{
	ulint	n_fields;
	ulint	i;

	ut_ad(rec_offs_validate(rec, index, offsets));
	ut_ad(mtr_memo_contains_page(mtr, rec, MTR_MEMO_PAGE_X_FIX));

	/* Free possible externally stored fields in the record */

	n_fields = upd_get_n_fields(update);

	for (i = 0; i < n_fields; i++) {
		const upd_field_t* ufield = upd_get_nth_field(update, i);

		if (rec_offs_nth_extern(offsets, ufield->field_no)) {
			ulint	len;
			byte*	data = rec_get_nth_field(
				rec, offsets, ufield->field_no, &len);
			ut_a(len >= BTR_EXTERN_FIELD_REF_SIZE);

			btr_free_externally_stored_field(
				index, data + len - BTR_EXTERN_FIELD_REF_SIZE,
				rec, offsets, page_zip,
				ufield->field_no, rb_ctx, mtr);
		}
	}
}

/*******************************************************************//**
Copies the prefix of an uncompressed BLOB.  The clustered index record
that points to this BLOB must be protected by a lock or a page latch.
@return	number of bytes written to buf */
static
ulint
btr_copy_blob_prefix(
/*=================*/
	byte*		buf,	/*!< out: the externally stored part of
				the field, or a prefix of it */
	ulint		len,	/*!< in: length of buf, in bytes */
	ulint		space_id,/*!< in: space id of the BLOB pages */
	ulint		page_no,/*!< in: page number of the first BLOB page */
	ulint		offset,	/*!< in: offset on the first BLOB page */
	trx_t*		trx)	/*!< in: transaction handle */
{
	ulint	copied_len	= 0;

	for (;;) {
		mtr_t		mtr;
		buf_block_t*	block;
		const page_t*	page;
		const byte*	blob_header;
		ulint		part_len;
		ulint		copy_len;

		mtr_start_trx(&mtr, trx);

		block = buf_page_get(space_id, 0, page_no, RW_S_LATCH, &mtr);
		buf_block_dbg_add_level(block, SYNC_EXTERN_STORAGE);
		page = buf_block_get_frame(block);

		btr_check_blob_fil_page_type(space_id, page_no, page, TRUE);

		blob_header = page + offset;
		part_len = btr_blob_get_part_len(blob_header);
		copy_len = ut_min(part_len, len - copied_len);

		memcpy(buf + copied_len,
		       blob_header + BTR_BLOB_HDR_SIZE, copy_len);
		copied_len += copy_len;

		page_no = btr_blob_get_next_page_no(blob_header);

		mtr_commit(&mtr);

		if (page_no == FIL_NULL || copy_len != part_len) {
			UNIV_MEM_ASSERT_RW(buf, copied_len);
			return(copied_len);
		}

		/* On other BLOB pages except the first the BLOB header
		always is at the page data start: */

		offset = FIL_PAGE_DATA;

		ut_ad(copied_len <= len);
	}
}

/*******************************************************************//**
Copies the prefix of a compressed BLOB.  The clustered index record
that points to this BLOB must be protected by a lock or a page latch.
@return	number of bytes written to buf */
static
ulint
btr_copy_zblob_prefix(
/*==================*/
	byte*		buf,	/*!< out: the externally stored part of
				the field, or a prefix of it */
	ulint		len,	/*!< in: length of buf, in bytes */
	ulint		zip_size,/*!< in: compressed BLOB page size */
	ulint		space_id,/*!< in: space id of the BLOB pages */
	ulint		page_no,/*!< in: page number of the first BLOB page */
	ulint		offset)	/*!< in: offset on the first BLOB page */
{
	ulint		page_type = FIL_PAGE_TYPE_ZBLOB;
	mem_heap_t*	heap;
	int		err;
	z_stream	d_stream;

	d_stream.next_out = buf;
	d_stream.avail_out = static_cast<uInt>(len);
	d_stream.next_in = Z_NULL;
	d_stream.avail_in = 0;

	/* Zlib inflate needs 32 kilobytes for the default
	window size, plus a few kilobytes for small objects. */
	heap = mem_heap_create(40000);
	page_zip_set_alloc(&d_stream, heap);

	ut_ad(ut_is_2pow(zip_size));
	ut_ad(zip_size >= UNIV_ZIP_SIZE_MIN);
	ut_ad(zip_size <= UNIV_ZIP_SIZE_MAX);
	ut_ad(space_id);

	err = inflateInit(&d_stream);
	ut_a(err == Z_OK);

	for (;;) {
		buf_page_t*	bpage;
		ulint		next_page_no;

