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
|
/*-
* Copyright (c) 2014-2018 MongoDB, Inc.
* Copyright (c) 2008-2014 WiredTiger, Inc.
* All rights reserved.
*
* See the file LICENSE for redistribution information.
*/
#include "wt_internal.h"
static int __log_newfile(WT_SESSION_IMPL *, bool, bool *);
static int __log_openfile(WT_SESSION_IMPL *, uint32_t, uint32_t, WT_FH **);
static int __log_write_internal(
WT_SESSION_IMPL *, WT_ITEM *, WT_LSN *, uint32_t);
#define WT_LOG_COMPRESS_SKIP (offsetof(WT_LOG_RECORD, record))
#define WT_LOG_ENCRYPT_SKIP (offsetof(WT_LOG_RECORD, record))
/* AUTOMATIC FLAG VALUE GENERATION START */
#define WT_LOG_OPEN_CREATE_OK 0x1u /* Flag to __log_openfile() */
/* AUTOMATIC FLAG VALUE GENERATION STOP */
/*
* __wt_log_printf --
* Write a text message to the log.
*/
int
__wt_log_printf(WT_SESSION_IMPL *session, const char *format, ...)
{
WT_DECL_RET;
va_list ap;
va_start(ap, format);
ret = __wt_log_vprintf(session, format, ap);
va_end(ap);
return (ret);
}
/*
* __log_checksum_match --
* Given a log record, return whether the checksum matches.
*/
static bool
__log_checksum_match(WT_SESSION_IMPL *session, WT_ITEM *buf, uint32_t reclen)
{
WT_LOG_RECORD *logrec;
uint32_t checksum_calculate, checksum_tmp;
WT_UNUSED(session);
logrec = (WT_LOG_RECORD *)buf->mem;
checksum_tmp = logrec->checksum;
logrec->checksum = 0;
checksum_calculate = __wt_checksum(logrec, reclen);
#ifdef WORDS_BIGENDIAN
checksum_calculate = __wt_bswap32(checksum_calculate);
#endif
logrec->checksum = checksum_tmp;
return (logrec->checksum == checksum_calculate);
}
/*
* __log_get_files --
* Retrieve the list of all log-related files of the given prefix type.
*/
static int
__log_get_files(WT_SESSION_IMPL *session,
const char *file_prefix, char ***filesp, u_int *countp)
{
WT_CONNECTION_IMPL *conn;
const char *log_path;
*countp = 0;
*filesp = NULL;
conn = S2C(session);
log_path = conn->log_path;
if (log_path == NULL)
log_path = "";
return (__wt_fs_directory_list(
session, log_path, file_prefix, filesp, countp));
}
/*
* __log_get_files_single --
* Retrieve a single log-related file of the given prefix type.
*/
static int
__log_get_files_single(WT_SESSION_IMPL *session,
const char *file_prefix, char ***filesp, u_int *countp)
{
WT_CONNECTION_IMPL *conn;
const char *log_path;
*countp = 0;
*filesp = NULL;
conn = S2C(session);
log_path = conn->log_path;
if (log_path == NULL)
log_path = "";
return (__wt_fs_directory_list_single(
session, log_path, file_prefix, filesp, countp));
}
/*
* __log_prealloc_remove --
* Remove all previously created pre-allocated files.
*/
static int
__log_prealloc_remove(WT_SESSION_IMPL *session)
{
WT_DECL_RET;
WT_LOG *log;
uint32_t lognum;
u_int i, logcount;
char **logfiles;
logfiles = NULL;
logcount = 0;
log = S2C(session)->log;
__wt_spin_lock(session, &log->log_fs_lock);
/*
* Clean up any old interim pre-allocated files. We clean
* up these files because settings may have changed upon reboot
* and we want those settings to take effect right away.
*/
WT_ERR(__log_get_files(session,
WT_LOG_TMPNAME, &logfiles, &logcount));
for (i = 0; i < logcount; i++) {
WT_ERR(__wt_log_extract_lognum(session, logfiles[i], &lognum));
WT_ERR(__wt_log_remove(session, WT_LOG_TMPNAME, lognum));
}
WT_ERR(__wt_fs_directory_list_free(session, &logfiles, logcount));
WT_ERR(__log_get_files(session,
WT_LOG_PREPNAME, &logfiles, &logcount));
for (i = 0; i < logcount; i++) {
WT_ERR(__wt_log_extract_lognum(session, logfiles[i], &lognum));
WT_ERR(__wt_log_remove(session, WT_LOG_PREPNAME, lognum));
}
err: WT_TRET(__wt_fs_directory_list_free(session, &logfiles, logcount));
__wt_spin_unlock(session, &log->log_fs_lock);
return (ret);
}
/*
* __log_wait_for_earlier_slot --
* Wait for write_lsn to catch up to this slot.
*/
static void
__log_wait_for_earlier_slot(WT_SESSION_IMPL *session, WT_LOGSLOT *slot)
{
WT_CONNECTION_IMPL *conn;
WT_LOG *log;
int yield_count;
conn = S2C(session);
log = conn->log;
yield_count = 0;
while (__wt_log_cmp(&log->write_lsn, &slot->slot_release_lsn) != 0) {
/*
* If we're on a locked path and the write LSN is not advancing,
* unlock in case an earlier thread is trying to switch its
* slot and complete its operation.
*/
if (F_ISSET(session, WT_SESSION_LOCKED_SLOT))
__wt_spin_unlock(session, &log->log_slot_lock);
__wt_cond_signal(session, conn->log_wrlsn_cond);
if (++yield_count < WT_THOUSAND)
__wt_yield();
else
__wt_cond_wait(session, log->log_write_cond, 200, NULL);
if (F_ISSET(session, WT_SESSION_LOCKED_SLOT))
__wt_spin_lock(session, &log->log_slot_lock);
}
}
/*
* __log_fs_write --
* Wrapper when writing to a log file. If we're writing to a new log
* file for the first time wait for writes to the previous log file.
*/
static int
__log_fs_write(WT_SESSION_IMPL *session,
WT_LOGSLOT *slot, wt_off_t offset, size_t len, const void *buf)
{
WT_DECL_RET;
/*
* If we're writing into a new log file and we're running in
* compatibility mode to an older release, we have to wait for all
* writes to the previous log file to complete otherwise there could
* be a hole at the end of the previous log file that we cannot detect.
*/
if (S2C(session)->log->log_version != WT_LOG_VERSION &&
slot->slot_release_lsn.l.file < slot->slot_start_lsn.l.file) {
__log_wait_for_earlier_slot(session, slot);
WT_RET(__wt_log_force_sync(session, &slot->slot_release_lsn));
}
if ((ret = __wt_write(session, slot->slot_fh, offset, len, buf)) != 0)
WT_PANIC_MSG(session, ret,
"%s: fatal log failure", slot->slot_fh->name);
return (ret);
}
/*
* __wt_log_ckpt --
* Record the given LSN as the checkpoint LSN and signal the archive
* thread as needed.
*/
void
__wt_log_ckpt(WT_SESSION_IMPL *session, WT_LSN *ckp_lsn)
{
WT_CONNECTION_IMPL *conn;
WT_LOG *log;
conn = S2C(session);
log = conn->log;
log->ckpt_lsn = *ckp_lsn;
if (conn->log_cond != NULL)
__wt_cond_signal(session, conn->log_cond);
}
/*
* __wt_log_flush_lsn --
* Force out buffered records and return the LSN, either the
* write_start_lsn or write_lsn depending on the argument.
*/
int
__wt_log_flush_lsn(WT_SESSION_IMPL *session, WT_LSN *lsn, bool start)
{
WT_CONNECTION_IMPL *conn;
WT_LOG *log;
conn = S2C(session);
log = conn->log;
WT_RET(__wt_log_force_write(session, 1, NULL));
__wt_log_wrlsn(session, NULL);
if (start)
*lsn = log->write_start_lsn;
else
*lsn = log->write_lsn;
return (0);
}
/*
* __wt_log_background --
* Record the given LSN as the background LSN and signal the
* thread as needed.
*/
void
__wt_log_background(WT_SESSION_IMPL *session, WT_LSN *lsn)
{
WT_CONNECTION_IMPL *conn;
WT_LOG *log;
conn = S2C(session);
log = conn->log;
/*
* If a thread already set the LSN to a bigger LSN, we're done.
*/
if (__wt_log_cmp(&session->bg_sync_lsn, lsn) > 0)
return;
session->bg_sync_lsn = *lsn;
/*
* Advance the logging subsystem background sync LSN if
* needed.
*/
__wt_spin_lock(session, &log->log_sync_lock);
if (__wt_log_cmp(lsn, &log->bg_sync_lsn) > 0)
log->bg_sync_lsn = *lsn;
__wt_spin_unlock(session, &log->log_sync_lock);
__wt_cond_signal(session, conn->log_file_cond);
}
/*
* __wt_log_force_sync --
* Force a sync of the log and files.
*/
int
__wt_log_force_sync(WT_SESSION_IMPL *session, WT_LSN *min_lsn)
{
WT_DECL_RET;
WT_FH *log_fh;
WT_LOG *log;
uint64_t fsync_duration_usecs, time_start, time_stop;
log = S2C(session)->log;
log_fh = NULL;
/*
* We need to wait for the previous log file to get written
* to disk before we sync out the current one and advance
* the LSN. Signal the worker thread because we know the
* LSN has moved into a later log file and there should be a
* log file ready to close.
*/
while (log->sync_lsn.l.file < min_lsn->l.file) {
__wt_cond_signal(session, S2C(session)->log_file_cond);
__wt_cond_wait(session, log->log_sync_cond, 10000, NULL);
}
__wt_spin_lock(session, &log->log_sync_lock);
WT_ASSERT(session, log->log_dir_fh != NULL);
/*
* Sync the directory if the log file entry hasn't been written
* into the directory.
*/
if (log->sync_dir_lsn.l.file < min_lsn->l.file) {
__wt_verbose(session, WT_VERB_LOG,
"log_force_sync: sync directory %s to LSN %" PRIu32
"/%" PRIu32,
log->log_dir_fh->name, min_lsn->l.file, min_lsn->l.offset);
time_start = __wt_clock(session);
WT_ERR(__wt_fsync(session, log->log_dir_fh, true));
time_stop = __wt_clock(session);
fsync_duration_usecs = WT_CLOCKDIFF_US(time_stop, time_start);
log->sync_dir_lsn = *min_lsn;
WT_STAT_CONN_INCR(session, log_sync_dir);
WT_STAT_CONN_INCRV(session,
log_sync_dir_duration, fsync_duration_usecs);
}
/*
* Sync the log file if needed.
*/
if (__wt_log_cmp(&log->sync_lsn, min_lsn) < 0) {
/*
* Get our own file handle to the log file. It is possible
* for the file handle in the log structure to change out
* from under us and either be NULL or point to a different
* file than we want.
