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
|
/*-
* Copyright (c) 2004 Poul-Henning Kamp
* Copyright (c) 1994,1997 John S. Dyson
* All rights reserved.
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions
* are met:
* 1. Redistributions of source code must retain the above copyright
* notice, this list of conditions and the following disclaimer.
* 2. Redistributions in binary form must reproduce the above copyright
* notice, this list of conditions and the following disclaimer in the
* documentation and/or other materials provided with the distribution.
*
* THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
* ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
* IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
* ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
* FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
* DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
* OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
* HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
* LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
* OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
* SUCH DAMAGE.
*/
/*
* this file contains a new buffer I/O scheme implementing a coherent
* VM object and buffer cache scheme. Pains have been taken to make
* sure that the performance degradation associated with schemes such
* as this is not realized.
*
* Author: John S. Dyson
* Significant help during the development and debugging phases
* had been provided by David Greenman, also of the FreeBSD core team.
*
* see man buf(9) for more info.
*/
#include <sys/cdefs.h>
__FBSDID("$FreeBSD$");
#include <sys/param.h>
#include <sys/systm.h>
#include <sys/bio.h>
#include <sys/conf.h>
#include <sys/buf.h>
#include <sys/devicestat.h>
#include <sys/eventhandler.h>
#include <sys/fail.h>
#include <sys/limits.h>
#include <sys/lock.h>
#include <sys/malloc.h>
#include <sys/mount.h>
#include <sys/mutex.h>
#include <sys/kernel.h>
#include <sys/kthread.h>
#include <sys/proc.h>
#include <sys/resourcevar.h>
#include <sys/sysctl.h>
#include <sys/vmmeter.h>
#include <sys/vnode.h>
#include <geom/geom.h>
#include <vm/vm.h>
#include <vm/vm_param.h>
#include <vm/vm_kern.h>
#include <vm/vm_pageout.h>
#include <vm/vm_page.h>
#include <vm/vm_object.h>
#include <vm/vm_extern.h>
#include <vm/vm_map.h>
#include "opt_compat.h"
#include "opt_directio.h"
#include "opt_swap.h"
static MALLOC_DEFINE(M_BIOBUF, "biobuf", "BIO buffer");
struct bio_ops bioops; /* I/O operation notification */
struct buf_ops buf_ops_bio = {
.bop_name = "buf_ops_bio",
.bop_write = bufwrite,
.bop_strategy = bufstrategy,
.bop_sync = bufsync,
.bop_bdflush = bufbdflush,
};
/*
* XXX buf is global because kern_shutdown.c and ffs_checkoverlap has
* carnal knowledge of buffers. This knowledge should be moved to vfs_bio.c.
*/
struct buf *buf; /* buffer header pool */
static struct proc *bufdaemonproc;
static int inmem(struct vnode *vp, daddr_t blkno);
static void vm_hold_free_pages(struct buf *bp, int newbsize);
static void vm_hold_load_pages(struct buf *bp, vm_offset_t from,
vm_offset_t to);
static void vfs_page_set_valid(struct buf *bp, vm_ooffset_t off, vm_page_t m);
static void vfs_page_set_validclean(struct buf *bp, vm_ooffset_t off,
vm_page_t m);
static void vfs_drain_busy_pages(struct buf *bp);
static void vfs_clean_pages_dirty_buf(struct buf *bp);
static void vfs_setdirty_locked_object(struct buf *bp);
static void vfs_vmio_release(struct buf *bp);
static int vfs_bio_clcheck(struct vnode *vp, int size,
daddr_t lblkno, daddr_t blkno);
static int buf_do_flush(struct vnode *vp);
static int flushbufqueues(struct vnode *, int, int);
static void buf_daemon(void);
static void bremfreel(struct buf *bp);
#if defined(COMPAT_FREEBSD4) || defined(COMPAT_FREEBSD5) || \
defined(COMPAT_FREEBSD6) || defined(COMPAT_FREEBSD7)
static int sysctl_bufspace(SYSCTL_HANDLER_ARGS);
#endif
int vmiodirenable = TRUE;
SYSCTL_INT(_vfs, OID_AUTO, vmiodirenable, CTLFLAG_RW, &vmiodirenable, 0,
"Use the VM system for directory writes");
long runningbufspace;
SYSCTL_LONG(_vfs, OID_AUTO, runningbufspace, CTLFLAG_RD, &runningbufspace, 0,
"Amount of presently outstanding async buffer io");
static long bufspace;
#if defined(COMPAT_FREEBSD4) || defined(COMPAT_FREEBSD5) || \
defined(COMPAT_FREEBSD6) || defined(COMPAT_FREEBSD7)
SYSCTL_PROC(_vfs, OID_AUTO, bufspace, CTLTYPE_LONG|CTLFLAG_MPSAFE|CTLFLAG_RD,
&bufspace, 0, sysctl_bufspace, "L", "Virtual memory used for buffers");
#else
SYSCTL_LONG(_vfs, OID_AUTO, bufspace, CTLFLAG_RD, &bufspace, 0,
"Virtual memory used for buffers");
#endif
static long maxbufspace;
SYSCTL_LONG(_vfs, OID_AUTO, maxbufspace, CTLFLAG_RD, &maxbufspace, 0,
"Maximum allowed value of bufspace (including buf_daemon)");
static long bufmallocspace;
SYSCTL_LONG(_vfs, OID_AUTO, bufmallocspace, CTLFLAG_RD, &bufmallocspace, 0,
"Amount of malloced memory for buffers");
static long maxbufmallocspace;
SYSCTL_LONG(_vfs, OID_AUTO, maxmallocbufspace, CTLFLAG_RW, &maxbufmallocspace, 0,
"Maximum amount of malloced memory for buffers");
static long lobufspace;
SYSCTL_LONG(_vfs, OID_AUTO, lobufspace, CTLFLAG_RD, &lobufspace, 0,
"Minimum amount of buffers we want to have");
long hibufspace;
SYSCTL_LONG(_vfs, OID_AUTO, hibufspace, CTLFLAG_RD, &hibufspace, 0,
"Maximum allowed value of bufspace (excluding buf_daemon)");
static int bufreusecnt;
SYSCTL_INT(_vfs, OID_AUTO, bufreusecnt, CTLFLAG_RW, &bufreusecnt, 0,
"Number of times we have reused a buffer");
static int buffreekvacnt;
SYSCTL_INT(_vfs, OID_AUTO, buffreekvacnt, CTLFLAG_RW, &buffreekvacnt, 0,
"Number of times we have freed the KVA space from some buffer");
static int bufdefragcnt;
SYSCTL_INT(_vfs, OID_AUTO, bufdefragcnt, CTLFLAG_RW, &bufdefragcnt, 0,
"Number of times we have had to repeat buffer allocation to defragment");
static long lorunningspace;
SYSCTL_LONG(_vfs, OID_AUTO, lorunningspace, CTLFLAG_RW, &lorunningspace, 0,
"Minimum preferred space used for in-progress I/O");
static long hirunningspace;
SYSCTL_LONG(_vfs, OID_AUTO, hirunningspace, CTLFLAG_RW, &hirunningspace, 0,
"Maximum amount of space to use for in-progress I/O");
int dirtybufferflushes;
SYSCTL_INT(_vfs, OID_AUTO, dirtybufferflushes, CTLFLAG_RW, &dirtybufferflushes,
0, "Number of bdwrite to bawrite conversions to limit dirty buffers");
int bdwriteskip;
SYSCTL_INT(_vfs, OID_AUTO, bdwriteskip, CTLFLAG_RW, &bdwriteskip,
0, "Number of buffers supplied to bdwrite with snapshot deadlock risk");
int altbufferflushes;
SYSCTL_INT(_vfs, OID_AUTO, altbufferflushes, CTLFLAG_RW, &altbufferflushes,
0, "Number of fsync flushes to limit dirty buffers");
static int recursiveflushes;
SYSCTL_INT(_vfs, OID_AUTO, recursiveflushes, CTLFLAG_RW, &recursiveflushes,
0, "Number of flushes skipped due to being recursive");
static int numdirtybuffers;
SYSCTL_INT(_vfs, OID_AUTO, numdirtybuffers, CTLFLAG_RD, &numdirtybuffers, 0,
"Number of buffers that are dirty (has unwritten changes) at the moment");
static int lodirtybuffers;
SYSCTL_INT(_vfs, OID_AUTO, lodirtybuffers, CTLFLAG_RW, &lodirtybuffers, 0,
"How many buffers we want to have free before bufdaemon can sleep");
static int hidirtybuffers;
SYSCTL_INT(_vfs, OID_AUTO, hidirtybuffers, CTLFLAG_RW, &hidirtybuffers, 0,
"When the number of dirty buffers is considered severe");
int dirtybufthresh;
SYSCTL_INT(_vfs, OID_AUTO, dirtybufthresh, CTLFLAG_RW, &dirtybufthresh,
0, "Number of bdwrite to bawrite conversions to clear dirty buffers");
static int numfreebuffers;
SYSCTL_INT(_vfs, OID_AUTO, numfreebuffers, CTLFLAG_RD, &numfreebuffers, 0,
"Number of free buffers");
static int lofreebuffers;
SYSCTL_INT(_vfs, OID_AUTO, lofreebuffers, CTLFLAG_RW, &lofreebuffers, 0,
"XXX Unused");
static int hifreebuffers;
SYSCTL_INT(_vfs, OID_AUTO, hifreebuffers, CTLFLAG_RW, &hifreebuffers, 0,
"XXX Complicatedly unused");
static int getnewbufcalls;
SYSCTL_INT(_vfs, OID_AUTO, getnewbufcalls, CTLFLAG_RW, &getnewbufcalls, 0,
"Number of calls to getnewbuf");
static int getnewbufrestarts;
SYSCTL_INT(_vfs, OID_AUTO, getnewbufrestarts, CTLFLAG_RW, &getnewbufrestarts, 0,
"Number of times getnewbuf has had to restart a buffer aquisition");
static int flushbufqtarget = 100;
SYSCTL_INT(_vfs, OID_AUTO, flushbufqtarget, CTLFLAG_RW, &flushbufqtarget, 0,
"Amount of work to do in flushbufqueues when helping bufdaemon");
static long notbufdflashes;
SYSCTL_LONG(_vfs, OID_AUTO, notbufdflashes, CTLFLAG_RD, ¬bufdflashes, 0,
"Number of dirty buffer flushes done by the bufdaemon helpers");
/*
* Wakeup point for bufdaemon, as well as indicator of whether it is already
* active. Set to 1 when the bufdaemon is already "on" the queue, 0 when it
* is idling.
*/
static int bd_request;
/*
* Request for the buf daemon to write more buffers than is indicated by
* lodirtybuf. This may be necessary to push out excess dependencies or
* defragment the address space where a simple count of the number of dirty
* buffers is insufficient to characterize the demand for flushing them.
*/
static int bd_speedupreq;
/*
* This lock synchronizes access to bd_request.
*/
static struct mtx bdlock;
/*
* bogus page -- for I/O to/from partially complete buffers
* this is a temporary solution to the problem, but it is not
* really that bad. it would be better to split the buffer
* for input in the case of buffers partially already in memory,
* but the code is intricate enough already.
*/
vm_page_t bogus_page;
/*
* Synchronization (sleep/wakeup) variable for active buffer space requests.
* Set when wait starts, cleared prior to wakeup().
* Used in runningbufwakeup() and waitrunningbufspace().
*/
static int runningbufreq;
/*
* This lock protects the runningbufreq and synchronizes runningbufwakeup and
* waitrunningbufspace().
*/
static struct mtx rbreqlock;
/*
* Synchronization (sleep/wakeup) variable for buffer requests.
* Can contain the VFS_BIO_NEED flags defined below; setting/clearing is done
* by and/or.
* Used in numdirtywakeup(), bufspacewakeup(), bufcountwakeup(), bwillwrite(),
* getnewbuf(), and getblk().
*/
static int needsbuffer;
/*
* Lock that protects needsbuffer and the sleeps/wakeups surrounding it.
*/
static struct mtx nblock;
/*
* Definitions for the buffer free lists.
*/
#define BUFFER_QUEUES 6 /* number of free buffer queues */
#define QUEUE_NONE 0 /* on no queue */
#define QUEUE_CLEAN 1 /* non-B_DELWRI buffers */
#define QUEUE_DIRTY 2 /* B_DELWRI buffers */
#define QUEUE_DIRTY_GIANT 3 /* B_DELWRI buffers that need giant */
#define QUEUE_EMPTYKVA 4 /* empty buffer headers w/KVA assignment */
#define QUEUE_EMPTY 5 /* empty buffer headers */
#define QUEUE_SENTINEL 1024 /* not an queue index, but mark for sentinel */
/* Queues for free buffers with various properties */
static TAILQ_HEAD(bqueues, buf) bufqueues[BUFFER_QUEUES] = { { 0 } };
/* Lock for the bufqueues */
static struct mtx bqlock;
/*
* Single global constant for BUF_WMESG, to avoid getting multiple references.
* buf_wmesg is referred from macros.
*/
const char *buf_wmesg = BUF_WMESG;
#define VFS_BIO_NEED_ANY 0x01 /* any freeable buffer */
#define VFS_BIO_NEED_DIRTYFLUSH 0x02 /* waiting for dirty buffer flush */
#define VFS_BIO_NEED_FREE 0x04 /* wait for free bufs, hi hysteresis */
#define VFS_BIO_NEED_BUFSPACE 0x08 /* wait for buf space, lo hysteresis */
#if defined(COMPAT_FREEBSD4) || defined(COMPAT_FREEBSD5) || \
defined(COMPAT_FREEBSD6) || defined(COMPAT_FREEBSD7)
static int
sysctl_bufspace(SYSCTL_HANDLER_ARGS)
{
long lvalue;
int ivalue;
if (sizeof(int) == sizeof(long) || req->oldlen >= sizeof(long))
return (sysctl_handle_long(oidp, arg1, arg2, req));
lvalue = *(long *)arg1;
if (lvalue > INT_MAX)
/* On overflow, still write out a long to trigger ENOMEM. */
return (sysctl_handle_long(oidp, &lvalue, 0, req));
ivalue = lvalue;
return (sysctl_handle_int(oidp, &ivalue, 0, req));
}
#endif
#ifdef DIRECTIO
extern void ffs_rawread_setup(void);
#endif /* DIRECTIO */
/*
* numdirtywakeup:
*
* If someone is blocked due to there being too many dirty buffers,
* and numdirtybuffers is now reasonable, wake them up.
*/
static __inline void
numdirtywakeup(int level)
{
if (numdirtybuffers <= level) {
mtx_lock(&nblock);
if (needsbuffer & VFS_BIO_NEED_DIRTYFLUSH) {
needsbuffer &= ~VFS_BIO_NEED_DIRTYFLUSH;
wakeup(&needsbuffer);
}
mtx_unlock(&nblock);
}
}
/*
* bufspacewakeup:
*
* Called when buffer space is potentially available for recovery.
* getnewbuf() will block on this flag when it is unable to free
* sufficient buffer space. Buffer space becomes recoverable when
* bp's get placed back in the queues.