		/* There is no latch on bpage directly.  Instead,
		bpage is protected by the B-tree page latch that
		is being held on the clustered index record, or,
		in row_merge_copy_blobs(), by an exclusive table lock. */
		bpage = buf_page_get_zip(space_id, zip_size, page_no);

		if (UNIV_UNLIKELY(!bpage)) {
			ut_print_timestamp(stderr);
			fprintf(stderr,
				"  InnoDB: Cannot load"
				" compressed BLOB"
				" page %lu space %lu\n",
				(ulong) page_no, (ulong) space_id);
			goto func_exit;
		}

		if (UNIV_UNLIKELY
		    (fil_page_get_type(bpage->zip.data) != page_type)) {
			ut_print_timestamp(stderr);
			fprintf(stderr,
				"  InnoDB: Unexpected type %lu of"
				" compressed BLOB"
				" page %lu space %lu\n",
				(ulong) fil_page_get_type(bpage->zip.data),
				(ulong) page_no, (ulong) space_id);
			ut_ad(0);
			goto end_of_blob;
		}

		next_page_no = mach_read_from_4(bpage->zip.data + offset);

		if (UNIV_LIKELY(offset == FIL_PAGE_NEXT)) {
			/* When the BLOB begins at page header,
			the compressed data payload does not
			immediately follow the next page pointer. */
			offset = FIL_PAGE_DATA;
		} else {
			offset += 4;
		}

		d_stream.next_in = bpage->zip.data + offset;
		d_stream.avail_in = static_cast<uInt>(zip_size - offset);

		err = inflate(&d_stream, Z_NO_FLUSH);
		switch (err) {
		case Z_OK:
			if (!d_stream.avail_out) {
				goto end_of_blob;
			}
			break;
		case Z_STREAM_END:
			if (next_page_no == FIL_NULL) {
				goto end_of_blob;
			}
			/* fall through */
		default:
inflate_error:
			ut_print_timestamp(stderr);
			fprintf(stderr,
				"  InnoDB: inflate() of"
				" compressed BLOB"
				" page %lu space %lu returned %d (%s)\n",
				(ulong) page_no, (ulong) space_id,
				err, d_stream.msg);
		case Z_BUF_ERROR:
			goto end_of_blob;
		}

		if (next_page_no == FIL_NULL) {
			if (!d_stream.avail_in) {
				ut_print_timestamp(stderr);
				fprintf(stderr,
					"  InnoDB: unexpected end of"
					" compressed BLOB"
					" page %lu space %lu\n",
					(ulong) page_no,
					(ulong) space_id);
			} else {
				err = inflate(&d_stream, Z_FINISH);
				switch (err) {
				case Z_STREAM_END:
				case Z_BUF_ERROR:
					break;
				default:
					goto inflate_error;
				}
			}

end_of_blob:
			buf_page_release_zip(bpage);
			goto func_exit;
		}

		buf_page_release_zip(bpage);

		/* On other BLOB pages except the first
		the BLOB header always is at the page header: */

		page_no = next_page_no;
		offset = FIL_PAGE_NEXT;
		page_type = FIL_PAGE_TYPE_ZBLOB2;
	}

func_exit:
	inflateEnd(&d_stream);
	mem_heap_free(heap);
	UNIV_MEM_ASSERT_RW(buf, d_stream.total_out);
	return(d_stream.total_out);
}

/*******************************************************************//**
Copies the prefix of an externally stored field of a record.  The
clustered index record that points to this BLOB must be protected by a
lock or a page latch.
@return	number of bytes written to buf */
static
ulint
btr_copy_externally_stored_field_prefix_low(
/*========================================*/
	byte*		buf,	/*!< out: the externally stored part of
				the field, or a prefix of it */
	ulint		len,	/*!< in: length of buf, in bytes */
	ulint		zip_size,/*!< in: nonzero=compressed BLOB page size,
				zero for uncompressed BLOBs */
	ulint		space_id,/*!< in: space id of the first BLOB page */
	ulint		page_no,/*!< in: page number of the first BLOB page */
	ulint		offset,	/*!< in: offset on the first BLOB page */
	trx_t*		trx)	/*!< in: transaction handle */
{
	if (UNIV_UNLIKELY(len == 0)) {
		return(0);
	}

	if (zip_size) {
		return(btr_copy_zblob_prefix(buf, len, zip_size,
					     space_id, page_no, offset));
	} else {
		return(btr_copy_blob_prefix(buf, len, space_id,
					    page_no, offset, trx));
	}
}

/*******************************************************************//**
Copies the prefix of an externally stored field of a record.  The
clustered index record must be protected by a lock or a page latch.
@return the length of the copied field, or 0 if the column was being
or has been deleted */
UNIV_INTERN
ulint
btr_copy_externally_stored_field_prefix(
/*====================================*/
	byte*		buf,	/*!< out: the field, or a prefix of it */
	ulint		len,	/*!< in: length of buf, in bytes */
	ulint		zip_size,/*!< in: nonzero=compressed BLOB page size,
				zero for uncompressed BLOBs */
	const byte*	data,	/*!< in: 'internally' stored part of the
				field containing also the reference to
				the external part; must be protected by
				a lock or a page latch */
	ulint		local_len,/*!< in: length of data, in bytes */
	trx_t*		trx)	/*!< in: transaction handle */
{
	ulint	space_id;
	ulint	page_no;
	ulint	offset;