*/
WT_ERR(__log_openfile(session, min_lsn->l.file, 0, &log_fh));
__wt_verbose(session, WT_VERB_LOG,
"log_force_sync: sync %s to LSN %" PRIu32 "/%" PRIu32,
log_fh->name, min_lsn->l.file, min_lsn->l.offset);
time_start = __wt_clock(session);
WT_ERR(__wt_fsync(session, log_fh, true));
time_stop = __wt_clock(session);
fsync_duration_usecs = WT_CLOCKDIFF_US(time_stop, time_start);
log->sync_lsn = *min_lsn;
WT_STAT_CONN_INCR(session, log_sync);
WT_STAT_CONN_INCRV(session,
log_sync_duration, fsync_duration_usecs);
__wt_cond_signal(session, log->log_sync_cond);
}
err:
__wt_spin_unlock(session, &log->log_sync_lock);
if (log_fh != NULL)
WT_TRET(__wt_close(session, &log_fh));
return (ret);
}
/*
* __wt_log_needs_recovery --
* Return 0 if we encounter a clean shutdown and 1 if recovery
* must be run in the given variable.
*/
int
__wt_log_needs_recovery(WT_SESSION_IMPL *session, WT_LSN *ckp_lsn, bool *recp)
{
WT_CONNECTION_IMPL *conn;
WT_CURSOR *c;
WT_DECL_RET;
WT_ITEM dummy_key, dummy_value;
WT_LOG *log;
uint64_t dummy_txnid;
uint32_t dummy_fileid, dummy_optype, rectype;
/*
* Default is to run recovery always (regardless of whether this
* connection has logging enabled).
*/
*recp = true;
conn = S2C(session);
log = conn->log;
if (log == NULL)
return (0);
/*
* See if there are any data modification records between the
* checkpoint LSN and the end of the log. If there are none then
* we can skip recovery.
*/
WT_RET(__wt_curlog_open(session, "log:", NULL, &c));
c->set_key(c, ckp_lsn->l.file, ckp_lsn->l.offset, 0);
if ((ret = c->search(c)) == 0) {
while ((ret = c->next(c)) == 0) {
/*
* The only thing we care about is the rectype.
*/
WT_ERR(c->get_value(c, &dummy_txnid, &rectype,
&dummy_optype, &dummy_fileid,
&dummy_key, &dummy_value));
if (rectype == WT_LOGREC_COMMIT)
break;
}
/*
* If we get to the end of the log, we can skip recovery.
*/
if (ret == WT_NOTFOUND) {
*recp = false;
ret = 0;
}
} else if (ret == WT_NOTFOUND)
/*
* We should always find the checkpoint LSN as it now points
* to the beginning of a written log record. But if we're
* running recovery on an earlier database we may not. In
* that case, we need to run recovery, don't return an error.
*/
ret = 0;
else
WT_ERR(ret);
err: WT_TRET(c->close(c));
return (ret);
}
/*
* __wt_log_written_reset --
* Interface to reset the amount of log written during this
* checkpoint period. Called from the checkpoint code.
*/
void
__wt_log_written_reset(WT_SESSION_IMPL *session)
{
WT_CONNECTION_IMPL *conn;
conn = S2C(session);
if (FLD_ISSET(conn->log_flags, WT_CONN_LOG_ENABLED))
conn->log->log_written = 0;
}
/*
* __wt_log_get_all_files --
* Retrieve the list of log files, either all of them or only the active
* ones (those that are not candidates for archiving). The caller is
* responsible for freeing the directory list returned.
*/
int
__wt_log_get_all_files(WT_SESSION_IMPL *session,
char ***filesp, u_int *countp, uint32_t *maxid, bool active_only)
{
WT_DECL_RET;
WT_LOG *log;
uint32_t id, max;
u_int count, i;
char **files;
*filesp = NULL;
*countp = 0;
*maxid = 0;
id = 0;
log = S2C(session)->log;
/*
* These may be files needed by backup. Force the current slot
* to get written to the file.
*/
WT_RET(__wt_log_force_write(session, 1, NULL));
WT_RET(__log_get_files(session, WT_LOG_FILENAME, &files, &count));
/* Filter out any files that are below the checkpoint LSN. */
for (max = 0, i = 0; i < count; ) {
WT_ERR(__wt_log_extract_lognum(session, files[i], &id));
if (active_only && id < log->ckpt_lsn.l.file) {
/*
* Any files not being returned are individually freed
* and the array adjusted.
*/
__wt_free(session, files[i]);
files[i] = files[count - 1];
files[--count] = NULL;
} else {
if (id > max)
max = id;
i++;
}
}
*maxid = max;
*filesp = files;
*countp = count;
/*
* Only free on error. The caller is responsible for calling free
* once it is done using the returned list.
*/
if (0) {
err: WT_TRET(__wt_fs_directory_list_free(session, &files, count));
}
return (ret);
}
/*
* __log_filename --
* Given a log number, return a WT_ITEM of a generated log file name
* of the given prefix type.
*/
static int
__log_filename(WT_SESSION_IMPL *session,
uint32_t id, const char *file_prefix, WT_ITEM *buf)
{
return (__wt_filename_construct(session,
S2C(session)->log_path, file_prefix, UINTMAX_MAX, id, buf));
}
/*
* __wt_log_extract_lognum --
* Given a log file name, extract out the log number.
*/
int
__wt_log_extract_lognum(
WT_SESSION_IMPL *session, const char *name, uint32_t *id)
{
const char *p;
if (id == NULL || name == NULL)
return (WT_ERROR);
if ((p = strrchr(name, '.')) == NULL ||
sscanf(++p, "%" SCNu32, id) != 1)
WT_RET_MSG(session, WT_ERROR, "Bad log file name '%s'", name);
return (0);
}
/*
* __wt_log_reset --
* Reset the existing log file to after the given file number.
* Called from recovery when toggling logging back on, it was off
* the previous open but it was on earlier before that toggle.
*/
int
__wt_log_reset(WT_SESSION_IMPL *session, uint32_t lognum)
{
WT_CONNECTION_IMPL *conn;
WT_DECL_RET;
WT_LOG *log;
uint32_t old_lognum;
u_int i, logcount;
char **logfiles;
conn = S2C(session);
log = conn->log;
if (!FLD_ISSET(conn->log_flags, WT_CONN_LOG_ENABLED) ||
log->fileid > lognum)
return (0);
WT_ASSERT(session, F_ISSET(conn, WT_CONN_RECOVERING));
WT_ASSERT(session, !F_ISSET(conn, WT_CONN_READONLY));
/*
* We know we're single threaded and called from recovery only when
* toggling logging back on. Therefore the only log files we have are
* old and outdated and the new one created when logging opened before
* recovery. We have to remove all old log files first and then create
* the new one so that log file numbers are contiguous in the file
* system.
*/
WT_RET(__wt_close(session, &log->log_fh));
WT_RET(__log_get_files(session, WT_LOG_FILENAME, &logfiles, &logcount));
for (i = 0; i < logcount; i++) {
WT_ERR(__wt_log_extract_lognum(
session, logfiles[i], &old_lognum));
WT_ASSERT(session, old_lognum < lognum || lognum == 1);
WT_ERR(__wt_log_remove(session, WT_LOG_FILENAME, old_lognum));
}
log->fileid = lognum;
/* Send in true to update connection creation LSNs. */
WT_WITH_SLOT_LOCK(session, log,
ret = __log_newfile(session, true, NULL));
WT_ERR(__wt_log_slot_init(session, false));
err: WT_TRET(__wt_fs_directory_list_free(session, &logfiles, logcount));
return (ret);
}
/*
* __log_zero --
* Zero a log file.
*/
static int
__log_zero(WT_SESSION_IMPL *session,
WT_FH *fh, wt_off_t start_off, wt_off_t len)
{
WT_CONNECTION_IMPL *conn;
WT_DECL_ITEM(zerobuf);
WT_DECL_RET;
WT_LOG *log;
uint32_t allocsize, bufsz, off, partial, wrlen;
conn = S2C(session);
log = conn->log;
allocsize = log->allocsize;
zerobuf = NULL;
if (allocsize < WT_MEGABYTE)
bufsz = WT_MEGABYTE;
else
bufsz = allocsize;
/*
* If they're using smaller log files, cap it at the file size.
*/
if (conn->log_file_max < bufsz)
bufsz = (uint32_t)conn->log_file_max;
WT_RET(__wt_scr_alloc(session, bufsz, &zerobuf));
memset(zerobuf->mem, 0, zerobuf->memsize);
WT_STAT_CONN_INCR(session, log_zero_fills);
/*
* Read in a chunk starting at the end of the file. Keep going until
* we reach the beginning or we find a chunk that contains any non-zero
* bytes. Compare against a known zero byte chunk.
*/
off = (uint32_t)start_off;
while (off < (uint32_t)len) {
/*
* Typically we start to zero the file after the log header
* and the bufsz is a sector-aligned size. So we want to
* align our writes when we can.
*/
partial = off % bufsz;
if (partial != 0)
wrlen = bufsz - partial;
else
wrlen = bufsz;
/*
* Check if we're writing a partial amount at the end too.
*/
if ((uint32_t)len - off < bufsz)
wrlen = (uint32_t)len - off;
WT_ERR(__wt_write(session,
fh, (wt_off_t)off, wrlen, zerobuf->mem));
off += wrlen;
}
err: __wt_scr_free(session, &zerobuf);
return (ret);
}
/*
* __log_prealloc --
* Pre-allocate a log file.
*/
static int
__log_prealloc(WT_SESSION_IMPL *session, WT_FH *fh)
{
WT_CONNECTION_IMPL *conn;
WT_DECL_RET;
WT_LOG *log;
conn = S2C(session);
log = conn->log;
/*
* If the user configured zero filling, pre-allocate the log file
* manually. Otherwise use the file extension method to create
* and zero the log file based on what is available.
*/
if (FLD_ISSET(conn->log_flags, WT_CONN_LOG_ZERO_FILL))
return (__log_zero(session, fh,
WT_LOG_END_HEADER, conn->log_file_max));
/*
* We have exclusive access to the log file and there are no other
* writes happening concurrently, so there are no locking issues.
*/
ret = __wt_fextend(session, fh, conn->log_file_max);
return (ret == EBUSY || ret == ENOTSUP ? 0 : ret);
}
/*
* __log_size_fit --
* Return whether or not recsize will fit in the log file.
*/
static int
__log_size_fit(WT_SESSION_IMPL *session, WT_LSN *lsn, uint64_t recsize)
{
WT_CONNECTION_IMPL *conn;
WT_LOG *log;
conn = S2C(session);
log = conn->log;
return (lsn->l.offset == log->first_record ||
lsn->l.offset + (wt_off_t)recsize < conn->log_file_max);
}
/*
* __log_decompress --
* Decompress a log record.
*/
static int
__log_decompress(WT_SESSION_IMPL *session, WT_ITEM *in, WT_ITEM *out)
{
WT_COMPRESSOR *compressor;
WT_CONNECTION_IMPL *conn;
WT_LOG_RECORD *logrec;
size_t result_len, skip;
uint32_t uncompressed_size;
conn = S2C(session);
logrec = (WT_LOG_RECORD *)in->mem;
skip = WT_LOG_COMPRESS_SKIP;
compressor = conn->log_compressor;
if (compressor == NULL || compressor->decompress == NULL)
WT_RET_MSG(session, WT_ERROR,
"log_decompress: Compressed record with "
"no configured compressor");
uncompressed_size = logrec->mem_len;
WT_RET(__wt_buf_initsize(session, out, uncompressed_size));
memcpy(out->mem, in->mem, skip);
WT_RET(compressor->decompress(compressor, &session->iface,
(uint8_t *)in->mem + skip, in->size - skip,
(uint8_t *)out->mem + skip,
uncompressed_size - skip, &result_len));
/*
* If checksums were turned off because we're depending on the
* decompression to fail on any corrupted data, we'll end up
* here after corruption happens. If we're salvaging the file,
* it's OK, otherwise it's really, really bad.