*/
static __inline void
bufspacewakeup(void)
{
/*
* If someone is waiting for BUF space, wake them up. Even
* though we haven't freed the kva space yet, the waiting
* process will be able to now.
*/
mtx_lock(&nblock);
if (needsbuffer & VFS_BIO_NEED_BUFSPACE) {
needsbuffer &= ~VFS_BIO_NEED_BUFSPACE;
wakeup(&needsbuffer);
}
mtx_unlock(&nblock);
}
/*
* runningbufwakeup() - in-progress I/O accounting.
*
*/
void
runningbufwakeup(struct buf *bp)
{
if (bp->b_runningbufspace) {
atomic_subtract_long(&runningbufspace, bp->b_runningbufspace);
bp->b_runningbufspace = 0;
mtx_lock(&rbreqlock);
if (runningbufreq && runningbufspace <= lorunningspace) {
runningbufreq = 0;
wakeup(&runningbufreq);
}
mtx_unlock(&rbreqlock);
}
}
/*
* bufcountwakeup:
*
* Called when a buffer has been added to one of the free queues to
* account for the buffer and to wakeup anyone waiting for free buffers.
* This typically occurs when large amounts of metadata are being handled
* by the buffer cache ( else buffer space runs out first, usually ).
*/
static __inline void
bufcountwakeup(struct buf *bp)
{
int old;
KASSERT((bp->b_vflags & BV_INFREECNT) == 0,
("buf %p already counted as free", bp));
if (bp->b_bufobj != NULL)
mtx_assert(BO_MTX(bp->b_bufobj), MA_OWNED);
bp->b_vflags |= BV_INFREECNT;
old = atomic_fetchadd_int(&numfreebuffers, 1);
KASSERT(old >= 0 && old < nbuf,
("numfreebuffers climbed to %d", old + 1));
mtx_lock(&nblock);
if (needsbuffer) {
needsbuffer &= ~VFS_BIO_NEED_ANY;
if (numfreebuffers >= hifreebuffers)
needsbuffer &= ~VFS_BIO_NEED_FREE;
wakeup(&needsbuffer);
}
mtx_unlock(&nblock);
}
/*
* waitrunningbufspace()
*
* runningbufspace is a measure of the amount of I/O currently
* running. This routine is used in async-write situations to
* prevent creating huge backups of pending writes to a device.
* Only asynchronous writes are governed by this function.
*
* Reads will adjust runningbufspace, but will not block based on it.
* The read load has a side effect of reducing the allowed write load.
*
* This does NOT turn an async write into a sync write. It waits
* for earlier writes to complete and generally returns before the
* caller's write has reached the device.
*/
void
waitrunningbufspace(void)
{
mtx_lock(&rbreqlock);
while (runningbufspace > hirunningspace) {
++runningbufreq;
msleep(&runningbufreq, &rbreqlock, PVM, "wdrain", 0);
}
mtx_unlock(&rbreqlock);
}
/*
* vfs_buf_test_cache:
*
* Called when a buffer is extended. This function clears the B_CACHE
* bit if the newly extended portion of the buffer does not contain
* valid data.
*/
static __inline
void
vfs_buf_test_cache(struct buf *bp,
vm_ooffset_t foff, vm_offset_t off, vm_offset_t size,
vm_page_t m)
{
VM_OBJECT_LOCK_ASSERT(m->object, MA_OWNED);
if (bp->b_flags & B_CACHE) {
int base = (foff + off) & PAGE_MASK;
if (vm_page_is_valid(m, base, size) == 0)
bp->b_flags &= ~B_CACHE;
}
}
/* Wake up the buffer daemon if necessary */
static __inline
void
bd_wakeup(int dirtybuflevel)
{
mtx_lock(&bdlock);
if (bd_request == 0 && numdirtybuffers >= dirtybuflevel) {
bd_request = 1;
wakeup(&bd_request);
}
mtx_unlock(&bdlock);
}
/*
* bd_speedup - speedup the buffer cache flushing code
*/
void
bd_speedup(void)
{
int needwake;
mtx_lock(&bdlock);
needwake = 0;
if (bd_speedupreq == 0 || bd_request == 0)
needwake = 1;
bd_speedupreq = 1;
bd_request = 1;
if (needwake)
wakeup(&bd_request);
mtx_unlock(&bdlock);
}
/*
* Calculating buffer cache scaling values and reserve space for buffer
* headers. This is called during low level kernel initialization and
* may be called more then once. We CANNOT write to the memory area
* being reserved at this time.
*/
caddr_t
kern_vfs_bio_buffer_alloc(caddr_t v, long physmem_est)
{
int tuned_nbuf;
long maxbuf;
/*
* physmem_est is in pages. Convert it to kilobytes (assumes
* PAGE_SIZE is >= 1K)
*/
physmem_est = physmem_est * (PAGE_SIZE / 1024);
/*
* The nominal buffer size (and minimum KVA allocation) is BKVASIZE.
* For the first 64MB of ram nominally allocate sufficient buffers to
* cover 1/4 of our ram. Beyond the first 64MB allocate additional
* buffers to cover 1/10 of our ram over 64MB. When auto-sizing
* the buffer cache we limit the eventual kva reservation to
* maxbcache bytes.
*
* factor represents the 1/4 x ram conversion.
*/
if (nbuf == 0) {
int factor = 4 * BKVASIZE / 1024;
nbuf = 50;
if (physmem_est > 4096)
nbuf += min((physmem_est - 4096) / factor,
65536 / factor);
if (physmem_est > 65536)
nbuf += (physmem_est - 65536) * 2 / (factor * 5);
if (maxbcache && nbuf > maxbcache / BKVASIZE)
nbuf = maxbcache / BKVASIZE;
tuned_nbuf = 1;
} else
tuned_nbuf = 0;
/* XXX Avoid unsigned long overflows later on with maxbufspace. */
maxbuf = (LONG_MAX / 3) / BKVASIZE;
if (nbuf > maxbuf) {
if (!tuned_nbuf)
printf("Warning: nbufs lowered from %d to %ld\n", nbuf,
maxbuf);
nbuf = maxbuf;
}
/*
* swbufs are used as temporary holders for I/O, such as paging I/O.
* We have no less then 16 and no more then 256.
*/
nswbuf = max(min(nbuf/4, 256), 16);
#ifdef NSWBUF_MIN
if (nswbuf < NSWBUF_MIN)
nswbuf = NSWBUF_MIN;
#endif
#ifdef DIRECTIO
ffs_rawread_setup();
#endif
/*
* Reserve space for the buffer cache buffers
*/
swbuf = (void *)v;
v = (caddr_t)(swbuf + nswbuf);
buf = (void *)v;
v = (caddr_t)(buf + nbuf);
return(v);
}
/* Initialize the buffer subsystem. Called before use of any buffers. */
void
bufinit(void)
{
struct buf *bp;
int i;
mtx_init(&bqlock, "buf queue lock", NULL, MTX_DEF);
mtx_init(&rbreqlock, "runningbufspace lock", NULL, MTX_DEF);
mtx_init(&nblock, "needsbuffer lock", NULL, MTX_DEF);
mtx_init(&bdlock, "buffer daemon lock", NULL, MTX_DEF);
/* next, make a null set of free lists */
for (i = 0; i < BUFFER_QUEUES; i++)
TAILQ_INIT(&bufqueues[i]);
/* finally, initialize each buffer header and stick on empty q */
for (i = 0; i < nbuf; i++) {
bp = &buf[i];
bzero(bp, sizeof *bp);
bp->b_flags = B_INVAL; /* we're just an empty header */
bp->b_rcred = NOCRED;
bp->b_wcred = NOCRED;
bp->b_qindex = QUEUE_EMPTY;
bp->b_vflags = BV_INFREECNT; /* buf is counted as free */
bp->b_xflags = 0;
LIST_INIT(&bp->b_dep);
BUF_LOCKINIT(bp);
TAILQ_INSERT_TAIL(&bufqueues[QUEUE_EMPTY], bp, b_freelist);
}
/*
* maxbufspace is the absolute maximum amount of buffer space we are
* allowed to reserve in KVM and in real terms. The absolute maximum
* is nominally used by buf_daemon. hibufspace is the nominal maximum
* used by most other processes. The differential is required to
* ensure that buf_daemon is able to run when other processes might
* be blocked waiting for buffer space.
*
* maxbufspace is based on BKVASIZE. Allocating buffers larger then
* this may result in KVM fragmentation which is not handled optimally
* by the system.
*/
maxbufspace = (long)nbuf * BKVASIZE;
hibufspace = lmax(3 * maxbufspace / 4, maxbufspace - MAXBSIZE * 10);
lobufspace = hibufspace - MAXBSIZE;
/*
* Note: The 16 MiB upper limit for hirunningspace was chosen
* arbitrarily and may need further tuning. It corresponds to
* 128 outstanding write IO requests (if IO size is 128 KiB),
* which fits with many RAID controllers' tagged queuing limits.
* The lower 1 MiB limit is the historical upper limit for
* hirunningspace.
*/
hirunningspace = lmax(lmin(roundup(hibufspace / 64, MAXBSIZE),
16 * 1024 * 1024), 1024 * 1024);
lorunningspace = roundup((hirunningspace * 2) / 3, MAXBSIZE);
/*
* Limit the amount of malloc memory since it is wired permanently into
* the kernel space. Even though this is accounted for in the buffer
* allocation, we don't want the malloced region to grow uncontrolled.
* The malloc scheme improves memory utilization significantly on average
* (small) directories.
*/
maxbufmallocspace = hibufspace / 20;
/*
* Reduce the chance of a deadlock occuring by limiting the number
* of delayed-write dirty buffers we allow to stack up.
*/
hidirtybuffers = nbuf / 4 + 20;
dirtybufthresh = hidirtybuffers * 9 / 10;
numdirtybuffers = 0;
/*
* To support extreme low-memory systems, make sure hidirtybuffers cannot
* eat up all available buffer space. This occurs when our minimum cannot
* be met. We try to size hidirtybuffers to 3/4 our buffer space assuming
* BKVASIZE'd (8K) buffers.
*/
while ((long)hidirtybuffers * BKVASIZE > 3 * hibufspace / 4) {
hidirtybuffers >>= 1;
}
lodirtybuffers = hidirtybuffers / 2;
/*
* Try to keep the number of free buffers in the specified range,
* and give special processes (e.g. like buf_daemon) access to an
* emergency reserve.
*/
lofreebuffers = nbuf / 18 + 5;
hifreebuffers = 2 * lofreebuffers;
numfreebuffers = nbuf;
bogus_page = vm_page_alloc(NULL, 0, VM_ALLOC_NOOBJ |
VM_ALLOC_NORMAL | VM_ALLOC_WIRED);
}
/*
* bfreekva() - free the kva allocation for a buffer.
*
* Since this call frees up buffer space, we call bufspacewakeup().
*/
static void
bfreekva(struct buf *bp)
{
if (bp->b_kvasize) {
atomic_add_int(&buffreekvacnt, 1);
atomic_subtract_long(&bufspace, bp->b_kvasize);
vm_map_remove(buffer_map, (vm_offset_t) bp->b_kvabase,
(vm_offset_t) bp->b_kvabase + bp->b_kvasize);
bp->b_kvasize = 0;
bufspacewakeup();
}
}
/*
* bremfree:
*
* Mark the buffer for removal from the appropriate free list in brelse.
*
*/
void
bremfree(struct buf *bp)
{
int old;
CTR3(KTR_BUF, "bremfree(%p) vp %p flags %X", bp, bp->b_vp, bp->b_flags);
KASSERT((bp->b_flags & B_REMFREE) == 0,
("bremfree: buffer %p already marked for delayed removal.", bp));
KASSERT(bp->b_qindex != QUEUE_NONE,
("bremfree: buffer %p not on a queue.", bp));
BUF_ASSERT_HELD(bp);
bp->b_flags |= B_REMFREE;
/* Fixup numfreebuffers count. */
if ((bp->b_flags & B_INVAL) || (bp->b_flags & B_DELWRI) == 0) {
KASSERT((bp->b_vflags & BV_INFREECNT) != 0,
("buf %p not counted in numfreebuffers", bp));
if (bp->b_bufobj != NULL)
mtx_assert(BO_MTX(bp->b_bufobj), MA_OWNED);
bp->b_vflags &= ~BV_INFREECNT;
old = atomic_fetchadd_int(&numfreebuffers, -1);
KASSERT(old > 0, ("numfreebuffers dropped to %d", old - 1));
}
}
/*
* bremfreef:
*
* Force an immediate removal from a free list. Used only in nfs when
* it abuses the b_freelist pointer.
*/
void
bremfreef(struct buf *bp)
{
mtx_lock(&bqlock);
bremfreel(bp);
mtx_unlock(&bqlock);
}
/*
* bremfreel:
*
* Removes a buffer from the free list, must be called with the
* bqlock held.
*/
static void
bremfreel(struct buf *bp)
{
int old;
CTR3(KTR_BUF, "bremfreel(%p) vp %p flags %X",
bp, bp->b_vp, bp->b_flags);
KASSERT(bp->b_qindex != QUEUE_NONE,
("bremfreel: buffer %p not on a queue.", bp));
BUF_ASSERT_HELD(bp);
mtx_assert(&bqlock, MA_OWNED);
TAILQ_REMOVE(&bufqueues[bp->b_qindex], bp, b_freelist);
bp->b_qindex = QUEUE_NONE;
/*
* If this was a delayed bremfree() we only need to remove the buffer
* from the queue and return the stats are already done.
*/
if (bp->b_flags & B_REMFREE) {
bp->b_flags &= ~B_REMFREE;
return;
}
/*
* Fixup numfreebuffers count. If the buffer is invalid or not
* delayed-write, the buffer was free and we must decrement
* numfreebuffers.
*/
if ((bp->b_flags & B_INVAL) || (bp->b_flags & B_DELWRI) == 0) {
KASSERT((bp->b_vflags & BV_INFREECNT) != 0,
("buf %p not counted in numfreebuffers", bp));
if (bp->b_bufobj != NULL)
mtx_assert(BO_MTX(bp->b_bufobj), MA_OWNED);
bp->b_vflags &= ~BV_INFREECNT;
old = atomic_fetchadd_int(&numfreebuffers, -1);
KASSERT(old > 0, ("numfreebuffers dropped to %d", old - 1));
}
}
/*
* Get a buffer with the specified data. Look in the cache first. We
* must clear BIO_ERROR and B_INVAL prior to initiating I/O. If B_CACHE
* is set, the buffer is valid and we do not have to do anything ( see
* getblk() ). This is really just a special case of breadn().
*/
int
bread(struct vnode * vp, daddr_t blkno, int size, struct ucred * cred,
struct buf **bpp)
{
return (breadn(vp, blkno, size, 0, 0, 0, cred, bpp));
}
/*
* Attempt to initiate asynchronous I/O on read-ahead blocks. We must
* clear BIO_ERROR and B_INVAL prior to initiating I/O . If B_CACHE is set,
* the buffer is valid and we do not have to do anything.