	ut_a(local_len >= BTR_EXTERN_FIELD_REF_SIZE);

	local_len -= BTR_EXTERN_FIELD_REF_SIZE;

	if (UNIV_UNLIKELY(local_len >= len)) {
		memcpy(buf, data, len);
		return(len);
	}

	memcpy(buf, data, local_len);
	data += local_len;

	ut_a(memcmp(data, field_ref_zero, BTR_EXTERN_FIELD_REF_SIZE));

	if (!mach_read_from_4(data + BTR_EXTERN_LEN + 4)) {
		/* The externally stored part of the column has been
		(partially) deleted.  Signal the half-deleted BLOB
		to the caller. */

		return(0);
	}

	space_id = mach_read_from_4(data + BTR_EXTERN_SPACE_ID);

	page_no = mach_read_from_4(data + BTR_EXTERN_PAGE_NO);

	offset = mach_read_from_4(data + BTR_EXTERN_OFFSET);

	return(local_len
	       + btr_copy_externally_stored_field_prefix_low(buf + local_len,
							     len - local_len,
							     zip_size,
							     space_id, page_no,
							     offset, trx));
}

/*******************************************************************//**
Copies an externally stored field of a record to mem heap.  The
clustered index record must be protected by a lock or a page latch.
@return	the whole field copied to heap */
UNIV_INTERN
byte*
btr_copy_externally_stored_field(
/*=============================*/
	ulint*		len,	/*!< out: length of the whole field */
	const byte*	data,	/*!< in: 'internally' stored part of the
				field containing also the reference to
				the external part; must be protected by
				a lock or a page latch */
	ulint		zip_size,/*!< in: nonzero=compressed BLOB page size,
				zero for uncompressed BLOBs */
	ulint		local_len,/*!< in: length of data */
	mem_heap_t*	heap,	/*!< in: mem heap */
	trx_t*		trx)	/*!< in: transaction handle */
{
	ulint	space_id;
	ulint	page_no;
	ulint	offset;
	ulint	extern_len;
	byte*	buf;

	ut_a(local_len >= BTR_EXTERN_FIELD_REF_SIZE);

	local_len -= BTR_EXTERN_FIELD_REF_SIZE;

	space_id = mach_read_from_4(data + local_len + BTR_EXTERN_SPACE_ID);

	page_no = mach_read_from_4(data + local_len + BTR_EXTERN_PAGE_NO);

	offset = mach_read_from_4(data + local_len + BTR_EXTERN_OFFSET);

	/* Currently a BLOB cannot be bigger than 4 GB; we
	leave the 4 upper bytes in the length field unused */

	extern_len = mach_read_from_4(data + local_len + BTR_EXTERN_LEN + 4);

	buf = (byte*) mem_heap_alloc(heap, local_len + extern_len);

	memcpy(buf, data, local_len);
	*len = local_len
		+ btr_copy_externally_stored_field_prefix_low(buf + local_len,
							      extern_len,
							      zip_size,
							      space_id,
							      page_no, offset,
							      trx);

	return(buf);
}

/*******************************************************************//**
Copies an externally stored field of a record to mem heap.
@return	the field copied to heap, or NULL if the field is incomplete */
UNIV_INTERN
byte*
btr_rec_copy_externally_stored_field(
/*=================================*/
	const rec_t*	rec,	/*!< in: record in a clustered index;
				must be protected by a lock or a page latch */
	const ulint*	offsets,/*!< in: array returned by rec_get_offsets() */
	ulint		zip_size,/*!< in: nonzero=compressed BLOB page size,
				zero for uncompressed BLOBs */
	ulint		no,	/*!< in: field number */
	ulint*		len,	/*!< out: length of the field */
	mem_heap_t*	heap,	/*!< in: mem heap */
	trx_t*		trx)	/*!< in: transaction handle */
{
	ulint		local_len;
	const byte*	data;

	ut_a(rec_offs_nth_extern(offsets, no));

	/* An externally stored field can contain some initial
	data from the field, and in the last 20 bytes it has the
	space id, page number, and offset where the rest of the
	field data is stored, and the data length in addition to
	the data stored locally. We may need to store some data
	locally to get the local record length above the 128 byte
	limit so that field offsets are stored in two bytes, and
	the extern bit is available in those two bytes. */

	data = rec_get_nth_field(rec, offsets, no, &local_len);

	ut_a(local_len >= BTR_EXTERN_FIELD_REF_SIZE);

	if (UNIV_UNLIKELY
	    (!memcmp(data + local_len - BTR_EXTERN_FIELD_REF_SIZE,
		     field_ref_zero, BTR_EXTERN_FIELD_REF_SIZE))) {
		/* The externally stored field was not written yet.
		This record should only be seen by
		recv_recovery_rollback_active() or any
		TRX_ISO_READ_UNCOMMITTED transactions. */
		return(NULL);
	}

	return(btr_copy_externally_stored_field(len, data,
						zip_size, local_len, heap,
						trx));
}
#endif /* !UNIV_HOTBACKUP */