*/
if (result_len != uncompressed_size - WT_LOG_COMPRESS_SKIP)
return (WT_ERROR);
return (0);
}
/*
* __log_decrypt --
* Decrypt a log record.
*/
static int
__log_decrypt(WT_SESSION_IMPL *session, WT_ITEM *in, WT_ITEM *out)
{
WT_CONNECTION_IMPL *conn;
WT_ENCRYPTOR *encryptor;
WT_KEYED_ENCRYPTOR *kencryptor;
conn = S2C(session);
kencryptor = conn->kencryptor;
if (kencryptor == NULL ||
(encryptor = kencryptor->encryptor) == NULL ||
encryptor->decrypt == NULL)
WT_RET_MSG(session, WT_ERROR,
"log_decrypt: Encrypted record with "
"no configured decrypt method");
return (__wt_decrypt(session, encryptor, WT_LOG_ENCRYPT_SKIP, in, out));
}
/*
* __wt_log_fill --
* Copy a thread's log records into the assigned slot.
*/
int
__wt_log_fill(WT_SESSION_IMPL *session,
WT_MYSLOT *myslot, bool force, WT_ITEM *record, WT_LSN *lsnp)
{
WT_DECL_RET;
/*
* Call write or copy into the buffer. For now the offset is the
* real byte offset. If the offset becomes a unit of WT_LOG_ALIGN this
* is where we would multiply by WT_LOG_ALIGN to get the real file byte
* offset for write().
*/
if (!force && !F_ISSET(myslot, WT_MYSLOT_UNBUFFERED))
memcpy((char *)myslot->slot->slot_buf.mem + myslot->offset,
record->mem, record->size);
else
/*
* If this is a force or unbuffered write, write it now.
*/
WT_ERR(__log_fs_write(session, myslot->slot,
myslot->offset + myslot->slot->slot_start_offset,
record->size, record->mem));
WT_STAT_CONN_INCRV(session, log_bytes_written, record->size);
if (lsnp != NULL) {
*lsnp = myslot->slot->slot_start_lsn;
lsnp->l.offset += (uint32_t)myslot->offset;
}
err:
if (ret != 0 && myslot->slot->slot_error == 0)
myslot->slot->slot_error = ret;
return (ret);
}
/*
* __log_file_header --
* Create and write a log file header into a file handle. If writing
* into the main log, it will be called locked. If writing into a
* pre-allocated log, it will be called unlocked.
*/
static int
__log_file_header(
WT_SESSION_IMPL *session, WT_FH *fh, WT_LSN *end_lsn, bool prealloc)
{
WT_CONNECTION_IMPL *conn;
WT_DECL_ITEM(buf);
WT_DECL_RET;
WT_LOG *log;
WT_LOGSLOT tmp;
WT_LOG_DESC *desc;
WT_LOG_RECORD *logrec;
WT_MYSLOT myslot;
conn = S2C(session);
log = conn->log;
/*
* Set up the log descriptor record. Use a scratch buffer to
* get correct alignment for direct I/O.
*/
WT_ASSERT(session, sizeof(WT_LOG_DESC) < log->allocsize);
WT_RET(__wt_scr_alloc(session, log->allocsize, &buf));
memset(buf->mem, 0, log->allocsize);
buf->size = log->allocsize;
logrec = (WT_LOG_RECORD *)buf->mem;
desc = (WT_LOG_DESC *)logrec->record;
desc->log_magic = WT_LOG_MAGIC;
desc->version = log->log_version;
desc->log_size = (uint64_t)conn->log_file_max;
__wt_log_desc_byteswap(desc);
/*
* Now that the record is set up, initialize the record header.
*
* Checksum a little-endian version of the header, and write everything
* in little-endian format. The checksum is (potentially) returned in a
* big-endian format, swap it into place in a separate step.
*/
logrec->len = log->allocsize;
logrec->checksum = 0;
__wt_log_record_byteswap(logrec);
logrec->checksum = __wt_checksum(logrec, log->allocsize);
#ifdef WORDS_BIGENDIAN
logrec->checksum = __wt_bswap32(logrec->checksum);
#endif
WT_CLEAR(tmp);
memset(&myslot, 0, sizeof(myslot));
myslot.slot = &tmp;
/*
* We may recursively call __wt_log_acquire to allocate log space for
* the log descriptor record. Call __wt_log_fill to write it, but we
* do not need to call __wt_log_release because we're not waiting for
* any earlier operations to complete.
*/
if (prealloc) {
WT_ASSERT(session, fh != NULL);
tmp.slot_fh = fh;
} else {
WT_ASSERT(session, fh == NULL);
WT_ERR(__wt_log_acquire(session, log->allocsize, &tmp));
}
WT_ERR(__wt_log_fill(session, &myslot, true, buf, NULL));
/*
* Make sure the header gets to disk.
*/
WT_ERR(__wt_fsync(session, tmp.slot_fh, true));
if (end_lsn != NULL)
*end_lsn = tmp.slot_end_lsn;
err: __wt_scr_free(session, &buf);
return (ret);
}
/*
* __log_openfile --
* Open a log file with the given log file number and return the WT_FH.
*/
static int
__log_openfile(
WT_SESSION_IMPL *session, uint32_t id, uint32_t flags, WT_FH **fhp)
{
WT_CONNECTION_IMPL *conn;
WT_DECL_ITEM(buf);
WT_DECL_RET;
u_int wtopen_flags;
conn = S2C(session);
WT_RET(__wt_scr_alloc(session, 0, &buf));
/*
* If we are creating the file then we use a temporary file name.
* Otherwise it is a log file name.
*/
if (LF_ISSET(WT_LOG_OPEN_CREATE_OK)) {
wtopen_flags = WT_FS_OPEN_CREATE;
WT_ERR(__log_filename(session, id, WT_LOG_TMPNAME, buf));
} else {
wtopen_flags = 0;
WT_ERR(__log_filename(session, id, WT_LOG_FILENAME, buf));
}
__wt_verbose(session, WT_VERB_LOG,
"opening log %s", (const char *)buf->data);
if (FLD_ISSET(conn->direct_io, WT_DIRECT_IO_LOG))
FLD_SET(wtopen_flags, WT_FS_OPEN_DIRECTIO);
WT_ERR(__wt_open(
session, buf->data, WT_FS_OPEN_FILE_TYPE_LOG, wtopen_flags, fhp));
err: __wt_scr_free(session, &buf);
return (ret);
}
/*
* __log_open_verify --
* Open a log file with the given log file number, verify its
* header and return various pieces of system information about
* this log file.
*/
static int
__log_open_verify(WT_SESSION_IMPL *session, uint32_t id, WT_FH **fhp,
WT_LSN *lsnp, uint16_t *versionp)
{
WT_CONNECTION_IMPL *conn;
WT_DECL_ITEM(buf);
WT_DECL_RET;
WT_FH *fh;
WT_LOG *log;
WT_LOG_DESC *desc;
WT_LOG_RECORD *logrec;
uint32_t allocsize, rectype;
const uint8_t *end, *p;
conn = S2C(session);
log = conn->log;
WT_RET(__wt_scr_alloc(session, 0, &buf));
WT_ERR(__log_openfile(session, id, 0, &fh));
if (log == NULL)
allocsize = WT_LOG_ALIGN;
else
allocsize = log->allocsize;
if (lsnp != NULL)
WT_ZERO_LSN(lsnp);
WT_ERR(__wt_buf_grow(session, buf, allocsize));
memset(buf->mem, 0, allocsize);
/*
* Read in the log file header and verify it.
*/
WT_ERR(__wt_read(session, fh, 0, allocsize, buf->mem));
logrec = (WT_LOG_RECORD *)buf->mem;
__wt_log_record_byteswap(logrec);
desc = (WT_LOG_DESC *)logrec->record;
__wt_log_desc_byteswap(desc);
if (desc->log_magic != WT_LOG_MAGIC)
WT_PANIC_RET(session, WT_ERROR,
"log file %s corrupted: Bad magic number %" PRIu32,
fh->name, desc->log_magic);
/*
* We cannot read future log file formats.
*/
if (desc->version > WT_LOG_VERSION)
WT_ERR_MSG(session, WT_ERROR,
"unsupported WiredTiger file version: this build "
" only supports versions up to %d,"
" and the file is version %" PRIu16,
WT_LOG_VERSION, desc->version);
/*
* Set up the return values if the magic number is valid.
*/
if (versionp != NULL)
*versionp = desc->version;
/*
* Skip reading in the previous LSN if log file is an old version
* or if the caller doesn't care about the LSN. Otherwise read that
* record in and set up the LSN. We already have a buffer that is
* the correct size. Reuse it.
*/
if (lsnp == NULL ||
(desc->version < WT_LOG_VERSION_SYSTEM))
goto err;
memset(buf->mem, 0, allocsize);
WT_ERR(__wt_read(session, fh, allocsize, allocsize, buf->mem));
logrec = (WT_LOG_RECORD *)buf->mem;
/*
* We have a valid header but the system record is not there.
* The log ends here. Return without setting the LSN.
*/
if (logrec->len == 0) {
__wt_verbose(session, WT_VERB_LOG,
"Log %s found empty log after header", fh->name);
goto err;
}
if (!__log_checksum_match(session, buf, allocsize))
WT_ERR_MSG(session, WT_ERROR,
"%s: System log record checksum mismatch", fh->name);
__wt_log_record_byteswap(logrec);
p = WT_LOG_SKIP_HEADER(buf->data);
end = (const uint8_t *)buf->data + allocsize;
WT_ERR(__wt_logrec_read(session, &p, end, &rectype));
if (rectype != WT_LOGREC_SYSTEM)
WT_ERR_MSG(session, WT_ERROR, "System log record missing");
WT_ERR(__wt_log_recover_system(session, &p, end, lsnp));
err: __wt_scr_free(session, &buf);
/*
* Return the file handle if needed, otherwise close it.
*/
if (fhp != NULL && ret == 0)
*fhp = fh;
else
WT_TRET(__wt_close(session, &fh));
return (ret);
}
/*
* __log_alloc_prealloc --
* Look for a pre-allocated log file and rename it to use as the next
* real log file. Called locked.
*/
static int
__log_alloc_prealloc(WT_SESSION_IMPL *session, uint32_t to_num)
{
WT_CONNECTION_IMPL *conn;
WT_DECL_ITEM(from_path);
WT_DECL_ITEM(to_path);
WT_DECL_RET;
WT_LOG *log;
uint32_t from_num;
u_int logcount;
char **logfiles;
/*
* If there are no pre-allocated files, return WT_NOTFOUND.