*/
void
breada(struct vnode * vp, daddr_t * rablkno, int * rabsize,
int cnt, struct ucred * cred)
{
struct buf *rabp;
int i;
for (i = 0; i < cnt; i++, rablkno++, rabsize++) {
if (inmem(vp, *rablkno))
continue;
rabp = getblk(vp, *rablkno, *rabsize, 0, 0, 0);
if ((rabp->b_flags & B_CACHE) == 0) {
if (!TD_IS_IDLETHREAD(curthread))
curthread->td_ru.ru_inblock++;
rabp->b_flags |= B_ASYNC;
rabp->b_flags &= ~B_INVAL;
rabp->b_ioflags &= ~BIO_ERROR;
rabp->b_iocmd = BIO_READ;
if (rabp->b_rcred == NOCRED && cred != NOCRED)
rabp->b_rcred = crhold(cred);
vfs_busy_pages(rabp, 0);
BUF_KERNPROC(rabp);
rabp->b_iooffset = dbtob(rabp->b_blkno);
bstrategy(rabp);
} else {
brelse(rabp);
}
}
}
/*
* Operates like bread, but also starts asynchronous I/O on
* read-ahead blocks.
*/
int
breadn(struct vnode * vp, daddr_t blkno, int size,
daddr_t * rablkno, int *rabsize,
int cnt, struct ucred * cred, struct buf **bpp)
{
struct buf *bp;
int rv = 0, readwait = 0;
CTR3(KTR_BUF, "breadn(%p, %jd, %d)", vp, blkno, size);
*bpp = bp = getblk(vp, blkno, size, 0, 0, 0);
/* if not found in cache, do some I/O */
if ((bp->b_flags & B_CACHE) == 0) {
if (!TD_IS_IDLETHREAD(curthread))
curthread->td_ru.ru_inblock++;
bp->b_iocmd = BIO_READ;
bp->b_flags &= ~B_INVAL;
bp->b_ioflags &= ~BIO_ERROR;
if (bp->b_rcred == NOCRED && cred != NOCRED)
bp->b_rcred = crhold(cred);
vfs_busy_pages(bp, 0);
bp->b_iooffset = dbtob(bp->b_blkno);
bstrategy(bp);
++readwait;
}
breada(vp, rablkno, rabsize, cnt, cred);
if (readwait) {
rv = bufwait(bp);
}
return (rv);
}
/*
* Write, release buffer on completion. (Done by iodone
* if async). Do not bother writing anything if the buffer
* is invalid.
*
* Note that we set B_CACHE here, indicating that buffer is
* fully valid and thus cacheable. This is true even of NFS
* now so we set it generally. This could be set either here
* or in biodone() since the I/O is synchronous. We put it
* here.
*/
int
bufwrite(struct buf *bp)
{
int oldflags;
struct vnode *vp;
int vp_md;
CTR3(KTR_BUF, "bufwrite(%p) vp %p flags %X", bp, bp->b_vp, bp->b_flags);
if (bp->b_flags & B_INVAL) {
brelse(bp);
return (0);
}
oldflags = bp->b_flags;
BUF_ASSERT_HELD(bp);
if (bp->b_pin_count > 0)
bunpin_wait(bp);
KASSERT(!(bp->b_vflags & BV_BKGRDINPROG),
("FFS background buffer should not get here %p", bp));
vp = bp->b_vp;
if (vp)
vp_md = vp->v_vflag & VV_MD;
else
vp_md = 0;
/* Mark the buffer clean */
bundirty(bp);
bp->b_flags &= ~B_DONE;
bp->b_ioflags &= ~BIO_ERROR;
bp->b_flags |= B_CACHE;
bp->b_iocmd = BIO_WRITE;
bufobj_wref(bp->b_bufobj);
vfs_busy_pages(bp, 1);
/*
* Normal bwrites pipeline writes
*/
bp->b_runningbufspace = bp->b_bufsize;
atomic_add_long(&runningbufspace, bp->b_runningbufspace);
if (!TD_IS_IDLETHREAD(curthread))
curthread->td_ru.ru_oublock++;
if (oldflags & B_ASYNC)
BUF_KERNPROC(bp);
bp->b_iooffset = dbtob(bp->b_blkno);
bstrategy(bp);
if ((oldflags & B_ASYNC) == 0) {
int rtval = bufwait(bp);
brelse(bp);
return (rtval);
} else {
/*
* don't allow the async write to saturate the I/O
* system. We will not deadlock here because
* we are blocking waiting for I/O that is already in-progress
* to complete. We do not block here if it is the update
* or syncer daemon trying to clean up as that can lead
* to deadlock.
*/
if ((curthread->td_pflags & TDP_NORUNNINGBUF) == 0 && !vp_md)
waitrunningbufspace();
}
return (0);
}
void
bufbdflush(struct bufobj *bo, struct buf *bp)
{
struct buf *nbp;
if (bo->bo_dirty.bv_cnt > dirtybufthresh + 10) {
(void) VOP_FSYNC(bp->b_vp, MNT_NOWAIT, curthread);
altbufferflushes++;
} else if (bo->bo_dirty.bv_cnt > dirtybufthresh) {
BO_LOCK(bo);
/*
* Try to find a buffer to flush.
*/
TAILQ_FOREACH(nbp, &bo->bo_dirty.bv_hd, b_bobufs) {
if ((nbp->b_vflags & BV_BKGRDINPROG) ||
BUF_LOCK(nbp,
LK_EXCLUSIVE | LK_NOWAIT, NULL))
continue;
if (bp == nbp)
panic("bdwrite: found ourselves");
BO_UNLOCK(bo);
/* Don't countdeps with the bo lock held. */
if (buf_countdeps(nbp, 0)) {
BO_LOCK(bo);
BUF_UNLOCK(nbp);
continue;
}
if (nbp->b_flags & B_CLUSTEROK) {
vfs_bio_awrite(nbp);
} else {
bremfree(nbp);
bawrite(nbp);
}
dirtybufferflushes++;
break;
}
if (nbp == NULL)
BO_UNLOCK(bo);
}
}
/*
* Delayed write. (Buffer is marked dirty). Do not bother writing
* anything if the buffer is marked invalid.
*
* Note that since the buffer must be completely valid, we can safely
* set B_CACHE. In fact, we have to set B_CACHE here rather then in
* biodone() in order to prevent getblk from writing the buffer
* out synchronously.
*/
void
bdwrite(struct buf *bp)
{
struct thread *td = curthread;
struct vnode *vp;
struct bufobj *bo;
CTR3(KTR_BUF, "bdwrite(%p) vp %p flags %X", bp, bp->b_vp, bp->b_flags);
KASSERT(bp->b_bufobj != NULL, ("No b_bufobj %p", bp));
BUF_ASSERT_HELD(bp);
if (bp->b_flags & B_INVAL) {
brelse(bp);
return;
}
/*
* If we have too many dirty buffers, don't create any more.
* If we are wildly over our limit, then force a complete
* cleanup. Otherwise, just keep the situation from getting
* out of control. Note that we have to avoid a recursive
* disaster and not try to clean up after our own cleanup!
*/
vp = bp->b_vp;
bo = bp->b_bufobj;
if ((td->td_pflags & (TDP_COWINPROGRESS|TDP_INBDFLUSH)) == 0) {
td->td_pflags |= TDP_INBDFLUSH;
BO_BDFLUSH(bo, bp);
td->td_pflags &= ~TDP_INBDFLUSH;
} else
recursiveflushes++;
bdirty(bp);
/*
* Set B_CACHE, indicating that the buffer is fully valid. This is
* true even of NFS now.
*/
bp->b_flags |= B_CACHE;
/*
* This bmap keeps the system from needing to do the bmap later,
* perhaps when the system is attempting to do a sync. Since it
* is likely that the indirect block -- or whatever other datastructure
* that the filesystem needs is still in memory now, it is a good
* thing to do this. Note also, that if the pageout daemon is
* requesting a sync -- there might not be enough memory to do
* the bmap then... So, this is important to do.
*/
if (vp->v_type != VCHR && bp->b_lblkno == bp->b_blkno) {
VOP_BMAP(vp, bp->b_lblkno, NULL, &bp->b_blkno, NULL, NULL);
}
/*
* Set the *dirty* buffer range based upon the VM system dirty
* pages.
*
* Mark the buffer pages as clean. We need to do this here to
* satisfy the vnode_pager and the pageout daemon, so that it
* thinks that the pages have been "cleaned". Note that since
* the pages are in a delayed write buffer -- the VFS layer
* "will" see that the pages get written out on the next sync,
* or perhaps the cluster will be completed.
*/
vfs_clean_pages_dirty_buf(bp);
bqrelse(bp);
/*
* Wakeup the buffer flushing daemon if we have a lot of dirty
* buffers (midpoint between our recovery point and our stall
* point).
*/
bd_wakeup((lodirtybuffers + hidirtybuffers) / 2);
/*
* note: we cannot initiate I/O from a bdwrite even if we wanted to,
* due to the softdep code.
*/
}
/*
* bdirty:
*
* Turn buffer into delayed write request. We must clear BIO_READ and
* B_RELBUF, and we must set B_DELWRI. We reassign the buffer to
* itself to properly update it in the dirty/clean lists. We mark it
* B_DONE to ensure that any asynchronization of the buffer properly
* clears B_DONE ( else a panic will occur later ).
*
* bdirty() is kinda like bdwrite() - we have to clear B_INVAL which
* might have been set pre-getblk(). Unlike bwrite/bdwrite, bdirty()
* should only be called if the buffer is known-good.
*
* Since the buffer is not on a queue, we do not update the numfreebuffers
* count.
*
* The buffer must be on QUEUE_NONE.
*/
void
bdirty(struct buf *bp)
{
CTR3(KTR_BUF, "bdirty(%p) vp %p flags %X",
bp, bp->b_vp, bp->b_flags);
KASSERT(bp->b_bufobj != NULL, ("No b_bufobj %p", bp));
KASSERT(bp->b_flags & B_REMFREE || bp->b_qindex == QUEUE_NONE,
("bdirty: buffer %p still on queue %d", bp, bp->b_qindex));
BUF_ASSERT_HELD(bp);
bp->b_flags &= ~(B_RELBUF);
bp->b_iocmd = BIO_WRITE;
if ((bp->b_flags & B_DELWRI) == 0) {
bp->b_flags |= /* XXX B_DONE | */ B_DELWRI;
reassignbuf(bp);
atomic_add_int(&numdirtybuffers, 1);
bd_wakeup((lodirtybuffers + hidirtybuffers) / 2);
}
}
/*
* bundirty:
*
* Clear B_DELWRI for buffer.
*
* Since the buffer is not on a queue, we do not update the numfreebuffers
* count.
*
* The buffer must be on QUEUE_NONE.
*/
void
bundirty(struct buf *bp)
{
CTR3(KTR_BUF, "bundirty(%p) vp %p flags %X", bp, bp->b_vp, bp->b_flags);
KASSERT(bp->b_bufobj != NULL, ("No b_bufobj %p", bp));
KASSERT(bp->b_flags & B_REMFREE || bp->b_qindex == QUEUE_NONE,
("bundirty: buffer %p still on queue %d", bp, bp->b_qindex));
BUF_ASSERT_HELD(bp);
if (bp->b_flags & B_DELWRI) {
bp->b_flags &= ~B_DELWRI;
reassignbuf(bp);
atomic_subtract_int(&numdirtybuffers, 1);
numdirtywakeup(lodirtybuffers);
}
/*
* Since it is now being written, we can clear its deferred write flag.
*/
bp->b_flags &= ~B_DEFERRED;
}
/*
* bawrite:
*
* Asynchronous write. Start output on a buffer, but do not wait for
* it to complete. The buffer is released when the output completes.
*
* bwrite() ( or the VOP routine anyway ) is responsible for handling
* B_INVAL buffers. Not us.
*/
void
bawrite(struct buf *bp)
{
bp->b_flags |= B_ASYNC;
(void) bwrite(bp);
}
/*
* bwillwrite:
*
* Called prior to the locking of any vnodes when we are expecting to
* write. We do not want to starve the buffer cache with too many
* dirty buffers so we block here. By blocking prior to the locking
* of any vnodes we attempt to avoid the situation where a locked vnode
* prevents the various system daemons from flushing related buffers.
*/
void
bwillwrite(void)
{
if (numdirtybuffers >= hidirtybuffers) {
mtx_lock(&nblock);
while (numdirtybuffers >= hidirtybuffers) {
bd_wakeup(1);
needsbuffer |= VFS_BIO_NEED_DIRTYFLUSH;
msleep(&needsbuffer, &nblock,
(PRIBIO + 4), "flswai", 0);
}
mtx_unlock(&nblock);
}
}
/*
* Return true if we have too many dirty buffers.
*/
int
buf_dirty_count_severe(void)
{
return(numdirtybuffers >= hidirtybuffers);
}
static __noinline int
buf_vm_page_count_severe(void)
{
KFAIL_POINT_CODE(DEBUG_FP, buf_pressure, return 1);
return vm_page_count_severe();
}
/*
* brelse:
*
* Release a busy buffer and, if requested, free its resources. The
* buffer will be stashed in the appropriate bufqueue[] allowing it
* to be accessed later as a cache entity or reused for other purposes.
*/
void
brelse(struct buf *bp)
{
CTR3(KTR_BUF, "brelse(%p) vp %p flags %X",
bp, bp->b_vp, bp->b_flags);
KASSERT(!(bp->b_flags & (B_CLUSTER|B_PAGING)),
("brelse: inappropriate B_PAGING or B_CLUSTER bp %p", bp));
if (bp->b_flags & B_MANAGED) {
bqrelse(bp);
return;
}
if (bp->b_iocmd == BIO_WRITE && (bp->b_ioflags & BIO_ERROR) &&
bp->b_error == EIO && !(bp->b_flags & B_INVAL)) {
/*
* Failed write, redirty. Must clear BIO_ERROR to prevent
* pages from being scrapped. If the error is anything
* other than an I/O error (EIO), assume that retrying
* is futile.
*/
bp->b_ioflags &= ~BIO_ERROR;
bdirty(bp);
} else if ((bp->b_flags & (B_NOCACHE | B_INVAL)) ||
(bp->b_ioflags & BIO_ERROR) || (bp->b_bufsize <= 0)) {
/*
* Either a failed I/O or we were asked to free or not
* cache the buffer.
*/
bp->b_flags |= B_INVAL;
if (!LIST_EMPTY(&bp->b_dep))
buf_deallocate(bp);
if (bp->b_flags & B_DELWRI) {
atomic_subtract_int(&numdirtybuffers, 1);
numdirtywakeup(lodirtybuffers);
}
bp->b_flags &= ~(B_DELWRI | B_CACHE);
if ((bp->b_flags & B_VMIO) == 0) {
if (bp->b_bufsize)
allocbuf(bp, 0);
if (bp->b_vp)
brelvp(bp);
}
}
/*
* We must clear B_RELBUF if B_DELWRI is set. If vfs_vmio_release()
* is called with B_DELWRI set, the underlying pages may wind up
* getting freed causing a previous write (bdwrite()) to get 'lost'
* because pages associated with a B_DELWRI bp are marked clean.