*/
conn = S2C(session);
log = conn->log;
logfiles = NULL;
WT_ERR(__log_get_files_single(
session, WT_LOG_PREPNAME, &logfiles, &logcount));
if (logcount == 0)
return (WT_NOTFOUND);
/* We have a file to use. */
WT_ERR(__wt_log_extract_lognum(session, logfiles[0], &from_num));
WT_ERR(__wt_scr_alloc(session, 0, &from_path));
WT_ERR(__wt_scr_alloc(session, 0, &to_path));
WT_ERR(__log_filename(session, from_num, WT_LOG_PREPNAME, from_path));
WT_ERR(__log_filename(session, to_num, WT_LOG_FILENAME, to_path));
__wt_spin_lock(session, &log->log_fs_lock);
__wt_verbose(session, WT_VERB_LOG,
"log_alloc_prealloc: rename log %s to %s",
(const char *)from_path->data, (const char *)to_path->data);
WT_STAT_CONN_INCR(session, log_prealloc_used);
/*
* All file setup, writing the header and pre-allocation was done
* before. We only need to rename it.
*/
WT_ERR(__wt_fs_rename(session, from_path->data, to_path->data, false));
err: __wt_scr_free(session, &from_path);
__wt_scr_free(session, &to_path);
__wt_spin_unlock(session, &log->log_fs_lock);
WT_TRET(__wt_fs_directory_list_free(session, &logfiles, logcount));
return (ret);
}
/*
* __log_newfile --
* Create the next log file and write the file header record into it.
*/
static int
__log_newfile(WT_SESSION_IMPL *session, bool conn_open, bool *created)
{
WT_CONNECTION_IMPL *conn;
WT_DECL_RET;
WT_FH *log_fh;
WT_LOG *log;
WT_LSN end_lsn, logrec_lsn;
u_int yield_cnt;
bool create_log;
conn = S2C(session);
log = conn->log;
/*
* Set aside the log file handle to be closed later. Other threads
* may still be using it to write to the log. If the log file size
* is small we could fill a log file before the previous one is closed.
* Wait for that to close.
*/
WT_ASSERT(session, F_ISSET(session, WT_SESSION_LOCKED_SLOT));
for (yield_cnt = 0; log->log_close_fh != NULL;) {
WT_STAT_CONN_INCR(session, log_close_yields);
__wt_log_wrlsn(session, NULL);
if (++yield_cnt > 10000)
return (EBUSY);
__wt_yield();
}
/*
* Note, the file server worker thread requires the LSN be set once the
* close file handle is set, force that ordering.
*/
if (log->log_fh == NULL)
log->log_close_fh = NULL;
else {
log->log_close_lsn = log->alloc_lsn;
WT_PUBLISH(log->log_close_fh, log->log_fh);
}
log->fileid++;
/*
* If pre-allocating log files look for one; otherwise, or if we don't
* find one, create a log file. We can't use pre-allocated log files
* while a hot backup is in progress: applications can copy the files
* in any way they choose, and a log file rename might confuse things.
*/
create_log = true;
if (conn->log_prealloc > 0 && !conn->hot_backup) {
__wt_readlock(session, &conn->hot_backup_lock);
if (conn->hot_backup)
__wt_readunlock(session, &conn->hot_backup_lock);
else {
ret = __log_alloc_prealloc(session, log->fileid);
__wt_readunlock(session, &conn->hot_backup_lock);
/*
* If ret is 0 it means we found a pre-allocated file.
* If ret is WT_NOTFOUND, create the new log file and
* signal the server, we missed our pre-allocation.
* If ret is non-zero but not WT_NOTFOUND, return the
* error.
*/
WT_RET_NOTFOUND_OK(ret);
if (ret == 0)
create_log = false;
else {
WT_STAT_CONN_INCR(session, log_prealloc_missed);
if (conn->log_cond != NULL)
__wt_cond_signal(
session, conn->log_cond);
}
}
}
/*
* If we need to create the log file, do so now.
*/
if (create_log) {
/*
* Increment the missed pre-allocated file counter only
* if a hot backup is not in progress. We are deliberately
* not using pre-allocated log files during backup
* (see comment above).
*/
if (!conn->hot_backup)
log->prep_missed++;
WT_RET(__wt_log_allocfile(
session, log->fileid, WT_LOG_FILENAME));
}
/*
* Since the file system clears the output file handle pointer before
* searching the handle list and filling in the new file handle,
* we must pass in a local file handle. Otherwise there is a wide
* window where another thread could see a NULL log file handle.
*/
WT_RET(__log_open_verify(session, log->fileid, &log_fh, NULL, NULL));
/*
* Write the LSN at the end of the last record in the previous log file
* as the first record in this log file.
*/
if (log->fileid == 1)
WT_INIT_LSN(&logrec_lsn);
else
logrec_lsn = log->alloc_lsn;
/*
* We need to setup the LSNs. Set the end LSN and alloc LSN to
* the end of the header.
*/
WT_SET_LSN(&log->alloc_lsn, log->fileid, WT_LOG_END_HEADER);
/*
* If we're running the version where we write a system record
* do so now and update the alloc_lsn.
*/
if (log->log_version == WT_LOG_VERSION) {
WT_RET(__wt_log_system_record(session,
log_fh, &logrec_lsn));
WT_SET_LSN(&log->alloc_lsn, log->fileid, log->first_record);
}
end_lsn = log->alloc_lsn;
WT_PUBLISH(log->log_fh, log_fh);
/*
* If we're called from connection creation code, we need to update
* the LSNs since we're the only write in progress.
*/
if (conn_open) {
WT_RET(__wt_fsync(session, log->log_fh, true));
log->sync_lsn = end_lsn;
log->write_lsn = end_lsn;
log->write_start_lsn = end_lsn;
}
if (created != NULL)
*created = create_log;
return (0);
}
/*
* __log_set_version --
* Set version related information under lock.
*/
static int
__log_set_version(WT_SESSION_IMPL *session, uint16_t version,
uint32_t first_rec, bool live_chg, bool downgrade)
{
WT_CONNECTION_IMPL *conn;
WT_LOG *log;
conn = S2C(session);
log = conn->log;
log->log_version = version;
log->first_record = first_rec;
if (downgrade)
FLD_SET(conn->log_flags, WT_CONN_LOG_DOWNGRADED);
else
FLD_CLR(conn->log_flags, WT_CONN_LOG_DOWNGRADED);
if (live_chg)
F_SET(log, WT_LOG_FORCE_NEWFILE);
if (!F_ISSET(conn, WT_CONN_READONLY))
return (__log_prealloc_remove(session));
return (0);
}
/*
* __wt_log_set_version --
* Change the version number in logging. Will be done with locking.
* We need to force the log file to advance and remove all old
* pre-allocated files.
*/
int
__wt_log_set_version(WT_SESSION_IMPL *session, uint16_t version,
uint32_t first_rec, bool downgrade, bool live_chg, uint32_t *lognump)
{
WT_CONNECTION_IMPL *conn;
WT_DECL_RET;
WT_LOG *log;
conn = S2C(session);
log = conn->log;
/*
* The steps are:
* - Set up versions and remove files under lock.
* - Set a flag so that the next slot change forces a file change.
* - Force out the slot that is currently active in the current log.
* - Write a log record to force a record into the new log file.
*/
WT_WITH_SLOT_LOCK(session, log,
ret = __log_set_version(session,
version, first_rec, live_chg, downgrade));
if (!live_chg)
return (ret);
WT_ERR(ret);
/*
* A new log file will be used when we force out the earlier slot.
*/
WT_ERR(__wt_log_force_write(session, 1, NULL));
/*
* We need to write a record to the new version log file so that
* a potential checkpoint finds LSNs in that new log file and
* an archive correctly removes all earlier logs.
* Write an internal printf record.
*/
WT_ERR(__wt_log_printf(session,
"COMPATIBILITY: Version now %" PRIu16, log->log_version));
if (lognump != NULL)
*lognump = log->alloc_lsn.l.file;
err:
return (ret);
}
/*
* __wt_log_acquire --
* Called serially when switching slots. Can be called recursively
* from __log_newfile when we change log files.
*/
int
__wt_log_acquire(WT_SESSION_IMPL *session, uint64_t recsize, WT_LOGSLOT *slot)
{
WT_CONNECTION_IMPL *conn;
WT_LOG *log;
bool created_log;
conn = S2C(session);
log = conn->log;
created_log = true;
/*
* Add recsize to alloc_lsn. Save our starting LSN
* where the previous allocation finished for the release LSN.
* That way when log files switch, we're waiting for the correct LSN
* from outstanding writes.
*/
WT_ASSERT(session, F_ISSET(session, WT_SESSION_LOCKED_SLOT));
/*
* We need to set the release LSN earlier, before a log file change.
*/
slot->slot_release_lsn = log->alloc_lsn;
/*
* Make sure that the size can fit in the file. Proactively switch
* if it cannot. This reduces, but does not eliminate, log files
* that exceed the maximum file size. We want to minimize the risk
* of an error due to no space.
*/
if (F_ISSET(log, WT_LOG_FORCE_NEWFILE) ||
!__log_size_fit(session, &log->alloc_lsn, recsize)) {
WT_RET(__log_newfile(session, false, &created_log));
F_CLR(log, WT_LOG_FORCE_NEWFILE);
if (log->log_close_fh != NULL)
F_SET(slot, WT_SLOT_CLOSEFH);
}
/*
* Pre-allocate on the first real write into the log file, if it
* was just created (i.e. not pre-allocated).
*/
if (log->alloc_lsn.l.offset == log->first_record && created_log)
WT_RET(__log_prealloc(session, log->log_fh));
/*
* Initialize the slot for activation.
*/
__wt_log_slot_activate(session, slot);
return (0);
}
/*
* __log_truncate_file --
* Truncate a log file to the specified offset.
*
* If the underlying file system doesn't support truncate then we need to
* zero out the rest of the file, doing an effective truncate.
*/
static int
__log_truncate_file(WT_SESSION_IMPL *session, WT_FH *log_fh, wt_off_t offset)
{
WT_CONNECTION_IMPL *conn;
WT_DECL_RET;
WT_LOG *log;
conn = S2C(session);
log = conn->log;
if (!F_ISSET(log, WT_LOG_TRUNCATE_NOTSUP) && !conn->hot_backup) {
__wt_readlock(session, &conn->hot_backup_lock);
if (conn->hot_backup)
__wt_readunlock(session, &conn->hot_backup_lock);
else {
ret = __wt_ftruncate(session, log_fh, offset);
__wt_readunlock(session, &conn->hot_backup_lock);
if (ret != ENOTSUP)
return (ret);
F_SET(log, WT_LOG_TRUNCATE_NOTSUP);
}
}
return (__log_zero(session, log_fh, offset, conn->log_file_max));
}
/*
* __log_truncate --
* Truncate the log to the given LSN. If this_log is set, it will only
* truncate the log file indicated in the given LSN. If not set,
* it will truncate between the given LSN and the trunc_lsn. That is,
* since we pre-allocate log files, it will free that space and allow the
* log to be traversed. We use the trunc_lsn because logging has already
* opened the new/next log file before recovery ran. This function assumes
* we are in recovery or other dedicated time and not during live running.
*/
static int
__log_truncate(WT_SESSION_IMPL *session, WT_LSN *lsn, bool this_log)
{
WT_CONNECTION_IMPL *conn;
WT_DECL_RET;
WT_FH *log_fh;
WT_LOG *log;
uint32_t lognum;
u_int i, logcount;
char **logfiles;
conn = S2C(session);
log = conn->log;
log_fh = NULL;
logcount = 0;
logfiles = NULL;
/*
* Truncate the log file to the given LSN.
*
* It's possible the underlying file system doesn't support truncate
* (there are existing examples), which is fine, but we don't want to
* repeatedly do the setup work just to find that out every time. Check
* before doing work, and if there's a not-supported error, turn off
* future truncates.
*/
WT_ERR(__log_openfile(session, lsn->l.file, 0, &log_fh));
WT_ERR(__log_truncate_file(session, log_fh, lsn->l.offset));
WT_ERR(__wt_fsync(session, log_fh, true));
WT_ERR(__wt_close(session, &log_fh));
/*
* If we just want to truncate the current log, return and skip
* looking for intervening logs.
*/
if (this_log)
goto err;
WT_ERR(__log_get_files(session, WT_LOG_FILENAME, &logfiles, &logcount));
for (i = 0; i < logcount; i++) {
WT_ERR(__wt_log_extract_lognum(session, logfiles[i], &lognum));
if (lognum > lsn->l.file && lognum < log->trunc_lsn.l.file) {
WT_ERR(__log_openfile(session, lognum, 0, &log_fh));
/*
* If there are intervening files pre-allocated,
* truncate them to the end of the log file header.
*/
WT_ERR(__log_truncate_file(
session, log_fh, log->first_record));
WT_ERR(__wt_fsync(session, log_fh, true));
WT_ERR(__wt_close(session, &log_fh));
}
}
err: WT_TRET(__wt_close(session, &log_fh));
WT_TRET(__wt_fs_directory_list_free(session, &logfiles, logcount));
return (ret);
}
/*
* __wt_log_allocfile --
* Given a log number, create a new log file by writing the header,
* pre-allocating the file and moving it to the destination name.
*/
int
__wt_log_allocfile(
WT_SESSION_IMPL *session, uint32_t lognum, const char *dest)
{
WT_CONNECTION_IMPL *conn;
WT_DECL_ITEM(from_path);
WT_DECL_ITEM(to_path);
WT_DECL_RET;
WT_FH *log_fh;
WT_LOG *log;
uint32_t tmp_id;
conn = S2C(session);
log = conn->log;
log_fh = NULL;
/*
* Preparing a log file entails creating a temporary file:
* - Writing the header.
* - Truncating to the offset of the first record.
* - Pre-allocating the file if needed.
* - Renaming it to the desired file name.
*/
WT_RET(__wt_scr_alloc(session, 0, &from_path));
WT_ERR(__wt_scr_alloc(session, 0, &to_path));
tmp_id = __wt_atomic_add32(&log->tmp_fileid, 1);
WT_ERR(__log_filename(session, tmp_id, WT_LOG_TMPNAME, from_path));
WT_ERR(__log_filename(session, lognum, dest, to_path));
__wt_spin_lock(session, &log->log_fs_lock);
/*
* Set up the temporary file.
*/
WT_ERR(__log_openfile(session, tmp_id, WT_LOG_OPEN_CREATE_OK, &log_fh));
WT_ERR(__log_file_header(session, log_fh, NULL, true));
WT_ERR(__log_prealloc(session, log_fh));
WT_ERR(__wt_fsync(session, log_fh, true));
WT_ERR(__wt_close(session, &log_fh));
__wt_verbose(session, WT_VERB_LOG,
"log_allocfile: rename %s to %s",
(const char *)from_path->data, (const char *)to_path->data);
/*
* Rename it into place and make it available.
*/
WT_ERR(__wt_fs_rename(session, from_path->data, to_path->data, false));
err: __wt_scr_free(session, &from_path);
__wt_scr_free(session, &to_path);
__wt_spin_unlock(session, &log->log_fs_lock);
WT_TRET(__wt_close(session, &log_fh));
return (ret);
}
/*
* __wt_log_remove --
* Given a log number, remove that log file.
*/
int
__wt_log_remove(WT_SESSION_IMPL *session,
const char *file_prefix, uint32_t lognum)
{
WT_DECL_ITEM(path);
WT_DECL_RET;
WT_RET(__wt_scr_alloc(session, 0, &path));
WT_ERR(__log_filename(session, lognum, file_prefix, path));
__wt_verbose(session, WT_VERB_LOG,
"log_remove: remove log %s", (const char *)path->data);
WT_ERR(__wt_fs_remove(session, path->data, false));
err: __wt_scr_free(session, &path);
return (ret);
}
/*
* __wt_log_open --
* Open the appropriate log file for the connection. The purpose is
* to find the last log file that exists, open it and set our initial
* LSNs to the end of that file. If none exist, call __log_newfile
* to create it.
*/
int
__wt_log_open(WT_SESSION_IMPL *session)
{
WT_CONNECTION_IMPL *conn;
WT_DECL_RET;
WT_LOG *log;
uint32_t firstlog, lastlog, lognum;
uint16_t version;
u_int i, logcount;
char **logfiles;
conn = S2C(session);
log = conn->log;
logfiles = NULL;
logcount = 0;
lastlog = 0;
firstlog = UINT32_MAX;
/*
* Open up a file handle to the log directory if we haven't.
*/
if (log->log_dir_fh == NULL) {
__wt_verbose(session, WT_VERB_LOG,
"log_open: open fh to directory %s", conn->log_path);
WT_RET(__wt_open(session, conn->log_path,
WT_FS_OPEN_FILE_TYPE_DIRECTORY, 0, &log->log_dir_fh));
}
if (!F_ISSET(conn, WT_CONN_READONLY))
WT_ERR(__log_prealloc_remove(session));
/*
* Now look at the log files and set our LSNs.
*/
WT_ERR(__log_get_files(session, WT_LOG_FILENAME, &logfiles, &logcount));
for (i = 0; i < logcount; i++) {
WT_ERR(__wt_log_extract_lognum(session, logfiles[i], &lognum));
lastlog = WT_MAX(lastlog, lognum);
firstlog = WT_MIN(firstlog, lognum);
}
log->fileid = lastlog;
__wt_verbose(session, WT_VERB_LOG,
"log_open: first log %" PRIu32 " last log %" PRIu32,
firstlog, lastlog);
if (firstlog == UINT32_MAX) {
WT_ASSERT(session, logcount == 0);
WT_INIT_LSN(&log->first_lsn);
} else
WT_SET_LSN(&log->first_lsn, firstlog, 0);
/*
* Start logging at the beginning of the next log file, no matter
* where the previous log file ends.
*/
if (!F_ISSET(conn, WT_CONN_READONLY)) {
WT_WITH_SLOT_LOCK(session, log,
ret = __log_newfile(session, true, NULL));
WT_ERR(ret);
}
/* If we found log files, save the new state. */
if (logcount > 0) {
/*
* If we're running in a downgraded mode and there are earlier
* logs detect if they're at a higher version. If so, we need
* to force recovery (to write a full checkpoint) and force
* archiving to remove all higher version logs.
*/
if (FLD_ISSET(conn->log_flags, WT_CONN_LOG_DOWNGRADED)) {
for (i = 0; i < logcount; ++i) {
WT_ERR(__wt_log_extract_lognum(
session, logfiles[i], &lognum));
/*
* By sending in a NULL file handle, we don't
* have to close the file.
*/
WT_ERR(__log_open_verify(session,
lognum, NULL, NULL, &version));
/*
* If we find any log file at the wrong version
* set the flag and we're done.
*/
if (log->log_version != version) {
FLD_SET(conn->log_flags,
WT_CONN_LOG_FORCE_DOWNGRADE);
break;
}
}
}
log->trunc_lsn = log->alloc_lsn;
FLD_SET(conn->log_flags, WT_CONN_LOG_EXISTED);
}
err: WT_TRET(__wt_fs_directory_list_free(session, &logfiles, logcount));
if (ret == 0)
F_SET(log, WT_LOG_OPENED);
return (ret);
}
/*
* __wt_log_close --
* Close the log file.
*/
int
__wt_log_close(WT_SESSION_IMPL *session)
{
WT_CONNECTION_IMPL *conn;
WT_LOG *log;
conn = S2C(session);
log = conn->log;
if (log->log_close_fh != NULL && log->log_close_fh != log->log_fh) {
__wt_verbose(session, WT_VERB_LOG,
"closing old log %s", log->log_close_fh->name);
if (!F_ISSET(conn, WT_CONN_READONLY))
WT_RET(__wt_fsync(session, log->log_close_fh, true));
WT_RET(__wt_close(session, &log->log_close_fh));
}
if (log->log_fh != NULL) {
__wt_verbose(session, WT_VERB_LOG,
"closing log %s", log->log_fh->name);
if (!F_ISSET(conn, WT_CONN_READONLY))
WT_RET(__wt_fsync(session, log->log_fh, true));
WT_RET(__wt_close(session, &log->log_fh));
log->log_fh = NULL;
}
if (log->log_dir_fh != NULL) {
__wt_verbose(session, WT_VERB_LOG,
"closing log directory %s", log->log_dir_fh->name);
if (!F_ISSET(conn, WT_CONN_READONLY))
WT_RET(__wt_fsync(session, log->log_dir_fh, true));
WT_RET(__wt_close(session, &log->log_dir_fh));
log->log_dir_fh = NULL;
}
F_CLR(log, WT_LOG_OPENED);
return (0);
}
/*
* __log_has_hole --
* Determine if the current offset represents a hole in the log
* file (i.e. there is valid data somewhere after the hole), or
* if this is the end of this log file and the remainder of the
* file is zeroes.
*/
static int
__log_has_hole(WT_SESSION_IMPL *session,
WT_FH *fh, wt_off_t log_size, wt_off_t offset, bool *hole)
{
WT_CONNECTION_IMPL *conn;
WT_DECL_RET;
WT_LOG *log;
wt_off_t off, remainder;
size_t bufsz, rdlen;
char *buf, *zerobuf;
*hole = false;
conn = S2C(session);
log = conn->log;
remainder = log_size - offset;
/*
* It can be very slow looking for the last real record in the log
* in very small chunks. Walk a megabyte at a time. If we find a
* part of the log that is not just zeroes we know this log file
* has a hole in it.
*/
buf = zerobuf = NULL;
if (log == NULL || log->allocsize < WT_MEGABYTE)
bufsz = WT_MEGABYTE;
else
bufsz = log->allocsize;
if ((size_t)remainder < bufsz)
bufsz = (size_t)remainder;
WT_RET(__wt_calloc_def(session, bufsz, &buf));
WT_ERR(__wt_calloc_def(session, bufsz, &zerobuf));
/*
* Read in a chunk starting at the given offset.
* Compare against a known zero byte chunk.
*/
for (off = offset; remainder > 0;
remainder -= (wt_off_t)rdlen, off += (wt_off_t)rdlen) {
rdlen = WT_MIN(bufsz, (size_t)remainder);
WT_ERR(__wt_read(session, fh, off, rdlen, buf));
if (memcmp(buf, zerobuf, rdlen) != 0) {
*hole = true;
break;
}
}
err: __wt_free(session, buf);
__wt_free(session, zerobuf);
return (ret);
}
/*
* __wt_log_release --
* Release a log slot.
*/
int
__wt_log_release(WT_SESSION_IMPL *session, WT_LOGSLOT *slot, bool *freep)
{
WT_CONNECTION_IMPL *conn;
WT_DECL_RET;
WT_LOG *log;
WT_LSN sync_lsn;
uint64_t fsync_duration_usecs, time_start, time_stop;
int64_t release_buffered, release_bytes;
bool locked;
conn = S2C(session);
log = conn->log;
locked = false;
if (freep != NULL)
*freep = 1;
release_buffered = WT_LOG_SLOT_RELEASED_BUFFERED(slot->slot_state);
release_bytes = release_buffered + slot->slot_unbuffered;
/*
* Checkpoints can be configured based on amount of log written.