*
* We still allow the B_INVAL case to call vfs_vmio_release(), even
* if B_DELWRI is set.
*
* If B_DELWRI is not set we may have to set B_RELBUF if we are low
* on pages to return pages to the VM page queues.
*/
if (bp->b_flags & B_DELWRI)
bp->b_flags &= ~B_RELBUF;
else if (buf_vm_page_count_severe()) {
/*
* The locking of the BO_LOCK is not necessary since
* BKGRDINPROG cannot be set while we hold the buf
* lock, it can only be cleared if it is already
* pending.
*/
if (bp->b_vp) {
if (!(bp->b_vflags & BV_BKGRDINPROG))
bp->b_flags |= B_RELBUF;
} else
bp->b_flags |= B_RELBUF;
}
/*
* VMIO buffer rundown. It is not very necessary to keep a VMIO buffer
* constituted, not even NFS buffers now. Two flags effect this. If
* B_INVAL, the struct buf is invalidated but the VM object is kept
* around ( i.e. so it is trivial to reconstitute the buffer later ).
*
* If BIO_ERROR or B_NOCACHE is set, pages in the VM object will be
* invalidated. BIO_ERROR cannot be set for a failed write unless the
* buffer is also B_INVAL because it hits the re-dirtying code above.
*
* Normally we can do this whether a buffer is B_DELWRI or not. If
* the buffer is an NFS buffer, it is tracking piecemeal writes or
* the commit state and we cannot afford to lose the buffer. If the
* buffer has a background write in progress, we need to keep it
* around to prevent it from being reconstituted and starting a second
* background write.
*/
if ((bp->b_flags & B_VMIO)
&& !(bp->b_vp->v_mount != NULL &&
(bp->b_vp->v_mount->mnt_vfc->vfc_flags & VFCF_NETWORK) != 0 &&
!vn_isdisk(bp->b_vp, NULL) &&
(bp->b_flags & B_DELWRI))
) {
int i, j, resid;
vm_page_t m;
off_t foff;
vm_pindex_t poff;
vm_object_t obj;
obj = bp->b_bufobj->bo_object;
/*
* Get the base offset and length of the buffer. Note that
* in the VMIO case if the buffer block size is not
* page-aligned then b_data pointer may not be page-aligned.
* But our b_pages[] array *IS* page aligned.
*
* block sizes less then DEV_BSIZE (usually 512) are not
* supported due to the page granularity bits (m->valid,
* m->dirty, etc...).
*
* See man buf(9) for more information
*/
resid = bp->b_bufsize;
foff = bp->b_offset;
VM_OBJECT_LOCK(obj);
for (i = 0; i < bp->b_npages; i++) {
int had_bogus = 0;
m = bp->b_pages[i];
/*
* If we hit a bogus page, fixup *all* the bogus pages
* now.
*/
if (m == bogus_page) {
poff = OFF_TO_IDX(bp->b_offset);
had_bogus = 1;
for (j = i; j < bp->b_npages; j++) {
vm_page_t mtmp;
mtmp = bp->b_pages[j];
if (mtmp == bogus_page) {
mtmp = vm_page_lookup(obj, poff + j);
if (!mtmp) {
panic("brelse: page missing\n");
}
bp->b_pages[j] = mtmp;
}
}
if ((bp->b_flags & B_INVAL) == 0) {
pmap_qenter(
trunc_page((vm_offset_t)bp->b_data),
bp->b_pages, bp->b_npages);
}
m = bp->b_pages[i];
}
if ((bp->b_flags & B_NOCACHE) ||
(bp->b_ioflags & BIO_ERROR &&
bp->b_iocmd == BIO_READ)) {
int poffset = foff & PAGE_MASK;
int presid = resid > (PAGE_SIZE - poffset) ?
(PAGE_SIZE - poffset) : resid;
KASSERT(presid >= 0, ("brelse: extra page"));
vm_page_set_invalid(m, poffset, presid);
if (had_bogus)
printf("avoided corruption bug in bogus_page/brelse code\n");
}
resid -= PAGE_SIZE - (foff & PAGE_MASK);
foff = (foff + PAGE_SIZE) & ~(off_t)PAGE_MASK;
}
VM_OBJECT_UNLOCK(obj);
if (bp->b_flags & (B_INVAL | B_RELBUF))
vfs_vmio_release(bp);
} else if (bp->b_flags & B_VMIO) {
if (bp->b_flags & (B_INVAL | B_RELBUF)) {
vfs_vmio_release(bp);
}
} else if ((bp->b_flags & (B_INVAL | B_RELBUF)) != 0) {
if (bp->b_bufsize != 0)
allocbuf(bp, 0);
if (bp->b_vp != NULL)
brelvp(bp);
}
if (BUF_LOCKRECURSED(bp)) {
/* do not release to free list */
BUF_UNLOCK(bp);
return;
}
/* enqueue */
mtx_lock(&bqlock);
/* Handle delayed bremfree() processing. */
if (bp->b_flags & B_REMFREE) {
struct bufobj *bo;
bo = bp->b_bufobj;
if (bo != NULL)
BO_LOCK(bo);
bremfreel(bp);
if (bo != NULL)
BO_UNLOCK(bo);
}
if (bp->b_qindex != QUEUE_NONE)
panic("brelse: free buffer onto another queue???");
/*
* If the buffer has junk contents signal it and eventually
* clean up B_DELWRI and diassociate the vnode so that gbincore()
* doesn't find it.
*/
if (bp->b_bufsize == 0 || (bp->b_ioflags & BIO_ERROR) != 0 ||
(bp->b_flags & (B_INVAL | B_NOCACHE | B_RELBUF)) != 0)
bp->b_flags |= B_INVAL;
if (bp->b_flags & B_INVAL) {
if (bp->b_flags & B_DELWRI)
bundirty(bp);
if (bp->b_vp)
brelvp(bp);
}
/* buffers with no memory */
if (bp->b_bufsize == 0) {
bp->b_xflags &= ~(BX_BKGRDWRITE | BX_ALTDATA);
if (bp->b_vflags & BV_BKGRDINPROG)
panic("losing buffer 1");
if (bp->b_kvasize) {
bp->b_qindex = QUEUE_EMPTYKVA;
} else {
bp->b_qindex = QUEUE_EMPTY;
}
TAILQ_INSERT_HEAD(&bufqueues[bp->b_qindex], bp, b_freelist);
/* buffers with junk contents */
} else if (bp->b_flags & (B_INVAL | B_NOCACHE | B_RELBUF) ||
(bp->b_ioflags & BIO_ERROR)) {
bp->b_xflags &= ~(BX_BKGRDWRITE | BX_ALTDATA);
if (bp->b_vflags & BV_BKGRDINPROG)
panic("losing buffer 2");
bp->b_qindex = QUEUE_CLEAN;
TAILQ_INSERT_HEAD(&bufqueues[QUEUE_CLEAN], bp, b_freelist);
/* remaining buffers */
} else {
if ((bp->b_flags & (B_DELWRI|B_NEEDSGIANT)) ==
(B_DELWRI|B_NEEDSGIANT))
bp->b_qindex = QUEUE_DIRTY_GIANT;
else if (bp->b_flags & B_DELWRI)
bp->b_qindex = QUEUE_DIRTY;
else
bp->b_qindex = QUEUE_CLEAN;
if (bp->b_flags & B_AGE)
TAILQ_INSERT_HEAD(&bufqueues[bp->b_qindex], bp, b_freelist);
else
TAILQ_INSERT_TAIL(&bufqueues[bp->b_qindex], bp, b_freelist);
}
mtx_unlock(&bqlock);
/*
* Fixup numfreebuffers count. The bp is on an appropriate queue
* unless locked. We then bump numfreebuffers if it is not B_DELWRI.
* We've already handled the B_INVAL case ( B_DELWRI will be clear
* if B_INVAL is set ).
*/
if (!(bp->b_flags & B_DELWRI)) {
struct bufobj *bo;
bo = bp->b_bufobj;
if (bo != NULL)
BO_LOCK(bo);
bufcountwakeup(bp);
if (bo != NULL)
BO_UNLOCK(bo);
}
/*
* Something we can maybe free or reuse
*/
if (bp->b_bufsize || bp->b_kvasize)
bufspacewakeup();
bp->b_flags &= ~(B_ASYNC | B_NOCACHE | B_AGE | B_RELBUF | B_DIRECT);
if ((bp->b_flags & B_DELWRI) == 0 && (bp->b_xflags & BX_VNDIRTY))
panic("brelse: not dirty");
/* unlock */
BUF_UNLOCK(bp);
}
/*
* Release a buffer back to the appropriate queue but do not try to free
* it. The buffer is expected to be used again soon.
*
* bqrelse() is used by bdwrite() to requeue a delayed write, and used by
* biodone() to requeue an async I/O on completion. It is also used when
* known good buffers need to be requeued but we think we may need the data
* again soon.
*
* XXX we should be able to leave the B_RELBUF hint set on completion.
*/
void
bqrelse(struct buf *bp)
{
struct bufobj *bo;
CTR3(KTR_BUF, "bqrelse(%p) vp %p flags %X", bp, bp->b_vp, bp->b_flags);
KASSERT(!(bp->b_flags & (B_CLUSTER|B_PAGING)),
("bqrelse: inappropriate B_PAGING or B_CLUSTER bp %p", bp));
if (BUF_LOCKRECURSED(bp)) {
/* do not release to free list */
BUF_UNLOCK(bp);
return;
}
bo = bp->b_bufobj;
if (bp->b_flags & B_MANAGED) {
if (bp->b_flags & B_REMFREE) {
mtx_lock(&bqlock);
if (bo != NULL)
BO_LOCK(bo);
bremfreel(bp);
if (bo != NULL)
BO_UNLOCK(bo);
mtx_unlock(&bqlock);
}
bp->b_flags &= ~(B_ASYNC | B_NOCACHE | B_AGE | B_RELBUF);
BUF_UNLOCK(bp);
return;
}
mtx_lock(&bqlock);
/* Handle delayed bremfree() processing. */
if (bp->b_flags & B_REMFREE) {
if (bo != NULL)
BO_LOCK(bo);
bremfreel(bp);
if (bo != NULL)
BO_UNLOCK(bo);
}
if (bp->b_qindex != QUEUE_NONE)
panic("bqrelse: free buffer onto another queue???");
/* buffers with stale but valid contents */
if (bp->b_flags & B_DELWRI) {
if (bp->b_flags & B_NEEDSGIANT)
bp->b_qindex = QUEUE_DIRTY_GIANT;
else
bp->b_qindex = QUEUE_DIRTY;
TAILQ_INSERT_TAIL(&bufqueues[bp->b_qindex], bp, b_freelist);
} else {
/*
* The locking of the BO_LOCK for checking of the
* BV_BKGRDINPROG is not necessary since the
* BV_BKGRDINPROG cannot be set while we hold the buf
* lock, it can only be cleared if it is already
* pending.
*/
if (!buf_vm_page_count_severe() || (bp->b_vflags & BV_BKGRDINPROG)) {
bp->b_qindex = QUEUE_CLEAN;
TAILQ_INSERT_TAIL(&bufqueues[QUEUE_CLEAN], bp,
b_freelist);
} else {
/*
* We are too low on memory, we have to try to free
* the buffer (most importantly: the wired pages
* making up its backing store) *now*.
*/
mtx_unlock(&bqlock);
brelse(bp);
return;
}
}
mtx_unlock(&bqlock);
if ((bp->b_flags & B_INVAL) || !(bp->b_flags & B_DELWRI)) {
if (bo != NULL)
BO_LOCK(bo);
bufcountwakeup(bp);
if (bo != NULL)
BO_UNLOCK(bo);
}
/*
* Something we can maybe free or reuse.
*/
if (bp->b_bufsize && !(bp->b_flags & B_DELWRI))
bufspacewakeup();
bp->b_flags &= ~(B_ASYNC | B_NOCACHE | B_AGE | B_RELBUF);
if ((bp->b_flags & B_DELWRI) == 0 && (bp->b_xflags & BX_VNDIRTY))
panic("bqrelse: not dirty");
/* unlock */
BUF_UNLOCK(bp);
}
/* Give pages used by the bp back to the VM system (where possible) */
static void
vfs_vmio_release(struct buf *bp)
{
int i;
vm_page_t m;
VM_OBJECT_LOCK(bp->b_bufobj->bo_object);
for (i = 0; i < bp->b_npages; i++) {
m = bp->b_pages[i];
bp->b_pages[i] = NULL;
/*
* In order to keep page LRU ordering consistent, put
* everything on the inactive queue.
*/
vm_page_lock(m);
vm_page_unwire(m, 0);
/*
* We don't mess with busy pages, it is
* the responsibility of the process that
* busied the pages to deal with them.
*/
if ((m->oflags & VPO_BUSY) == 0 && m->busy == 0 &&
m->wire_count == 0) {
/*
* Might as well free the page if we can and it has
* no valid data. We also free the page if the
* buffer was used for direct I/O
*/
if ((bp->b_flags & B_ASYNC) == 0 && !m->valid) {
vm_page_free(m);
} else if (bp->b_flags & B_DIRECT) {
vm_page_try_to_free(m);
} else if (buf_vm_page_count_severe()) {
vm_page_try_to_cache(m);
}
}
vm_page_unlock(m);
}
VM_OBJECT_UNLOCK(bp->b_bufobj->bo_object);
pmap_qremove(trunc_page((vm_offset_t) bp->b_data), bp->b_npages);
if (bp->b_bufsize) {
bufspacewakeup();
bp->b_bufsize = 0;
}
bp->b_npages = 0;
bp->b_flags &= ~B_VMIO;
if (bp->b_vp)
brelvp(bp);
}
/*
* Check to see if a block at a particular lbn is available for a clustered
* write.
*/
static int
vfs_bio_clcheck(struct vnode *vp, int size, daddr_t lblkno, daddr_t blkno)
{
struct buf *bpa;
int match;
match = 0;
/* If the buf isn't in core skip it */
if ((bpa = gbincore(&vp->v_bufobj, lblkno)) == NULL)
return (0);
/* If the buf is busy we don't want to wait for it */
if (BUF_LOCK(bpa, LK_EXCLUSIVE | LK_NOWAIT, NULL) != 0)
return (0);
/* Only cluster with valid clusterable delayed write buffers */
if ((bpa->b_flags & (B_DELWRI | B_CLUSTEROK | B_INVAL)) !=
(B_DELWRI | B_CLUSTEROK))
goto done;
if (bpa->b_bufsize != size)
goto done;
/*
* Check to see if it is in the expected place on disk and that the
* block has been mapped.
*/
if ((bpa->b_blkno != bpa->b_lblkno) && (bpa->b_blkno == blkno))
match = 1;
done:
BUF_UNLOCK(bpa);
return (match);
}
/*
* vfs_bio_awrite:
*
* Implement clustered async writes for clearing out B_DELWRI buffers.
* This is much better then the old way of writing only one buffer at
* a time. Note that we may not be presented with the buffers in the
* correct order, so we search for the cluster in both directions.
*/
int
vfs_bio_awrite(struct buf *bp)
{
struct bufobj *bo;
int i;
int j;
daddr_t lblkno = bp->b_lblkno;
struct vnode *vp = bp->b_vp;
int ncl;
int nwritten;
int size;
int maxcl;
bo = &vp->v_bufobj;
/*
* right now we support clustered writing only to regular files. If
* we find a clusterable block we could be in the middle of a cluster
* rather then at the beginning.
*/
if ((vp->v_type == VREG) &&
(vp->v_mount != 0) && /* Only on nodes that have the size info */
(bp->b_flags & (B_CLUSTEROK | B_INVAL)) == B_CLUSTEROK) {
size = vp->v_mount->mnt_stat.f_iosize;
maxcl = MAXPHYS / size;
BO_LOCK(bo);
for (i = 1; i < maxcl; i++)
if (vfs_bio_clcheck(vp, size, lblkno + i,
bp->b_blkno + ((i * size) >> DEV_BSHIFT)) == 0)
break;
for (j = 1; i + j <= maxcl && j <= lblkno; j++)
if (vfs_bio_clcheck(vp, size, lblkno - j,
bp->b_blkno - ((j * size) >> DEV_BSHIFT)) == 0)
break;
BO_UNLOCK(bo);
--j;
ncl = i + j;
/*
* this is a possible cluster write
*/
if (ncl != 1) {
BUF_UNLOCK(bp);
nwritten = cluster_wbuild(vp, size, lblkno - j, ncl);
return nwritten;
}
}
bremfree(bp);
bp->b_flags |= B_ASYNC;
/*
* default (old) behavior, writing out only one block
*
* XXX returns b_bufsize instead of b_bcount for nwritten?
*/
nwritten = bp->b_bufsize;
(void) bwrite(bp);
return nwritten;
}
/*
* getnewbuf:
*
* Find and initialize a new buffer header, freeing up existing buffers
* in the bufqueues as necessary. The new buffer is returned locked.
*
* Important: B_INVAL is not set. If the caller wishes to throw the
* buffer away, the caller must set B_INVAL prior to calling brelse().
*
* We block if:
* We have insufficient buffer headers
* We have insufficient buffer space
* buffer_map is too fragmented ( space reservation fails )
* If we have to flush dirty buffers ( but we try to avoid this )
*
* To avoid VFS layer recursion we do not flush dirty buffers ourselves.
* Instead we ask the buf daemon to do it for us. We attempt to
* avoid piecemeal wakeups of the pageout daemon.
*/
static struct buf *
getnewbuf(struct vnode *vp, int slpflag, int slptimeo, int size, int maxsize,
int gbflags)
{
struct thread *td;
struct buf *bp;
struct buf *nbp;
int defrag = 0;
int nqindex;
static int flushingbufs;
td = curthread;
/*
* We can't afford to block since we might be holding a vnode lock,
* which may prevent system daemons from running. We deal with
* low-memory situations by proactively returning memory and running
* async I/O rather then sync I/O.
*/
atomic_add_int(&getnewbufcalls, 1);
atomic_subtract_int(&getnewbufrestarts, 1);
restart:
atomic_add_int(&getnewbufrestarts, 1);
/*
* Setup for scan. If we do not have enough free buffers,
* we setup a degenerate case that immediately fails. Note
* that if we are specially marked process, we are allowed to
* dip into our reserves.
*
* The scanning sequence is nominally: EMPTY->EMPTYKVA->CLEAN
*
* We start with EMPTYKVA. If the list is empty we backup to EMPTY.
* However, there are a number of cases (defragging, reusing, ...)
* where we cannot backup.
*/
mtx_lock(&bqlock);
nqindex = QUEUE_EMPTYKVA;
nbp = TAILQ_FIRST(&bufqueues[QUEUE_EMPTYKVA]);
if (nbp == NULL) {
/*
* If no EMPTYKVA buffers and we are either
* defragging or reusing, locate a CLEAN buffer
* to free or reuse. If bufspace useage is low
* skip this step so we can allocate a new buffer.
*/
if (defrag || bufspace >= lobufspace) {
nqindex = QUEUE_CLEAN;
nbp = TAILQ_FIRST(&bufqueues[QUEUE_CLEAN]);
}
/*
* If we could not find or were not allowed to reuse a
* CLEAN buffer, check to see if it is ok to use an EMPTY
* buffer. We can only use an EMPTY buffer if allocating
* its KVA would not otherwise run us out of buffer space.
*/
if (nbp == NULL && defrag == 0 &&
bufspace + maxsize < hibufspace) {
nqindex = QUEUE_EMPTY;
nbp = TAILQ_FIRST(&bufqueues[QUEUE_EMPTY]);
}
}
/*
* Run scan, possibly freeing data and/or kva mappings on the fly
* depending.
*/
while ((bp = nbp) != NULL) {
int qindex = nqindex;
/*
* Calculate next bp ( we can only use it if we do not block
* or do other fancy things ).
*/
if ((nbp = TAILQ_NEXT(bp, b_freelist)) == NULL) {
switch(qindex) {
case QUEUE_EMPTY:
nqindex = QUEUE_EMPTYKVA;
if ((nbp = TAILQ_FIRST(&bufqueues[QUEUE_EMPTYKVA])))
break;
/* FALLTHROUGH */
case QUEUE_EMPTYKVA:
nqindex = QUEUE_CLEAN;
if ((nbp = TAILQ_FIRST(&bufqueues[QUEUE_CLEAN])))
break;
/* FALLTHROUGH */
case QUEUE_CLEAN:
/*
* nbp is NULL.
*/
break;
}
}
/*
* If we are defragging then we need a buffer with
* b_kvasize != 0. XXX this situation should no longer
* occur, if defrag is non-zero the buffer's b_kvasize
* should also be non-zero at this point. XXX
*/
if (defrag && bp->b_kvasize == 0) {
printf("Warning: defrag empty buffer %p\n", bp);
continue;
}
/*
* Start freeing the bp. This is somewhat involved. nbp
* remains valid only for QUEUE_EMPTY[KVA] bp's.
*/
if (BUF_LOCK(bp, LK_EXCLUSIVE | LK_NOWAIT, NULL) != 0)
continue;
if (bp->b_vp) {
BO_LOCK(bp->b_bufobj);
if (bp->b_vflags & BV_BKGRDINPROG) {
BO_UNLOCK(bp->b_bufobj);
BUF_UNLOCK(bp);
continue;
}
BO_UNLOCK(bp->b_bufobj);
}
CTR6(KTR_BUF,
"getnewbuf(%p) vp %p flags %X kvasize %d bufsize %d "
"queue %d (recycling)", bp, bp->b_vp, bp->b_flags,
bp->b_kvasize, bp->b_bufsize, qindex);
/*
* Sanity Checks
*/
KASSERT(bp->b_qindex == qindex, ("getnewbuf: inconsistant queue %d bp %p", qindex, bp));
/*
* Note: we no longer distinguish between VMIO and non-VMIO
* buffers.
*/
KASSERT((bp->b_flags & B_DELWRI) == 0, ("delwri buffer %p found in queue %d", bp, qindex));
if (bp->b_bufobj != NULL)
BO_LOCK(bp->b_bufobj);
bremfreel(bp);
if (bp->b_bufobj != NULL)
BO_UNLOCK(bp->b_bufobj);
mtx_unlock(&bqlock);
if (qindex == QUEUE_CLEAN) {
if (bp->b_flags & B_VMIO) {
bp->b_flags &= ~B_ASYNC;
vfs_vmio_release(bp);
}
if (bp->b_vp)
brelvp(bp);
}
/*
* NOTE: nbp is now entirely invalid. We can only restart
* the scan from this point on.
*
* Get the rest of the buffer freed up. b_kva* is still
* valid after this operation.
*/
if (bp->b_rcred != NOCRED) {
crfree(bp->b_rcred);
bp->b_rcred = NOCRED;
}
if (bp->b_wcred != NOCRED) {
crfree(bp->b_wcred);
bp->b_wcred = NOCRED;
}
if (!LIST_EMPTY(&bp->b_dep))
buf_deallocate(bp);
if (bp->b_vflags & BV_BKGRDINPROG)
panic("losing buffer 3");
KASSERT(bp->b_vp == NULL,
("bp: %p still has vnode %p. qindex: %d",
bp, bp->b_vp, qindex));
KASSERT((bp->b_xflags & (BX_VNCLEAN|BX_VNDIRTY)) == 0,
("bp: %p still on a buffer list. xflags %X",
bp, bp->b_xflags));
if (bp->b_bufsize)
allocbuf(bp, 0);
bp->b_flags = 0;
bp->b_ioflags = 0;
bp->b_xflags = 0;
KASSERT((bp->b_vflags & BV_INFREECNT) == 0,
("buf %p still counted as free?", bp));
bp->b_vflags = 0;
bp->b_vp = NULL;
bp->b_blkno = bp->b_lblkno = 0;
bp->b_offset = NOOFFSET;
bp->b_iodone = 0;
bp->b_error = 0;
bp->b_resid = 0;
bp->b_bcount = 0;
bp->b_npages = 0;
bp->b_dirtyoff = bp->b_dirtyend = 0;
bp->b_bufobj = NULL;
bp->b_pin_count = 0;
bp->b_fsprivate1 = NULL;
bp->b_fsprivate2 = NULL;
bp->b_fsprivate3 = NULL;
LIST_INIT(&bp->b_dep);
/*
* If we are defragging then free the buffer.
*/
if (defrag) {
bp->b_flags |= B_INVAL;
bfreekva(bp);
brelse(bp);
defrag = 0;
goto restart;
}
/*
* Notify any waiters for the buffer lock about
* identity change by freeing the buffer.
*/
if (qindex == QUEUE_CLEAN && BUF_LOCKWAITERS(bp)) {
bp->b_flags |= B_INVAL;
bfreekva(bp);
brelse(bp);
goto restart;
}
/*
* If we are overcomitted then recover the buffer and its
* KVM space. This occurs in rare situations when multiple
* processes are blocked in getnewbuf() or allocbuf().
*/
if (bufspace >= hibufspace)
flushingbufs = 1;
if (flushingbufs && bp->b_kvasize != 0) {
bp->b_flags |= B_INVAL;
bfreekva(bp);
brelse(bp);
goto restart;
}
if (bufspace < lobufspace)
flushingbufs = 0;
break;
}
/*
* If we exhausted our list, sleep as appropriate. We may have to
* wakeup various daemons and write out some dirty buffers.
*
* Generally we are sleeping due to insufficient buffer space.
*/
if (bp == NULL) {
int flags, norunbuf;
char *waitmsg;
int fl;
if (defrag) {
flags = VFS_BIO_NEED_BUFSPACE;
waitmsg = "nbufkv";
} else if (bufspace >= hibufspace) {
waitmsg = "nbufbs";
flags = VFS_BIO_NEED_BUFSPACE;
} else {
waitmsg = "newbuf";
flags = VFS_BIO_NEED_ANY;
}
mtx_lock(&nblock);
needsbuffer |= flags;
mtx_unlock(&nblock);
mtx_unlock(&bqlock);
bd_speedup(); /* heeeelp */
if (gbflags & GB_NOWAIT_BD)
return (NULL);
mtx_lock(&nblock);
while (needsbuffer & flags) {
if (vp != NULL && (td->td_pflags & TDP_BUFNEED) == 0) {
mtx_unlock(&nblock);
/*
* getblk() is called with a vnode
* locked, and some majority of the
* dirty buffers may as well belong to
* the vnode. Flushing the buffers
* there would make a progress that
* cannot be achieved by the
* buf_daemon, that cannot lock the
* vnode.
*/
norunbuf = ~(TDP_BUFNEED | TDP_NORUNNINGBUF) |
(td->td_pflags & TDP_NORUNNINGBUF);
/* play bufdaemon */
td->td_pflags |= TDP_BUFNEED | TDP_NORUNNINGBUF;
fl = buf_do_flush(vp);
td->td_pflags &= norunbuf;
mtx_lock(&nblock);
if (fl != 0)
continue;
if ((needsbuffer & flags) == 0)
break;
}
if (msleep(&needsbuffer, &nblock,
(PRIBIO + 4) | slpflag, waitmsg, slptimeo)) {
mtx_unlock(&nblock);
return (NULL);
}
}
mtx_unlock(&nblock);
} else {
/*
* We finally have a valid bp. We aren't quite out of the
* woods, we still have to reserve kva space. In order
* to keep fragmentation sane we only allocate kva in
* BKVASIZE chunks.
*/
maxsize = (maxsize + BKVAMASK) & ~BKVAMASK;
if (maxsize != bp->b_kvasize) {
vm_offset_t addr = 0;
bfreekva(bp);
vm_map_lock(buffer_map);
if (vm_map_findspace(buffer_map,
vm_map_min(buffer_map), maxsize, &addr)) {
/*
* Uh oh. Buffer map is to fragmented. We
* must defragment the map.
*/
atomic_add_int(&bufdefragcnt, 1);
vm_map_unlock(buffer_map);
defrag = 1;
bp->b_flags |= B_INVAL;
brelse(bp);
goto restart;
}
if (addr) {
vm_map_insert(buffer_map, NULL, 0,
addr, addr + maxsize,
VM_PROT_ALL, VM_PROT_ALL, MAP_NOFAULT);
bp->b_kvabase = (caddr_t) addr;
bp->b_kvasize = maxsize;
atomic_add_long(&bufspace, bp->b_kvasize);
atomic_add_int(&bufreusecnt, 1);
}
vm_map_unlock(buffer_map);
}
bp->b_saveaddr = bp->b_kvabase;
bp->b_data = bp->b_saveaddr;
}
return(bp);
}
/*
* buf_daemon:
*
* buffer flushing daemon. Buffers are normally flushed by the
* update daemon but if it cannot keep up this process starts to
* take the load in an attempt to prevent getnewbuf() from blocking.
*/
static struct kproc_desc buf_kp = {
"bufdaemon",
buf_daemon,
&bufdaemonproc
};
SYSINIT(bufdaemon, SI_SUB_KTHREAD_BUF, SI_ORDER_FIRST, kproc_start, &buf_kp);
static int
buf_do_flush(struct vnode *vp)
{
int flushed;
flushed = flushbufqueues(vp, QUEUE_DIRTY, 0);
/* The list empty check here is slightly racy */
if (!TAILQ_EMPTY(&bufqueues[QUEUE_DIRTY_GIANT])) {
mtx_lock(&Giant);
flushed += flushbufqueues(vp, QUEUE_DIRTY_GIANT, 0);
mtx_unlock(&Giant);
}
if (flushed == 0) {
/*
* Could not find any buffers without rollback
* dependencies, so just write the first one
* in the hopes of eventually making progress.
*/
flushbufqueues(vp, QUEUE_DIRTY, 1);
if (!TAILQ_EMPTY(
&bufqueues[QUEUE_DIRTY_GIANT])) {
mtx_lock(&Giant);
flushbufqueues(vp, QUEUE_DIRTY_GIANT, 1);
mtx_unlock(&Giant);
}
}
return (flushed);
}
static void
buf_daemon()
{
int lodirtysave;
/*
* This process needs to be suspended prior to shutdown sync.