* Add in this log record to the sum and if needed, signal the
* checkpoint condition. The logging subsystem manages the
* accumulated field. There is a bit of layering violation
* here checking the connection ckpt field and using its
* condition.
*/
if (WT_CKPT_LOGSIZE(conn)) {
log->log_written += (wt_off_t)release_bytes;
__wt_checkpoint_signal(session, log->log_written);
}
/* Write the buffered records */
if (release_buffered != 0)
WT_ERR(__log_fs_write(session, slot, slot->slot_start_offset,
(size_t)release_buffered, slot->slot_buf.mem));
/*
* If we have to wait for a synchronous operation, we do not pass
* handling of this slot off to the worker thread. The caller is
* responsible for freeing the slot in that case. Otherwise the
* worker thread will free it.
*/
if (!F_ISSET(slot, WT_SLOT_FLUSH | WT_SLOT_SYNC | WT_SLOT_SYNC_DIR)) {
if (freep != NULL)
*freep = 0;
slot->slot_state = WT_LOG_SLOT_WRITTEN;
/*
* After this point the worker thread owns the slot. There
* is nothing more to do but return.
*/
/*
* !!! Signalling the wrlsn_cond condition here results in
* worse performance because it causes more scheduling churn
* and more walking of the slot pool for a very small number
* of slots to process. Don't signal here.
*/
return (0);
}
/*
* Wait for earlier groups to finish, otherwise there could
* be holes in the log file.
*/
WT_STAT_CONN_INCR(session, log_release_write_lsn);
__log_wait_for_earlier_slot(session, slot);
log->write_start_lsn = slot->slot_start_lsn;
log->write_lsn = slot->slot_end_lsn;
WT_ASSERT(session, slot != log->active_slot);
__wt_cond_signal(session, log->log_write_cond);
F_CLR(slot, WT_SLOT_FLUSH);
/*
* Signal the close thread if needed.
*/
if (F_ISSET(slot, WT_SLOT_CLOSEFH))
__wt_cond_signal(session, conn->log_file_cond);
/*
* Try to consolidate calls to fsync to wait less. Acquire a spin lock
* so that threads finishing writing to the log will wait while the
* current fsync completes and advance log->sync_lsn.
*/
while (F_ISSET(slot, WT_SLOT_SYNC | WT_SLOT_SYNC_DIR)) {
/*
* We have to wait until earlier log files have finished their
* sync operations. The most recent one will set the LSN to the
* beginning of our file.
*/
if (log->sync_lsn.l.file < slot->slot_end_lsn.l.file ||
__wt_spin_trylock(session, &log->log_sync_lock) != 0) {
__wt_cond_wait(
session, log->log_sync_cond, 10000, NULL);
continue;
}
locked = true;
/*
* Record the current end of our update after the lock.
* That is how far our calls can guarantee.
*/
sync_lsn = slot->slot_end_lsn;
/*
* Check if we have to sync the parent directory. Some
* combinations of sync flags may result in the log file
* not yet stable in its parent directory. Do that
* now if needed.
*/
if (F_ISSET(slot, WT_SLOT_SYNC_DIR) &&
(log->sync_dir_lsn.l.file < sync_lsn.l.file)) {
WT_ASSERT(session, log->log_dir_fh != NULL);
__wt_verbose(session, WT_VERB_LOG,
"log_release: sync directory %s to LSN %" PRIu32
"/%" PRIu32,
log->log_dir_fh->name,
sync_lsn.l.file, sync_lsn.l.offset);
time_start = __wt_clock(session);
WT_ERR(__wt_fsync(session, log->log_dir_fh, true));
time_stop = __wt_clock(session);
fsync_duration_usecs =
WT_CLOCKDIFF_US(time_stop, time_start);
log->sync_dir_lsn = sync_lsn;
WT_STAT_CONN_INCR(session, log_sync_dir);
WT_STAT_CONN_INCRV(session,
log_sync_dir_duration, fsync_duration_usecs);
}
/*
* Sync the log file if needed.
*/
if (F_ISSET(slot, WT_SLOT_SYNC) &&
__wt_log_cmp(&log->sync_lsn, &slot->slot_end_lsn) < 0) {
__wt_verbose(session, WT_VERB_LOG,
"log_release: sync log %s to LSN %" PRIu32
"/%" PRIu32,
log->log_fh->name,
sync_lsn.l.file, sync_lsn.l.offset);
WT_STAT_CONN_INCR(session, log_sync);
time_start = __wt_clock(session);
WT_ERR(__wt_fsync(session, log->log_fh, true));
time_stop = __wt_clock(session);
fsync_duration_usecs =
WT_CLOCKDIFF_US(time_stop, time_start);
WT_STAT_CONN_INCRV(session,
log_sync_duration, fsync_duration_usecs);
log->sync_lsn = sync_lsn;
__wt_cond_signal(session, log->log_sync_cond);
}
/*
* Clear the flags before leaving the loop.
*/
F_CLR(slot, WT_SLOT_SYNC | WT_SLOT_SYNC_DIR);
locked = false;
__wt_spin_unlock(session, &log->log_sync_lock);
}
err: if (locked)
__wt_spin_unlock(session, &log->log_sync_lock);
if (ret != 0 && slot->slot_error == 0)
slot->slot_error = ret;
return (ret);
}
/*
* __wt_log_scan --
* Scan the logs, calling a function on each record found.
*/
int
__wt_log_scan(WT_SESSION_IMPL *session, WT_LSN *lsnp, uint32_t flags,
int (*func)(WT_SESSION_IMPL *session,
WT_ITEM *record, WT_LSN *lsnp, WT_LSN *next_lsnp,
void *cookie, int firstrecord), void *cookie)
{
WT_CONNECTION_IMPL *conn;
WT_DECL_ITEM(buf);
WT_DECL_ITEM(decryptitem);
WT_DECL_ITEM(uncitem);
WT_DECL_RET;
WT_FH *log_fh;
WT_ITEM *cbbuf;
WT_LOG *log;
WT_LOG_RECORD *logrec;
WT_LSN end_lsn, next_lsn, prev_eof, prev_lsn, rd_lsn, start_lsn;
wt_off_t log_size;
uint32_t allocsize, firstlog, lastlog, lognum, rdup_len, reclen;
uint16_t version;
u_int i, logcount;
int firstrecord;
char **logfiles;
bool eol, partial_record;
conn = S2C(session);
log = conn->log;
log_fh = NULL;
logcount = 0;
logfiles = NULL;
eol = false;
firstrecord = 1;
/*
* If the caller did not give us a callback function there is nothing
* to do.
*/
if (func == NULL)
return (0);
if (lsnp != NULL &&
LF_ISSET(WT_LOGSCAN_FIRST|WT_LOGSCAN_FROM_CKP))
WT_RET_MSG(session, WT_ERROR,
"choose either a start LSN or a start flag");
/*
* Set up the allocation size, starting and ending LSNs. The values
* for those depend on whether logging is currently enabled or not.
*/
lastlog = 0;
if (log != NULL) {
allocsize = log->allocsize;
end_lsn = log->alloc_lsn;
start_lsn = log->first_lsn;
if (lsnp == NULL) {
if (LF_ISSET(WT_LOGSCAN_FROM_CKP))
start_lsn = log->ckpt_lsn;
else if (!LF_ISSET(WT_LOGSCAN_FIRST))
return (WT_ERROR); /* Illegal usage */
}
lastlog = log->fileid;
} else {
/*
* If logging is not configured, we can still print out the log
* if log files exist. We just need to set the LSNs from what
* is in the files versus what is in the live connection.
*/
/*
* Set allocsize to the minimum alignment it could be. Larger
* records and larger allocation boundaries should always be
* a multiple of this.
*/
allocsize = WT_LOG_ALIGN;
firstlog = UINT32_MAX;
WT_RET(__log_get_files(session,
WT_LOG_FILENAME, &logfiles, &logcount));
if (logcount == 0)
WT_RET_MSG(session, ENOTSUP, "no log files found");
for (i = 0; i < logcount; i++) {
WT_ERR(__wt_log_extract_lognum(session, logfiles[i],
&lognum));
lastlog = WT_MAX(lastlog, lognum);
firstlog = WT_MIN(firstlog, lognum);
}
WT_SET_LSN(&start_lsn, firstlog, 0);
WT_SET_LSN(&end_lsn, lastlog, 0);
WT_ERR(
__wt_fs_directory_list_free(session, &logfiles, logcount));
}
if (lsnp != NULL) {
/*
* Offsets must be on allocation boundaries.
* An invalid LSN from a user should just return
* WT_NOTFOUND. It is not an error. But if it is
* from recovery, we expect valid LSNs so give more
* information about that.
*/
if (lsnp->l.offset % allocsize != 0) {
if (LF_ISSET(WT_LOGSCAN_RECOVER))
WT_ERR_MSG(session, WT_NOTFOUND,
"__wt_log_scan unaligned LSN %"
PRIu32 "/%" PRIu32,
lsnp->l.file, lsnp->l.offset);
else
return (WT_NOTFOUND);
}
/*
* If the file is in the future it doesn't exist.
* An invalid LSN from a user should just return
* WT_NOTFOUND. It is not an error. But if it is
* from recovery, we expect valid LSNs so give more
* information about that.
*/
if (lsnp->l.file > lastlog) {
if (LF_ISSET(WT_LOGSCAN_RECOVER))
WT_ERR_MSG(session, WT_NOTFOUND,
"__wt_log_scan LSN %" PRIu32 "/%" PRIu32
" larger than biggest log file %" PRIu32,
lsnp->l.file, lsnp->l.offset, lastlog);
else
return (WT_NOTFOUND);
}
/*
* Log cursors may not know the starting LSN. If an
* LSN is passed in that it is equal to the smallest
* LSN, start from the beginning of the log.
*/
if (!WT_IS_INIT_LSN(lsnp))
start_lsn = *lsnp;
}
WT_ERR(__log_open_verify(session,
start_lsn.l.file, &log_fh, &prev_lsn, NULL));
WT_ERR(__wt_filesize(session, log_fh, &log_size));
rd_lsn = start_lsn;
if (LF_ISSET(WT_LOGSCAN_RECOVER))
__wt_verbose(session, WT_VERB_RECOVERY_PROGRESS,
"Recovering log %" PRIu32 " through %" PRIu32,
rd_lsn.l.file, end_lsn.l.file);
WT_ERR(__wt_scr_alloc(session, WT_LOG_ALIGN, &buf));
WT_ERR(__wt_scr_alloc(session, 0, &decryptitem));
WT_ERR(__wt_scr_alloc(session, 0, &uncitem));
for (;;) {
if (rd_lsn.l.offset + allocsize > log_size) {
advance:
if (rd_lsn.l.offset == log_size)
partial_record = false;
else
/*
* See if there is anything non-zero at the
* end of this log file.
*/
WT_ERR(__log_has_hole(
session, log_fh, log_size,
rd_lsn.l.offset, &partial_record));
/*
* If we read the last record, go to the next file.