*/
EVENTHANDLER_REGISTER(shutdown_pre_sync, kproc_shutdown, bufdaemonproc,
SHUTDOWN_PRI_LAST);
/*
* This process is allowed to take the buffer cache to the limit
*/
curthread->td_pflags |= TDP_NORUNNINGBUF | TDP_BUFNEED;
mtx_lock(&bdlock);
for (;;) {
bd_request = 0;
mtx_unlock(&bdlock);
kproc_suspend_check(bufdaemonproc);
lodirtysave = lodirtybuffers;
if (bd_speedupreq) {
lodirtybuffers = numdirtybuffers / 2;
bd_speedupreq = 0;
}
/*
* Do the flush. Limit the amount of in-transit I/O we
* allow to build up, otherwise we would completely saturate
* the I/O system. Wakeup any waiting processes before we
* normally would so they can run in parallel with our drain.
*/
while (numdirtybuffers > lodirtybuffers) {
if (buf_do_flush(NULL) == 0)
break;
kern_yield(-1);
}
lodirtybuffers = lodirtysave;
/*
* Only clear bd_request if we have reached our low water
* mark. The buf_daemon normally waits 1 second and
* then incrementally flushes any dirty buffers that have
* built up, within reason.
*
* If we were unable to hit our low water mark and couldn't
* find any flushable buffers, we sleep half a second.
* Otherwise we loop immediately.
*/
mtx_lock(&bdlock);
if (numdirtybuffers <= lodirtybuffers) {
/*
* We reached our low water mark, reset the
* request and sleep until we are needed again.
* The sleep is just so the suspend code works.
*/
bd_request = 0;
msleep(&bd_request, &bdlock, PVM, "psleep", hz);
} else {
/*
* We couldn't find any flushable dirty buffers but
* still have too many dirty buffers, we
* have to sleep and try again. (rare)
*/
msleep(&bd_request, &bdlock, PVM, "qsleep", hz / 10);
}
}
}
/*
* flushbufqueues:
*
* Try to flush a buffer in the dirty queue. We must be careful to
* free up B_INVAL buffers instead of write them, which NFS is
* particularly sensitive to.
*/
static int flushwithdeps = 0;
SYSCTL_INT(_vfs, OID_AUTO, flushwithdeps, CTLFLAG_RW, &flushwithdeps,
0, "Number of buffers flushed with dependecies that require rollbacks");
static int
flushbufqueues(struct vnode *lvp, int queue, int flushdeps)
{
struct buf *sentinel;
struct vnode *vp;
struct mount *mp;
struct buf *bp;
int hasdeps;
int flushed;
int target;
if (lvp == NULL) {
target = numdirtybuffers - lodirtybuffers;
if (flushdeps && target > 2)
target /= 2;
} else
target = flushbufqtarget;
flushed = 0;
bp = NULL;
sentinel = malloc(sizeof(struct buf), M_TEMP, M_WAITOK | M_ZERO);
sentinel->b_qindex = QUEUE_SENTINEL;
mtx_lock(&bqlock);
TAILQ_INSERT_HEAD(&bufqueues[queue], sentinel, b_freelist);
while (flushed != target) {
bp = TAILQ_NEXT(sentinel, b_freelist);
if (bp != NULL) {
TAILQ_REMOVE(&bufqueues[queue], sentinel, b_freelist);
TAILQ_INSERT_AFTER(&bufqueues[queue], bp, sentinel,
b_freelist);
} else
break;
/*
* Skip sentinels inserted by other invocations of the
* flushbufqueues(), taking care to not reorder them.
*/
if (bp->b_qindex == QUEUE_SENTINEL)
continue;
/*
* Only flush the buffers that belong to the
* vnode locked by the curthread.
*/
if (lvp != NULL && bp->b_vp != lvp)
continue;
if (BUF_LOCK(bp, LK_EXCLUSIVE | LK_NOWAIT, NULL) != 0)
continue;
if (bp->b_pin_count > 0) {
BUF_UNLOCK(bp);
continue;
}
BO_LOCK(bp->b_bufobj);
if ((bp->b_vflags & BV_BKGRDINPROG) != 0 ||
(bp->b_flags & B_DELWRI) == 0) {
BO_UNLOCK(bp->b_bufobj);
BUF_UNLOCK(bp);
continue;
}
BO_UNLOCK(bp->b_bufobj);
if (bp->b_flags & B_INVAL) {
bremfreel(bp);
mtx_unlock(&bqlock);
brelse(bp);
flushed++;
numdirtywakeup((lodirtybuffers + hidirtybuffers) / 2);
mtx_lock(&bqlock);
continue;
}
if (!LIST_EMPTY(&bp->b_dep) && buf_countdeps(bp, 0)) {
if (flushdeps == 0) {
BUF_UNLOCK(bp);
continue;
}
hasdeps = 1;
} else
hasdeps = 0;
/*
* We must hold the lock on a vnode before writing
* one of its buffers. Otherwise we may confuse, or
* in the case of a snapshot vnode, deadlock the
* system.
*
* The lock order here is the reverse of the normal
* of vnode followed by buf lock. This is ok because
* the NOWAIT will prevent deadlock.
*/
vp = bp->b_vp;
if (vn_start_write(vp, &mp, V_NOWAIT) != 0) {
BUF_UNLOCK(bp);
continue;
}
if (vn_lock(vp, LK_EXCLUSIVE | LK_NOWAIT | LK_CANRECURSE) == 0) {
mtx_unlock(&bqlock);
CTR3(KTR_BUF, "flushbufqueue(%p) vp %p flags %X",
bp, bp->b_vp, bp->b_flags);
if (curproc == bufdaemonproc)
vfs_bio_awrite(bp);
else {
bremfree(bp);
bwrite(bp);
notbufdflashes++;
}
vn_finished_write(mp);
VOP_UNLOCK(vp, 0);
flushwithdeps += hasdeps;
flushed++;
/*
* Sleeping on runningbufspace while holding
* vnode lock leads to deadlock.
*/
if (curproc == bufdaemonproc)
waitrunningbufspace();
numdirtywakeup((lodirtybuffers + hidirtybuffers) / 2);
mtx_lock(&bqlock);
continue;
}
vn_finished_write(mp);
BUF_UNLOCK(bp);
}
TAILQ_REMOVE(&bufqueues[queue], sentinel, b_freelist);
mtx_unlock(&bqlock);
free(sentinel, M_TEMP);
return (flushed);
}
/*
* Check to see if a block is currently memory resident.
*/
struct buf *
incore(struct bufobj *bo, daddr_t blkno)
{
struct buf *bp;
BO_LOCK(bo);
bp = gbincore(bo, blkno);
BO_UNLOCK(bo);
return (bp);
}
/*
* Returns true if no I/O is needed to access the
* associated VM object. This is like incore except
* it also hunts around in the VM system for the data.
*/
static int
inmem(struct vnode * vp, daddr_t blkno)
{
vm_object_t obj;
vm_offset_t toff, tinc, size;
vm_page_t m;
vm_ooffset_t off;
ASSERT_VOP_LOCKED(vp, "inmem");
if (incore(&vp->v_bufobj, blkno))
return 1;
if (vp->v_mount == NULL)
return 0;
obj = vp->v_object;
if (obj == NULL)
return (0);
size = PAGE_SIZE;
if (size > vp->v_mount->mnt_stat.f_iosize)
size = vp->v_mount->mnt_stat.f_iosize;
off = (vm_ooffset_t)blkno * (vm_ooffset_t)vp->v_mount->mnt_stat.f_iosize;
VM_OBJECT_LOCK(obj);
for (toff = 0; toff < vp->v_mount->mnt_stat.f_iosize; toff += tinc) {
m = vm_page_lookup(obj, OFF_TO_IDX(off + toff));
if (!m)
goto notinmem;
tinc = size;
if (tinc > PAGE_SIZE - ((toff + off) & PAGE_MASK))
tinc = PAGE_SIZE - ((toff + off) & PAGE_MASK);
if (vm_page_is_valid(m,
(vm_offset_t) ((toff + off) & PAGE_MASK), tinc) == 0)
goto notinmem;
}
VM_OBJECT_UNLOCK(obj);
return 1;
notinmem:
VM_OBJECT_UNLOCK(obj);
return (0);
}
/*
* Set the dirty range for a buffer based on the status of the dirty
* bits in the pages comprising the buffer. The range is limited
* to the size of the buffer.
*
* Tell the VM system that the pages associated with this buffer
* are clean. This is used for delayed writes where the data is
* going to go to disk eventually without additional VM intevention.
*
* Note that while we only really need to clean through to b_bcount, we
* just go ahead and clean through to b_bufsize.
*/
static void
vfs_clean_pages_dirty_buf(struct buf *bp)
{
vm_ooffset_t foff, noff, eoff;
vm_page_t m;
int i;
if ((bp->b_flags & B_VMIO) == 0 || bp->b_bufsize == 0)
return;
foff = bp->b_offset;
KASSERT(bp->b_offset != NOOFFSET,
("vfs_clean_pages_dirty_buf: no buffer offset"));
VM_OBJECT_LOCK(bp->b_bufobj->bo_object);
vfs_drain_busy_pages(bp);
vfs_setdirty_locked_object(bp);
for (i = 0; i < bp->b_npages; i++) {
noff = (foff + PAGE_SIZE) & ~(off_t)PAGE_MASK;
eoff = noff;
if (eoff > bp->b_offset + bp->b_bufsize)
eoff = bp->b_offset + bp->b_bufsize;
m = bp->b_pages[i];
vfs_page_set_validclean(bp, foff, m);
/* vm_page_clear_dirty(m, foff & PAGE_MASK, eoff - foff); */
foff = noff;
}
VM_OBJECT_UNLOCK(bp->b_bufobj->bo_object);
}
static void
vfs_setdirty_locked_object(struct buf *bp)
{
vm_object_t object;
int i;
object = bp->b_bufobj->bo_object;
VM_OBJECT_LOCK_ASSERT(object, MA_OWNED);
/*
* We qualify the scan for modified pages on whether the
* object has been flushed yet.
*/
if ((object->flags & OBJ_MIGHTBEDIRTY) != 0) {
vm_offset_t boffset;
vm_offset_t eoffset;
/*
* test the pages to see if they have been modified directly
* by users through the VM system.
*/
for (i = 0; i < bp->b_npages; i++)
vm_page_test_dirty(bp->b_pages[i]);
/*
* Calculate the encompassing dirty range, boffset and eoffset,
* (eoffset - boffset) bytes.
*/
for (i = 0; i < bp->b_npages; i++) {
if (bp->b_pages[i]->dirty)
break;
}
boffset = (i << PAGE_SHIFT) - (bp->b_offset & PAGE_MASK);
for (i = bp->b_npages - 1; i >= 0; --i) {
if (bp->b_pages[i]->dirty) {
break;
}
}
eoffset = ((i + 1) << PAGE_SHIFT) - (bp->b_offset & PAGE_MASK);
/*
* Fit it to the buffer.
*/
if (eoffset > bp->b_bcount)
eoffset = bp->b_bcount;
/*
* If we have a good dirty range, merge with the existing
* dirty range.
*/
if (boffset < eoffset) {
if (bp->b_dirtyoff > boffset)
bp->b_dirtyoff = boffset;
if (bp->b_dirtyend < eoffset)
bp->b_dirtyend = eoffset;
}
}
}
/*
* getblk:
*
* Get a block given a specified block and offset into a file/device.
* The buffers B_DONE bit will be cleared on return, making it almost
* ready for an I/O initiation. B_INVAL may or may not be set on
* return. The caller should clear B_INVAL prior to initiating a
* READ.
*
* For a non-VMIO buffer, B_CACHE is set to the opposite of B_INVAL for
* an existing buffer.
*
* For a VMIO buffer, B_CACHE is modified according to the backing VM.
* If getblk()ing a previously 0-sized invalid buffer, B_CACHE is set
* and then cleared based on the backing VM. If the previous buffer is
* non-0-sized but invalid, B_CACHE will be cleared.
*
* If getblk() must create a new buffer, the new buffer is returned with
* both B_INVAL and B_CACHE clear unless it is a VMIO buffer, in which
* case it is returned with B_INVAL clear and B_CACHE set based on the
* backing VM.
*
* getblk() also forces a bwrite() for any B_DELWRI buffer whos
* B_CACHE bit is clear.
*
* What this means, basically, is that the caller should use B_CACHE to
* determine whether the buffer is fully valid or not and should clear
* B_INVAL prior to issuing a read. If the caller intends to validate
* the buffer by loading its data area with something, the caller needs
* to clear B_INVAL. If the caller does this without issuing an I/O,
* the caller should set B_CACHE ( as an optimization ), else the caller
* should issue the I/O and biodone() will set B_CACHE if the I/O was
* a write attempt or if it was a successfull read. If the caller
* intends to issue a READ, the caller must clear B_INVAL and BIO_ERROR
* prior to issuing the READ. biodone() will *not* clear B_INVAL.
*/
struct buf *
getblk(struct vnode * vp, daddr_t blkno, int size, int slpflag, int slptimeo,
int flags)
{
struct buf *bp;
struct bufobj *bo;
int error;
CTR3(KTR_BUF, "getblk(%p, %ld, %d)", vp, (long)blkno, size);
ASSERT_VOP_LOCKED(vp, "getblk");
if (size > MAXBSIZE)
panic("getblk: size(%d) > MAXBSIZE(%d)\n", size, MAXBSIZE);
bo = &vp->v_bufobj;
loop:
/*
* Block if we are low on buffers. Certain processes are allowed
* to completely exhaust the buffer cache.
*
* If this check ever becomes a bottleneck it may be better to
* move it into the else, when gbincore() fails. At the moment
* it isn't a problem.
*
* XXX remove if 0 sections (clean this up after its proven)
*/
if (numfreebuffers == 0) {
if (TD_IS_IDLETHREAD(curthread))
return NULL;
mtx_lock(&nblock);
needsbuffer |= VFS_BIO_NEED_ANY;
mtx_unlock(&nblock);
}
BO_LOCK(bo);
bp = gbincore(bo, blkno);
if (bp != NULL) {
int lockflags;
/*
* Buffer is in-core. If the buffer is not busy, it must
* be on a queue.
*/
lockflags = LK_EXCLUSIVE | LK_SLEEPFAIL | LK_INTERLOCK;
if (flags & GB_LOCK_NOWAIT)
lockflags |= LK_NOWAIT;
error = BUF_TIMELOCK(bp, lockflags,
BO_MTX(bo), "getblk", slpflag, slptimeo);
/*
* If we slept and got the lock we have to restart in case
* the buffer changed identities.
*/
if (error == ENOLCK)
goto loop;
/* We timed out or were interrupted. */
else if (error)
return (NULL);
/*
* The buffer is locked. B_CACHE is cleared if the buffer is
* invalid. Otherwise, for a non-VMIO buffer, B_CACHE is set
* and for a VMIO buffer B_CACHE is adjusted according to the
* backing VM cache.
*/
if (bp->b_flags & B_INVAL)
bp->b_flags &= ~B_CACHE;
else if ((bp->b_flags & (B_VMIO | B_INVAL)) == 0)
bp->b_flags |= B_CACHE;
BO_LOCK(bo);
bremfree(bp);
BO_UNLOCK(bo);
/*
* check for size inconsistancies for non-VMIO case.