*/
WT_ERR(__wt_close(session, &log_fh));
log_fh = NULL;
eol = true;
/*
* Truncate this log file before we move to the next.
*/
if (LF_ISSET(WT_LOGSCAN_RECOVER) &&
__wt_log_cmp(&rd_lsn, &log->trunc_lsn) < 0) {
__wt_verbose(session, WT_VERB_LOG,
"Truncate end of log %" PRIu32 "/%" PRIu32,
rd_lsn.l.file, rd_lsn.l.offset);
WT_ERR(__log_truncate(session, &rd_lsn, true));
}
/*
* If we had a partial record, we'll want to break
* now after closing and truncating. Although for now
* log_truncate does not modify the LSN passed in,
* this code does not assume it is unmodified after that
* call which is why it uses the boolean set earlier.
*/
if (partial_record)
break;
/*
* Avoid an error message when we reach end of log
* by checking here.
*/
prev_eof = rd_lsn;
WT_SET_LSN(&rd_lsn, rd_lsn.l.file + 1, 0);
if (rd_lsn.l.file > end_lsn.l.file)
break;
if (LF_ISSET(WT_LOGSCAN_RECOVER))
__wt_verbose(session, WT_VERB_RECOVERY_PROGRESS,
"Recovering log %" PRIu32
" through %" PRIu32,
rd_lsn.l.file, end_lsn.l.file);
WT_ERR(__log_open_verify(session,
rd_lsn.l.file, &log_fh, &prev_lsn, &version));
/*
* Opening the log file reads with verify sets up the
* previous LSN from the first record. This detects
* a "hole" at the end of the previous log file.
*/
if (LF_ISSET(WT_LOGSCAN_RECOVER) &&
!WT_IS_INIT_LSN(&prev_lsn) &&
!WT_IS_ZERO_LSN(&prev_lsn) &&
prev_lsn.l.offset != prev_eof.l.offset) {
WT_ASSERT(session,
prev_eof.l.file == prev_lsn.l.file);
break;
}
/*
* If we read a current version log file without a
* previous LSN record the log ended after writing
* that header. We're done.
*/
if (LF_ISSET(WT_LOGSCAN_RECOVER) &&
version == WT_LOG_VERSION_SYSTEM &&
WT_IS_ZERO_LSN(&prev_lsn)) {
__wt_verbose(session, WT_VERB_LOG,
"log_scan: Stopping, no system "
"record detected in %s.", log_fh->name);
break;
}
WT_ERR(__wt_filesize(session, log_fh, &log_size));
eol = false;
continue;
}
/*
* Read the minimum allocation size a record could be.
*/
WT_ASSERT(session, buf->memsize >= allocsize);
WT_ERR(__wt_read(session,
log_fh, rd_lsn.l.offset, (size_t)allocsize, buf->mem));
/*
* See if we need to read more than the allocation size. We
* expect that we rarely will have to read more. Most log
* records will be fairly small.
*/
reclen = ((WT_LOG_RECORD *)buf->mem)->len;
#ifdef WORDS_BIGENDIAN
reclen = __wt_bswap32(reclen);
#endif
/*
* Log files are pre-allocated. We need to detect the
* difference between a hole in the file (where this location
* would be considered the end of log) and the last record
* in the log and we're at the zeroed part of the file.
* If we find a zeroed record, scan forward in the log looking
* for any data. If we detect any we have a hole and stop.
* Otherwise if the rest is all zeroes advance to the next file.
* When recovery finds the end of the log, truncate the file
* and remove any later log files that may exist.
*/
if (reclen == 0) {
WT_ERR(__log_has_hole(
session, log_fh, log_size, rd_lsn.l.offset, &eol));
if (eol)
/* Found a hole. This LSN is the end. */
break;
/* Last record in log. Look for more. */
goto advance;
}
rdup_len = __wt_rduppo2(reclen, allocsize);
if (reclen > allocsize) {
/*
* The log file end could be the middle of this
* log record. If we have a partially written record
* then this is considered the end of the log.
*/
if (rd_lsn.l.offset + rdup_len > log_size) {
eol = true;
break;
}
/*
* We need to round up and read in the full padded
* record, especially for direct I/O.
*/
WT_ERR(__wt_buf_grow(session, buf, rdup_len));
WT_ERR(__wt_read(session, log_fh,
rd_lsn.l.offset, (size_t)rdup_len, buf->mem));
WT_STAT_CONN_INCR(session, log_scan_rereads);
}
/*
* We read in the record, verify checksum.
*
* Handle little- and big-endian objects. Objects are written
* in little-endian format: save the header checksum, and
* calculate the checksum for the header in its little-endian
* form. Then, restore the header's checksum, and byte-swap
* the whole thing as necessary, leaving us with a calculated
* checksum that should match the checksum in the header.
*/
buf->size = reclen;
logrec = (WT_LOG_RECORD *)buf->mem;
if (!__log_checksum_match(session, buf, reclen)) {
/*
* A checksum mismatch means we have reached the end of
* the useful part of the log. This should be found on
* the first pass through recovery. In the second pass
* where we truncate the log, this is where it should
* end.
*/
if (log != NULL)
log->trunc_lsn = rd_lsn;
/*
* If the user asked for a specific LSN and it is not
* a valid LSN, return WT_NOTFOUND.
*/
if (LF_ISSET(WT_LOGSCAN_ONE))
ret = WT_NOTFOUND;
break;
}
__wt_log_record_byteswap(logrec);
/*
* We have a valid log record. If it is not the log file
* header, invoke the callback.
*/
WT_STAT_CONN_INCR(session, log_scan_records);
next_lsn = rd_lsn;
next_lsn.l.offset += rdup_len;
if (rd_lsn.l.offset != 0) {
/*
* We need to manage the different buffers here.
* Buf is the buffer this function uses to read from
* the disk. The callback buffer may change based
* on whether encryption and compression are used.
*
* We want to free any buffers from compression and
* encryption but keep the one we use for reading.
*/
cbbuf = buf;
if (F_ISSET(logrec, WT_LOG_RECORD_ENCRYPTED)) {
WT_ERR(__log_decrypt(
session, cbbuf, decryptitem));
cbbuf = decryptitem;
}
if (F_ISSET(logrec, WT_LOG_RECORD_COMPRESSED)) {
WT_ERR(__log_decompress(
session, cbbuf, uncitem));
cbbuf = uncitem;
}
WT_ERR((*func)(session,
cbbuf, &rd_lsn, &next_lsn, cookie, firstrecord));
firstrecord = 0;
if (LF_ISSET(WT_LOGSCAN_ONE))
break;
}
rd_lsn = next_lsn;
}
/* Truncate if we're in recovery. */
if (LF_ISSET(WT_LOGSCAN_RECOVER) &&
__wt_log_cmp(&rd_lsn, &log->trunc_lsn) < 0) {
__wt_verbose(session, WT_VERB_LOG,
"End of recovery truncate end of log %" PRIu32 "/%" PRIu32,
rd_lsn.l.file, rd_lsn.l.offset);
WT_ERR(__log_truncate(session, &rd_lsn, false));
}
err: WT_STAT_CONN_INCR(session, log_scans);
/*
* If the first attempt to read a log record results in
* an error recovery is likely going to fail. Try to provide
* a helpful failure message.
*/
if (ret != 0 && firstrecord) {
__wt_errx(session,
"WiredTiger is unable to read the recovery log.");
__wt_errx(session, "This may be due to the log"
" files being encrypted, being from an older"
" version or due to corruption on disk");
__wt_errx(session, "You should confirm that you have"
" opened the database with the correct options including"
" all encryption and compression options");
}
WT_TRET(__wt_fs_directory_list_free(session, &logfiles, logcount));
__wt_scr_free(session, &buf);
__wt_scr_free(session, &decryptitem);
__wt_scr_free(session, &uncitem);
/*
* If the caller wants one record and it is at the end of log,
* return WT_NOTFOUND.
*/
if (LF_ISSET(WT_LOGSCAN_ONE) && eol && ret == 0)
ret = WT_NOTFOUND;
WT_TRET(__wt_close(session, &log_fh));
return (ret);
}
/*
* __wt_log_force_write --
* Force a switch and release and write of the current slot.
* Wrapper function that takes the lock.
*/
int
__wt_log_force_write(WT_SESSION_IMPL *session, bool retry, bool *did_work)
{
WT_LOG *log;
WT_MYSLOT myslot;
log = S2C(session)->log;
memset(&myslot, 0, sizeof(myslot));
WT_STAT_CONN_INCR(session, log_force_write);
if (did_work != NULL)
*did_work = true;
myslot.slot = log->active_slot;
return (__wt_log_slot_switch(session, &myslot, retry, true, did_work));
}
/*
* __wt_log_write --
* Write a record into the log, compressing as necessary.
*/
int
__wt_log_write(WT_SESSION_IMPL *session, WT_ITEM *record, WT_LSN *lsnp,
uint32_t flags)
{
WT_COMPRESSOR *compressor;
WT_CONNECTION_IMPL *conn;
WT_DECL_ITEM(citem);
WT_DECL_ITEM(eitem);
WT_DECL_RET;
WT_ITEM *ip;
WT_KEYED_ENCRYPTOR *kencryptor;
WT_LOG *log;
WT_LOG_RECORD *newlrp;
size_t dst_len, len, new_size, result_len, src_len;
uint8_t *dst, *src;
int compression_failed;
conn = S2C(session);
log = conn->log;
/*
* An error during opening the logging subsystem can result in it
* being enabled, but without an open log file. In that case,
* just return. We can also have logging opened for reading in a
* read-only database and attempt to write a record on close.
*/
if (!F_ISSET(log, WT_LOG_OPENED) || F_ISSET(conn, WT_CONN_READONLY))
return (0);
ip = record;
if ((compressor = conn->log_compressor) != NULL &&
record->size < log->allocsize) {
WT_STAT_CONN_INCR(session, log_compress_small);
} else if (compressor != NULL) {
/* Skip the log header */
src = (uint8_t *)record->mem + WT_LOG_COMPRESS_SKIP;
src_len = record->size - WT_LOG_COMPRESS_SKIP;
/*
* Compute the size needed for the destination buffer. We only
* allocate enough memory for a copy of the original by default,
* if any compressed version is bigger than the original, we
* won't use it. However, some compression engines (snappy is
* one example), may need more memory because they don't stop
* just because there's no more memory into which to compress.
*/
if (compressor->pre_size == NULL)
len = src_len;
else
WT_ERR(compressor->pre_size(compressor,
&session->iface, src, src_len, &len));
new_size = len + WT_LOG_COMPRESS_SKIP;
WT_ERR(__wt_scr_alloc(session, new_size, &citem));
/* Skip the header bytes of the destination data. */
dst = (uint8_t *)citem->mem + WT_LOG_COMPRESS_SKIP;
dst_len = len;
compression_failed = 0;
WT_ERR(compressor->compress(compressor, &session->iface,
src, src_len, dst, dst_len, &result_len,
&compression_failed));
result_len += WT_LOG_COMPRESS_SKIP;
/*
* If compression fails, or doesn't gain us at least one unit of
* allocation, fallback to the original version. This isn't
* unexpected: if compression doesn't work for some chunk of
* data for some reason (noting likely additional format/header
* information which compressed output requires), it just means
* the uncompressed version is as good as it gets, and that's
* what we use.
*/
if (compression_failed ||
result_len / log->allocsize >=
record->size / log->allocsize)
WT_STAT_CONN_INCR(session, log_compress_write_fails);
else {
WT_STAT_CONN_INCR(session, log_compress_writes);
WT_STAT_CONN_INCRV(session, log_compress_mem,
record->size);
WT_STAT_CONN_INCRV(session, log_compress_len,
result_len);
/*
* Copy in the skipped header bytes, set the final data
* size.
*/
memcpy(citem->mem, record->mem, WT_LOG_COMPRESS_SKIP);
citem->size = result_len;
ip = citem;
newlrp = (WT_LOG_RECORD *)citem->mem;
F_SET(newlrp, WT_LOG_RECORD_COMPRESSED);
WT_ASSERT(session, result_len < UINT32_MAX &&
record->size < UINT32_MAX);
newlrp->mem_len = WT_STORE_SIZE(record->size);
}
}
if ((kencryptor = conn->kencryptor) != NULL) {
/*
* Allocate enough space for the original record plus the
* encryption size constant plus the length we store.
*/
__wt_encrypt_size(session, kencryptor, ip->size, &new_size);
WT_ERR(__wt_scr_alloc(session, new_size, &eitem));
WT_ERR(__wt_encrypt(session, kencryptor,
WT_LOG_ENCRYPT_SKIP, ip, eitem));
/*
* Final setup of new buffer. Set the flag for
* encryption in the record header.
*/
ip = eitem;
newlrp = (WT_LOG_RECORD *)eitem->mem;
F_SET(newlrp, WT_LOG_RECORD_ENCRYPTED);
WT_ASSERT(session, new_size < UINT32_MAX &&
ip->size < UINT32_MAX);
}
ret = __log_write_internal(session, ip, lsnp, flags);
err: __wt_scr_free(session, &citem);
__wt_scr_free(session, &eitem);
return (ret);
}
/*
* __log_write_internal --
* Write a record into the log.
*/
static int
__log_write_internal(WT_SESSION_IMPL *session, WT_ITEM *record, WT_LSN *lsnp,
uint32_t flags)
{
WT_CONNECTION_IMPL *conn;
WT_DECL_RET;
WT_LOG *log;
WT_LOG_RECORD *logrec;
WT_LSN lsn;
WT_MYSLOT myslot;
int64_t release_size;
uint32_t force, rdup_len;
bool free_slot;
conn = S2C(session);
log = conn->log;
if (record->size > UINT32_MAX)
WT_RET_MSG(session, EFBIG,
"Log record size of %" WT_SIZET_FMT " exceeds the maximum "
"supported size of %" PRIu32,
record->size, UINT32_MAX);
WT_INIT_LSN(&lsn);
myslot.slot = NULL;
memset(&myslot, 0, sizeof(myslot));
/*
* Assume the WT_ITEM the caller passed is a WT_LOG_RECORD, which has a
* header at the beginning for us to fill in.
*
* If using direct_io, the caller should pass us an aligned record.
* But we need to make sure it is big enough and zero-filled so
* that we can write the full amount. Do this whether or not
* direct_io is in use because it makes the reading code cleaner.
*/
WT_STAT_CONN_INCRV(session, log_bytes_payload, record->size);
rdup_len = __wt_rduppo2((uint32_t)record->size, log->allocsize);
WT_ERR(__wt_buf_grow(session, record, rdup_len));
WT_ASSERT(session, record->data == record->mem);
/*
* If the caller's record only partially fills the necessary
* space, we need to zero-fill the remainder.
*/
if (record->size != rdup_len) {
memset((uint8_t *)record->mem + record->size, 0,
rdup_len - record->size);
record->size = rdup_len;
}
/*
* Checksum a little-endian version of the header, and write everything
* in little-endian format. The checksum is (potentially) returned in a
* big-endian format, swap it into place in a separate step.
*/
logrec = (WT_LOG_RECORD *)record->mem;
logrec->len = (uint32_t)record->size;
logrec->checksum = 0;
__wt_log_record_byteswap(logrec);
logrec->checksum = __wt_checksum(logrec, record->size);
#ifdef WORDS_BIGENDIAN
logrec->checksum = __wt_bswap32(logrec->checksum);
#endif
WT_STAT_CONN_INCR(session, log_writes);
/*
* The only time joining a slot should ever return an error is if it
* detects a panic.
*/
__wt_log_slot_join(session, rdup_len, flags, &myslot);
/*
* If the addition of this record crosses the buffer boundary,
* switch in a new slot.
*/
force = LF_ISSET(WT_LOG_FLUSH | WT_LOG_FSYNC);
ret = 0;
if (myslot.end_offset >= WT_LOG_SLOT_BUF_MAX ||
F_ISSET(&myslot, WT_MYSLOT_UNBUFFERED) || force)
ret = __wt_log_slot_switch(session, &myslot, true, false, NULL);
if (ret == 0)
ret = __wt_log_fill(session, &myslot, false, record, &lsn);
release_size = __wt_log_slot_release(&myslot, (int64_t)rdup_len);
/*
* If we get an error we still need to do proper accounting in
* the slot fields.
* XXX On error we may still need to call release and free.
*/
if (ret != 0)
myslot.slot->slot_error = ret;
WT_ASSERT(session, ret == 0);
if (WT_LOG_SLOT_DONE(release_size)) {
WT_ERR(__wt_log_release(session, myslot.slot, &free_slot));
if (free_slot)
__wt_log_slot_free(session, myslot.slot);
} else if (force) {
/*
* If we are going to wait for this slot to get written,
* signal the wrlsn thread.
*
* XXX I've seen times when conditions are NULL.
*/
if (conn->log_cond != NULL) {
__wt_cond_signal(session, conn->log_cond);
__wt_yield();
} else
WT_ERR(__wt_log_force_write(session, 1, NULL));
}
if (LF_ISSET(WT_LOG_FLUSH)) {
/* Wait for our writes to reach the OS */
while (__wt_log_cmp(&log->write_lsn, &lsn) <= 0 &&
myslot.slot->slot_error == 0)
__wt_cond_wait(
session, log->log_write_cond, 10000, NULL);
} else if (LF_ISSET(WT_LOG_FSYNC)) {
/* Wait for our writes to reach disk */
while (__wt_log_cmp(&log->sync_lsn, &lsn) <= 0 &&
myslot.slot->slot_error == 0)
__wt_cond_wait(
session, log->log_sync_cond, 10000, NULL);
}
/*
* Advance the background sync LSN if needed.
*/
if (LF_ISSET(WT_LOG_BACKGROUND))
__wt_log_background(session, &lsn);
err:
if (ret == 0 && lsnp != NULL)
*lsnp = lsn;
/*
* If we're synchronous and some thread had an error, we don't know
* if our write made it out to the file or not. The error could be
* before or after us. So, if anyone got an error, we report it.
* If we're not synchronous, only report if our own operation got
* an error.
*/
if (LF_ISSET(WT_LOG_DSYNC | WT_LOG_FSYNC) && ret == 0 &&
myslot.slot != NULL)
ret = myslot.slot->slot_error;
/*
* If one of the sync flags is set, assert the proper LSN has moved to
* match on success.
*/
WT_ASSERT(session, ret != 0 || !LF_ISSET(WT_LOG_FLUSH) ||
__wt_log_cmp(&log->write_lsn, &lsn) >= 0);
WT_ASSERT(session, ret != 0 || !LF_ISSET(WT_LOG_FSYNC) ||
__wt_log_cmp(&log->sync_lsn, &lsn) >= 0);
return (ret);
}
/*
* __wt_log_vprintf --
* Write a message into the log.
*/
int
__wt_log_vprintf(WT_SESSION_IMPL *session, const char *fmt, va_list ap)
{
WT_CONNECTION_IMPL *conn;
WT_DECL_ITEM(logrec);
WT_DECL_RET;
size_t header_size, len;
uint32_t rectype;
const char *rec_fmt;
va_list ap_copy;
conn = S2C(session);
rectype = WT_LOGREC_MESSAGE;
rec_fmt = WT_UNCHECKED_STRING(I);
if (!FLD_ISSET(conn->log_flags, WT_CONN_LOG_ENABLED))
return (0);
va_copy(ap_copy, ap);
len = 1;
ret = __wt_vsnprintf_len_incr(NULL, 0, &len, fmt, ap_copy);
va_end(ap_copy);
WT_RET(ret);
WT_RET(
__wt_logrec_alloc(session, sizeof(WT_LOG_RECORD) + len, &logrec));
/*
* We're writing a record with the type (an integer) followed by a
* string (NUL-terminated data). To avoid writing the string into
* a buffer before copying it, we write the header first, then the
* raw bytes of the string.
*/
WT_ERR(__wt_struct_size(session, &header_size, rec_fmt, rectype));
WT_ERR(__wt_struct_pack(session,
(uint8_t *)logrec->data + logrec->size, header_size,
rec_fmt, rectype));
logrec->size += (uint32_t)header_size;
WT_ERR(__wt_vsnprintf(
(char *)logrec->data + logrec->size, len, fmt, ap));
__wt_verbose(session, WT_VERB_LOG,
"log_printf: %s", (char *)logrec->data + logrec->size);
logrec->size += len;
WT_ERR(__wt_log_write(session, logrec, NULL, 0));
err: __wt_scr_free(session, &logrec);
return (ret);
}
/*
* __wt_log_flush --
* Forcibly flush the log to the synchronization level specified.
* Wait until it has been completed.
*/
int
__wt_log_flush(WT_SESSION_IMPL *session, uint32_t flags)
{
WT_CONNECTION_IMPL *conn;
WT_LOG *log;
WT_LSN last_lsn, lsn;
conn = S2C(session);
WT_ASSERT(session, FLD_ISSET(conn->log_flags, WT_CONN_LOG_ENABLED));
log = conn->log;
/*
* We need to flush out the current slot first to get the real
* end of log LSN in log->alloc_lsn.
*/
WT_RET(__wt_log_flush_lsn(session, &lsn, false));
last_lsn = log->alloc_lsn;
/*
* If the last write caused a switch to a new log file, we should only
* wait for the last write to be flushed. Otherwise, if the workload
* is single-threaded we could wait here forever because the write LSN
* doesn't switch into the new file until it contains a record.
*/
if (last_lsn.l.offset == log->first_record)
last_lsn = log->log_close_lsn;
/*
* Wait until all current outstanding writes have been written
* to the file system.
*/
while (__wt_log_cmp(&last_lsn, &lsn) > 0) {
__wt_sleep(0, WT_THOUSAND);
WT_RET(__wt_log_flush_lsn(session, &lsn, false));
}
__wt_verbose(session, WT_VERB_LOG,
"log_flush: flags %#" PRIx32 " LSN %" PRIu32 "/%" PRIu32,
flags, lsn.l.file, lsn.l.offset);
/*
* If the user wants write-no-sync, there is nothing more to do.
* If the user wants background sync, set the LSN and we're done.
* If the user wants sync, force it now.
*/
if (LF_ISSET(WT_LOG_BACKGROUND))
__wt_log_background(session, &lsn);
else if (LF_ISSET(WT_LOG_FSYNC))
WT_RET(__wt_log_force_sync(session, &lsn));
return (0);
}
|