*/
if (bp->b_bcount != size) {
if ((bp->b_flags & B_VMIO) == 0 ||
(size > bp->b_kvasize)) {
if (bp->b_flags & B_DELWRI) {
/*
* If buffer is pinned and caller does
* not want sleep waiting for it to be
* unpinned, bail out
* */
if (bp->b_pin_count > 0) {
if (flags & GB_LOCK_NOWAIT) {
bqrelse(bp);
return (NULL);
} else {
bunpin_wait(bp);
}
}
bp->b_flags |= B_NOCACHE;
bwrite(bp);
} else {
if (LIST_EMPTY(&bp->b_dep)) {
bp->b_flags |= B_RELBUF;
brelse(bp);
} else {
bp->b_flags |= B_NOCACHE;
bwrite(bp);
}
}
goto loop;
}
}
/*
* If the size is inconsistant in the VMIO case, we can resize
* the buffer. This might lead to B_CACHE getting set or
* cleared. If the size has not changed, B_CACHE remains
* unchanged from its previous state.
*/
if (bp->b_bcount != size)
allocbuf(bp, size);
KASSERT(bp->b_offset != NOOFFSET,
("getblk: no buffer offset"));
/*
* A buffer with B_DELWRI set and B_CACHE clear must
* be committed before we can return the buffer in
* order to prevent the caller from issuing a read
* ( due to B_CACHE not being set ) and overwriting
* it.
*
* Most callers, including NFS and FFS, need this to
* operate properly either because they assume they
* can issue a read if B_CACHE is not set, or because
* ( for example ) an uncached B_DELWRI might loop due
* to softupdates re-dirtying the buffer. In the latter
* case, B_CACHE is set after the first write completes,
* preventing further loops.
* NOTE! b*write() sets B_CACHE. If we cleared B_CACHE
* above while extending the buffer, we cannot allow the
* buffer to remain with B_CACHE set after the write
* completes or it will represent a corrupt state. To
* deal with this we set B_NOCACHE to scrap the buffer
* after the write.
*
* We might be able to do something fancy, like setting
* B_CACHE in bwrite() except if B_DELWRI is already set,
* so the below call doesn't set B_CACHE, but that gets real
* confusing. This is much easier.
*/
if ((bp->b_flags & (B_CACHE|B_DELWRI)) == B_DELWRI) {
bp->b_flags |= B_NOCACHE;
bwrite(bp);
goto loop;
}
bp->b_flags &= ~B_DONE;
} else {
int bsize, maxsize, vmio;
off_t offset;
/*
* Buffer is not in-core, create new buffer. The buffer
* returned by getnewbuf() is locked. Note that the returned
* buffer is also considered valid (not marked B_INVAL).
*/
BO_UNLOCK(bo);
/*
* If the user does not want us to create the buffer, bail out
* here.
*/
if (flags & GB_NOCREAT)
return NULL;
bsize = vn_isdisk(vp, NULL) ? DEV_BSIZE : bo->bo_bsize;
offset = blkno * bsize;
vmio = vp->v_object != NULL;
maxsize = vmio ? size + (offset & PAGE_MASK) : size;
maxsize = imax(maxsize, bsize);
bp = getnewbuf(vp, slpflag, slptimeo, size, maxsize, flags);
if (bp == NULL) {
if (slpflag || slptimeo)
return NULL;
goto loop;
}
/*
* This code is used to make sure that a buffer is not
* created while the getnewbuf routine is blocked.
* This can be a problem whether the vnode is locked or not.
* If the buffer is created out from under us, we have to
* throw away the one we just created.
*
* Note: this must occur before we associate the buffer
* with the vp especially considering limitations in
* the splay tree implementation when dealing with duplicate
* lblkno's.
*/
BO_LOCK(bo);
if (gbincore(bo, blkno)) {
BO_UNLOCK(bo);
bp->b_flags |= B_INVAL;
brelse(bp);
goto loop;
}
/*
* Insert the buffer into the hash, so that it can
* be found by incore.
*/
bp->b_blkno = bp->b_lblkno = blkno;
bp->b_offset = offset;
bgetvp(vp, bp);
BO_UNLOCK(bo);
/*
* set B_VMIO bit. allocbuf() the buffer bigger. Since the
* buffer size starts out as 0, B_CACHE will be set by
* allocbuf() for the VMIO case prior to it testing the
* backing store for validity.
*/
if (vmio) {
bp->b_flags |= B_VMIO;
KASSERT(vp->v_object == bp->b_bufobj->bo_object,
("ARGH! different b_bufobj->bo_object %p %p %p\n",
bp, vp->v_object, bp->b_bufobj->bo_object));
} else {
bp->b_flags &= ~B_VMIO;
KASSERT(bp->b_bufobj->bo_object == NULL,
("ARGH! has b_bufobj->bo_object %p %p\n",
bp, bp->b_bufobj->bo_object));
}
allocbuf(bp, size);
bp->b_flags &= ~B_DONE;
}
CTR4(KTR_BUF, "getblk(%p, %ld, %d) = %p", vp, (long)blkno, size, bp);
BUF_ASSERT_HELD(bp);
KASSERT(bp->b_bufobj == bo,
("bp %p wrong b_bufobj %p should be %p", bp, bp->b_bufobj, bo));
return (bp);
}
/*
* Get an empty, disassociated buffer of given size. The buffer is initially
* set to B_INVAL.
*/
struct buf *
geteblk(int size, int flags)
{
struct buf *bp;
int maxsize;
maxsize = (size + BKVAMASK) & ~BKVAMASK;
while ((bp = getnewbuf(NULL, 0, 0, size, maxsize, flags)) == NULL) {
if ((flags & GB_NOWAIT_BD) &&
(curthread->td_pflags & TDP_BUFNEED) != 0)
return (NULL);
}
allocbuf(bp, size);
bp->b_flags |= B_INVAL; /* b_dep cleared by getnewbuf() */
BUF_ASSERT_HELD(bp);
return (bp);
}
/*
* This code constitutes the buffer memory from either anonymous system
* memory (in the case of non-VMIO operations) or from an associated
* VM object (in the case of VMIO operations). This code is able to
* resize a buffer up or down.
*
* Note that this code is tricky, and has many complications to resolve
* deadlock or inconsistant data situations. Tread lightly!!!
* There are B_CACHE and B_DELWRI interactions that must be dealt with by
* the caller. Calling this code willy nilly can result in the loss of data.
*
* allocbuf() only adjusts B_CACHE for VMIO buffers. getblk() deals with
* B_CACHE for the non-VMIO case.
*/
int
allocbuf(struct buf *bp, int size)
{
int newbsize, mbsize;
int i;
BUF_ASSERT_HELD(bp);
if (bp->b_kvasize < size)
panic("allocbuf: buffer too small");
if ((bp->b_flags & B_VMIO) == 0) {
caddr_t origbuf;
int origbufsize;
/*
* Just get anonymous memory from the kernel. Don't
* mess with B_CACHE.
*/
mbsize = (size + DEV_BSIZE - 1) & ~(DEV_BSIZE - 1);
if (bp->b_flags & B_MALLOC)
newbsize = mbsize;
else
newbsize = round_page(size);
if (newbsize < bp->b_bufsize) {
/*
* malloced buffers are not shrunk
*/
if (bp->b_flags & B_MALLOC) {
if (newbsize) {
bp->b_bcount = size;
} else {
free(bp->b_data, M_BIOBUF);
if (bp->b_bufsize) {
atomic_subtract_long(
&bufmallocspace,
bp->b_bufsize);
bufspacewakeup();
bp->b_bufsize = 0;
}
bp->b_saveaddr = bp->b_kvabase;
bp->b_data = bp->b_saveaddr;
bp->b_bcount = 0;
bp->b_flags &= ~B_MALLOC;
}
return 1;
}
vm_hold_free_pages(bp, newbsize);
} else if (newbsize > bp->b_bufsize) {
/*
* We only use malloced memory on the first allocation.
* and revert to page-allocated memory when the buffer
* grows.
*/
/*
* There is a potential smp race here that could lead
* to bufmallocspace slightly passing the max. It
* is probably extremely rare and not worth worrying
* over.
*/
if ( (bufmallocspace < maxbufmallocspace) &&
(bp->b_bufsize == 0) &&
(mbsize <= PAGE_SIZE/2)) {
bp->b_data = malloc(mbsize, M_BIOBUF, M_WAITOK);
bp->b_bufsize = mbsize;
bp->b_bcount = size;
bp->b_flags |= B_MALLOC;
atomic_add_long(&bufmallocspace, mbsize);
return 1;
}
origbuf = NULL;
origbufsize = 0;
/*
* If the buffer is growing on its other-than-first allocation,
* then we revert to the page-allocation scheme.
*/
if (bp->b_flags & B_MALLOC) {
origbuf = bp->b_data;
origbufsize = bp->b_bufsize;
bp->b_data = bp->b_kvabase;
if (bp->b_bufsize) {
atomic_subtract_long(&bufmallocspace,
bp->b_bufsize);
bufspacewakeup();
bp->b_bufsize = 0;
}
bp->b_flags &= ~B_MALLOC;
newbsize = round_page(newbsize);
}
vm_hold_load_pages(
bp,
(vm_offset_t) bp->b_data + bp->b_bufsize,
(vm_offset_t) bp->b_data + newbsize);
if (origbuf) {
bcopy(origbuf, bp->b_data, origbufsize);
free(origbuf, M_BIOBUF);
}
}
} else {
int desiredpages;
newbsize = (size + DEV_BSIZE - 1) & ~(DEV_BSIZE - 1);
desiredpages = (size == 0) ? 0 :
num_pages((bp->b_offset & PAGE_MASK) + newbsize);
if (bp->b_flags & B_MALLOC)
panic("allocbuf: VMIO buffer can't be malloced");
/*
* Set B_CACHE initially if buffer is 0 length or will become
* 0-length.
*/
if (size == 0 || bp->b_bufsize == 0)
bp->b_flags |= B_CACHE;
if (newbsize < bp->b_bufsize) {
/*
* DEV_BSIZE aligned new buffer size is less then the
* DEV_BSIZE aligned existing buffer size. Figure out
* if we have to remove any pages.
*/
if (desiredpages < bp->b_npages) {
vm_page_t m;
VM_OBJECT_LOCK(bp->b_bufobj->bo_object);
for (i = desiredpages; i < bp->b_npages; i++) {
/*
* the page is not freed here -- it
* is the responsibility of
* vnode_pager_setsize
*/
m = bp->b_pages[i];
KASSERT(m != bogus_page,
("allocbuf: bogus page found"));
while (vm_page_sleep_if_busy(m, TRUE,
"biodep"))
continue;
bp->b_pages[i] = NULL;
vm_page_lock(m);
vm_page_unwire(m, 0);
vm_page_unlock(m);
}
VM_OBJECT_UNLOCK(bp->b_bufobj->bo_object);
pmap_qremove((vm_offset_t) trunc_page((vm_offset_t)bp->b_data) +
(desiredpages << PAGE_SHIFT), (bp->b_npages - desiredpages));
bp->b_npages = desiredpages;
}
} else if (size > bp->b_bcount) {
/*
* We are growing the buffer, possibly in a
* byte-granular fashion.
*/
vm_object_t obj;
vm_offset_t toff;
vm_offset_t tinc;
/*
* Step 1, bring in the VM pages from the object,
* allocating them if necessary. We must clear
* B_CACHE if these pages are not valid for the
* range covered by the buffer.
*/
obj = bp->b_bufobj->bo_object;
VM_OBJECT_LOCK(obj);
while (bp->b_npages < desiredpages) {
vm_page_t m;
/*
* We must allocate system pages since blocking
* here could intefere with paging I/O, no
* matter which process we are.
*
* We can only test VPO_BUSY here. Blocking on
* m->busy might lead to a deadlock:
* vm_fault->getpages->cluster_read->allocbuf
* Thus, we specify VM_ALLOC_IGN_SBUSY.
*/
m = vm_page_grab(obj, OFF_TO_IDX(bp->b_offset) +
bp->b_npages, VM_ALLOC_NOBUSY |
VM_ALLOC_SYSTEM | VM_ALLOC_WIRED |
VM_ALLOC_RETRY | VM_ALLOC_IGN_SBUSY |
VM_ALLOC_COUNT(desiredpages - bp->b_npages));
if (m->valid == 0)
bp->b_flags &= ~B_CACHE;
bp->b_pages[bp->b_npages] = m;
++bp->b_npages;
}
/*
* Step 2. We've loaded the pages into the buffer,
* we have to figure out if we can still have B_CACHE
* set. Note that B_CACHE is set according to the
* byte-granular range ( bcount and size ), new the
* aligned range ( newbsize ).
*
* The VM test is against m->valid, which is DEV_BSIZE
* aligned. Needless to say, the validity of the data
* needs to also be DEV_BSIZE aligned. Note that this
* fails with NFS if the server or some other client
* extends the file's EOF. If our buffer is resized,
* B_CACHE may remain set! XXX
*/
toff = bp->b_bcount;
tinc = PAGE_SIZE - ((bp->b_offset + toff) & PAGE_MASK);
while ((bp->b_flags & B_CACHE) && toff < size) {
vm_pindex_t pi;
if (tinc > (size - toff))
tinc = size - toff;
pi = ((bp->b_offset & PAGE_MASK) + toff) >>
PAGE_SHIFT;
vfs_buf_test_cache(
bp,
bp->b_offset,
toff,
tinc,
bp->b_pages[pi]
);
toff += tinc;
tinc = PAGE_SIZE;
}
VM_OBJECT_UNLOCK(obj);
/*
* Step 3, fixup the KVM pmap. Remember that
* bp->b_data is relative to bp->b_offset, but
* bp->b_offset may be offset into the first page.
*/
bp->b_data = (caddr_t)
trunc_page((vm_offset_t)bp->b_data);
pmap_qenter(
(vm_offset_t)bp->b_data,
bp->b_pages,
bp->b_npages
);
bp->b_data = (caddr_t)((vm_offset_t)bp->b_data |
(vm_offset_t)(bp->b_offset & PAGE_MASK));
}
}
if (newbsize < bp->b_bufsize)
bufspacewakeup();
bp->b_bufsize = newbsize; /* actual buffer allocation */
bp->b_bcount = size; /* requested buffer size */
return 1;
}
void
biodone(struct bio *bp)
{
struct mtx *mtxp;
void (*done)(struct bio *);
mtxp = mtx_pool_find(mtxpool_sleep, bp);
mtx_lock(mtxp);
bp->bio_flags |= BIO_DONE;
done = bp->bio_done;
if (done == NULL)
wakeup(bp);
mtx_unlock(mtxp);
if (done != NULL)
done(bp);
}
/*
* Wait for a BIO to finish.
*
* XXX: resort to a timeout for now. The optimal locking (if any) for this
* case is not yet clear.
*/
int
biowait(struct bio *bp, const char *wchan)
{
struct mtx *mtxp;
mtxp = mtx_pool_find(mtxpool_sleep, bp);
mtx_lock(mtxp);
while ((bp->bio_flags & BIO_DONE) == 0)
msleep(bp, mtxp, PRIBIO, wchan, hz / 10);
mtx_unlock(mtxp);
if (bp->bio_error != 0)
return (bp->bio_error);
if (!(bp->bio_flags & BIO_ERROR))
return (0);
return (EIO);
}
void
biofinish(struct bio *bp, struct devstat *stat, int error)
{
if (error) {
bp->bio_error = error;
bp->bio_flags |= BIO_ERROR;
}
if (stat != NULL)
devstat_end_transaction_bio(stat, bp);
biodone(bp);
}
/*
* bufwait:
*
* Wait for buffer I/O completion, returning error status. The buffer
* is left locked and B_DONE on return. B_EINTR is converted into an EINTR
* error and cleared.
*/
int
bufwait(struct buf *bp)
{
if (bp->b_iocmd == BIO_READ)
bwait(bp, PRIBIO, "biord");
else
bwait(bp, PRIBIO, "biowr");
if (bp->b_flags & B_EINTR) {
bp->b_flags &= ~B_EINTR;
return (EINTR);
}
if (bp->b_ioflags & BIO_ERROR) {
return (bp->b_error ? bp->b_error : EIO);
} else {
return (0);
}
}
/*
* Call back function from struct bio back up to struct buf.
*/
static void
bufdonebio(struct bio *bip)
{
struct buf *bp;
bp = bip->bio_caller2;
bp->b_resid = bp->b_bcount - bip->bio_completed;
bp->b_resid = bip->bio_resid; /* XXX: remove */
bp->b_ioflags = bip->bio_flags;
bp->b_error = bip->bio_error;
if (bp->b_error)
bp->b_ioflags |= BIO_ERROR;
bufdone(bp);
g_destroy_bio(bip);
}
void
dev_strategy(struct cdev *dev, struct buf *bp)
{
struct cdevsw *csw;
struct bio *bip;
int ref;
if ((!bp->b_iocmd) || (bp->b_iocmd & (bp->b_iocmd - 1)))
panic("b_iocmd botch");
for (;;) {
bip = g_new_bio();
if (bip != NULL)
break;
/* Try again later */
tsleep(&bp, PRIBIO, "dev_strat", hz/10);
}
bip->bio_cmd = bp->b_iocmd;
bip->bio_offset = bp->b_iooffset;
bip->bio_length = bp->b_bcount;
bip->bio_bcount = bp->b_bcount; /* XXX: remove */
bip->bio_data = bp->b_data;
bip->bio_done = bufdonebio;
bip->bio_caller2 = bp;
bip->bio_dev = dev;
KASSERT(dev->si_refcount > 0,
("dev_strategy on un-referenced struct cdev *(%s)",
devtoname(dev)));
csw = dev_refthread(dev, &ref);
if (csw == NULL) {
g_destroy_bio(bip);
bp->b_error = ENXIO;
bp->b_ioflags = BIO_ERROR;
bufdone(bp);
return;
}
(*csw->d_strategy)(bip);
dev_relthread(dev, ref);
}
/*
* bufdone:
*
* Finish I/O on a buffer, optionally calling a completion function.
* This is usually called from an interrupt so process blocking is
* not allowed.
*
* biodone is also responsible for setting B_CACHE in a B_VMIO bp.
* In a non-VMIO bp, B_CACHE will be set on the next getblk()
* assuming B_INVAL is clear.
*
* For the VMIO case, we set B_CACHE if the op was a read and no
* read error occured, or if the op was a write. B_CACHE is never
* set if the buffer is invalid or otherwise uncacheable.
*
* biodone does not mess with B_INVAL, allowing the I/O routine or the
* initiator to leave B_INVAL set to brelse the buffer out of existance
* in the biodone routine.
*/
void
bufdone(struct buf *bp)
{
struct bufobj *dropobj;
void (*biodone)(struct buf *);
CTR3(KTR_BUF, "bufdone(%p) vp %p flags %X", bp, bp->b_vp, bp->b_flags);
dropobj = NULL;
KASSERT(!(bp->b_flags & B_DONE), ("biodone: bp %p already done", bp));
BUF_ASSERT_HELD(bp);
runningbufwakeup(bp);
if (bp->b_iocmd == BIO_WRITE)
dropobj = bp->b_bufobj;
/* call optional completion function if requested */
if (bp->b_iodone != NULL) {
biodone = bp->b_iodone;
bp->b_iodone = NULL;
(*biodone) (bp);
if (dropobj)
bufobj_wdrop(dropobj);
return;
}
bufdone_finish(bp);
if (dropobj)
bufobj_wdrop(dropobj);
}
void
bufdone_finish(struct buf *bp)
{
BUF_ASSERT_HELD(bp);
if (!LIST_EMPTY(&bp->b_dep))
buf_complete(bp);
if (bp->b_flags & B_VMIO) {
vm_ooffset_t foff;
vm_page_t m;
vm_object_t obj;
struct vnode *vp;
int bogus, i, iosize;
obj = bp->b_bufobj->bo_object;
KASSERT(obj->paging_in_progress >= bp->b_npages,
("biodone_finish: paging in progress(%d) < b_npages(%d)",
obj->paging_in_progress, bp->b_npages));
vp = bp->b_vp;
KASSERT(vp->v_holdcnt > 0,
("biodone_finish: vnode %p has zero hold count", vp));
KASSERT(vp->v_object != NULL,
("biodone_finish: vnode %p has no vm_object", vp));
foff = bp->b_offset;
KASSERT(bp->b_offset != NOOFFSET,
("biodone_finish: bp %p has no buffer offset", bp));
/*
* Set B_CACHE if the op was a normal read and no error
* occured. B_CACHE is set for writes in the b*write()
* routines.
*/
iosize = bp->b_bcount - bp->b_resid;
if (bp->b_iocmd == BIO_READ &&
!(bp->b_flags & (B_INVAL|B_NOCACHE)) &&
!(bp->b_ioflags & BIO_ERROR)) {
bp->b_flags |= B_CACHE;
}
bogus = 0;
VM_OBJECT_LOCK(obj);
for (i = 0; i < bp->b_npages; i++) {
int bogusflag = 0;
int resid;
resid = ((foff + PAGE_SIZE) & ~(off_t)PAGE_MASK) - foff;
if (resid > iosize)
resid = iosize;
/*
* cleanup bogus pages, restoring the originals
*/
m = bp->b_pages[i];
if (m == bogus_page) {
bogus = bogusflag = 1;
m = vm_page_lookup(obj, OFF_TO_IDX(foff));
if (m == NULL)
panic("biodone: page disappeared!");
bp->b_pages[i] = m;
}
KASSERT(OFF_TO_IDX(foff) == m->pindex,
("biodone_finish: foff(%jd)/pindex(%ju) mismatch",
(intmax_t)foff, (uintmax_t)m->pindex));
/*
* In the write case, the valid and clean bits are
* already changed correctly ( see bdwrite() ), so we
* only need to do this here in the read case.
*/
if ((bp->b_iocmd == BIO_READ) && !bogusflag && resid > 0) {
KASSERT((m->dirty & vm_page_bits(foff &
PAGE_MASK, resid)) == 0, ("bufdone_finish:"
" page %p has unexpected dirty bits", m));
vfs_page_set_valid(bp, foff, m);
}
vm_page_io_finish(m);
vm_object_pip_subtract(obj, 1);
foff = (foff + PAGE_SIZE) & ~(off_t)PAGE_MASK;
iosize -= resid;
}
vm_object_pip_wakeupn(obj, 0);
VM_OBJECT_UNLOCK(obj);
if (bogus)
pmap_qenter(trunc_page((vm_offset_t)bp->b_data),
bp->b_pages, bp->b_npages);
}
/*
* For asynchronous completions, release the buffer now. The brelse
* will do a wakeup there if necessary - so no need to do a wakeup
* here in the async case. The sync case always needs to do a wakeup.
*/
if (bp->b_flags & B_ASYNC) {
if ((bp->b_flags & (B_NOCACHE | B_INVAL | B_RELBUF)) || (bp->b_ioflags & BIO_ERROR))
brelse(bp);
else
bqrelse(bp);
} else
bdone(bp);
}
/*
* This routine is called in lieu of iodone in the case of
* incomplete I/O. This keeps the busy status for pages
* consistant.
*/
void
vfs_unbusy_pages(struct buf *bp)
{
int i;
vm_object_t obj;
vm_page_t m;
runningbufwakeup(bp);
if (!(bp->b_flags & B_VMIO))
return;
obj = bp->b_bufobj->bo_object;
VM_OBJECT_LOCK(obj);
for (i = 0; i < bp->b_npages; i++) {
m = bp->b_pages[i];
if (m == bogus_page) {
m = vm_page_lookup(obj, OFF_TO_IDX(bp->b_offset) + i);
if (!m)
panic("vfs_unbusy_pages: page missing\n");
bp->b_pages[i] = m;
pmap_qenter(trunc_page((vm_offset_t)bp->b_data),
bp->b_pages, bp->b_npages);
}
vm_object_pip_subtract(obj, 1);
vm_page_io_finish(m);
}
vm_object_pip_wakeupn(obj, 0);
VM_OBJECT_UNLOCK(obj);
}
/*
* vfs_page_set_valid:
*
* Set the valid bits in a page based on the supplied offset. The
* range is restricted to the buffer's size.
*
* This routine is typically called after a read completes.
*/
static void
vfs_page_set_valid(struct buf *bp, vm_ooffset_t off, vm_page_t m)
{
vm_ooffset_t eoff;
/*
* Compute the end offset, eoff, such that [off, eoff) does not span a
* page boundary and eoff is not greater than the end of the buffer.
* The end of the buffer, in this case, is our file EOF, not the
* allocation size of the buffer.
*/
eoff = (off + PAGE_SIZE) & ~(vm_ooffset_t)PAGE_MASK;
if (eoff > bp->b_offset + bp->b_bcount)
eoff = bp->b_offset + bp->b_bcount;
/*
* Set valid range. This is typically the entire buffer and thus the
* entire page.
*/
if (eoff > off)
vm_page_set_valid(m, off & PAGE_MASK, eoff - off);
}
/*
* vfs_page_set_validclean:
*
* Set the valid bits and clear the dirty bits in a page based on the
* supplied offset. The range is restricted to the buffer's size.
*/
static void
vfs_page_set_validclean(struct buf *bp, vm_ooffset_t off, vm_page_t m)
{
vm_ooffset_t soff, eoff;
/*
* Start and end offsets in buffer. eoff - soff may not cross a
* page boundry or cross the end of the buffer. The end of the
* buffer, in this case, is our file EOF, not the allocation size
* of the buffer.
*/
soff = off;
eoff = (off + PAGE_SIZE) & ~(off_t)PAGE_MASK;
if (eoff > bp->b_offset + bp->b_bcount)
eoff = bp->b_offset + bp->b_bcount;
/*
* Set valid range. This is typically the entire buffer and thus the
* entire page.
*/
if (eoff > soff) {
vm_page_set_validclean(
m,
(vm_offset_t) (soff & PAGE_MASK),
(vm_offset_t) (eoff - soff)
);
}
}
/*
* Ensure that all buffer pages are not busied by VPO_BUSY flag. If
* any page is busy, drain the flag.
*/
static void
vfs_drain_busy_pages(struct buf *bp)
{
vm_page_t m;
int i, last_busied;
VM_OBJECT_LOCK_ASSERT(bp->b_bufobj->bo_object, MA_OWNED);
last_busied = 0;
for (i = 0; i < bp->b_npages; i++) {
m = bp->b_pages[i];
if ((m->oflags & VPO_BUSY) != 0) {
for (; last_busied < i; last_busied++)
vm_page_busy(bp->b_pages[last_busied]);
while ((m->oflags & VPO_BUSY) != 0)
vm_page_sleep(m, "vbpage");
}
}
for (i = 0; i < last_busied; i++)
vm_page_wakeup(bp->b_pages[i]);
}
/*
* This routine is called before a device strategy routine.
* It is used to tell the VM system that paging I/O is in
* progress, and treat the pages associated with the buffer
* almost as being VPO_BUSY. Also the object paging_in_progress
* flag is handled to make sure that the object doesn't become
* inconsistant.
*
* Since I/O has not been initiated yet, certain buffer flags
* such as BIO_ERROR or B_INVAL may be in an inconsistant state
* and should be ignored.
*/
void
vfs_busy_pages(struct buf *bp, int clear_modify)
{
int i, bogus;
vm_object_t obj;
vm_ooffset_t foff;
vm_page_t m;
if (!(bp->b_flags & B_VMIO))
return;
obj = bp->b_bufobj->bo_object;
foff = bp->b_offset;
KASSERT(bp->b_offset != NOOFFSET,
("vfs_busy_pages: no buffer offset"));
VM_OBJECT_LOCK(obj);
vfs_drain_busy_pages(bp);
if (bp->b_bufsize != 0)
vfs_setdirty_locked_object(bp);
bogus = 0;
for (i = 0; i < bp->b_npages; i++) {
m = bp->b_pages[i];
if ((bp->b_flags & B_CLUSTER) == 0) {
vm_object_pip_add(obj, 1);
vm_page_io_start(m);
}
/*
* When readying a buffer for a read ( i.e
* clear_modify == 0 ), it is important to do
* bogus_page replacement for valid pages in
* partially instantiated buffers. Partially
* instantiated buffers can, in turn, occur when
* reconstituting a buffer from its VM backing store
* base. We only have to do this if B_CACHE is
* clear ( which causes the I/O to occur in the
* first place ). The replacement prevents the read
* I/O from overwriting potentially dirty VM-backed
* pages. XXX bogus page replacement is, uh, bogus.
* It may not work properly with small-block devices.
* We need to find a better way.
*/
if (clear_modify) {
pmap_remove_write(m);
vfs_page_set_validclean(bp, foff, m);
} else if (m->valid == VM_PAGE_BITS_ALL &&
(bp->b_flags & B_CACHE) == 0) {
bp->b_pages[i] = bogus_page;
bogus++;
}
foff = (foff + PAGE_SIZE) & ~(off_t)PAGE_MASK;
}
VM_OBJECT_UNLOCK(obj);
if (bogus)
pmap_qenter(trunc_page((vm_offset_t)bp->b_data),
bp->b_pages, bp->b_npages);
}
/*
* vfs_bio_set_valid:
*
* Set the range within the buffer to valid. The range is
* relative to the beginning of the buffer, b_offset. Note that
* b_offset itself may be offset from the beginning of the first
* page.
*/
void
vfs_bio_set_valid(struct buf *bp, int base, int size)
{
int i, n;
vm_page_t m;
if (!(bp->b_flags & B_VMIO))
return;
/*
* Fixup base to be relative to beginning of first page.
* Set initial n to be the maximum number of bytes in the
* first page that can be validated.
*/
base += (bp->b_offset & PAGE_MASK);
n = PAGE_SIZE - (base & PAGE_MASK);
VM_OBJECT_LOCK(bp->b_bufobj->bo_object);
for (i = base / PAGE_SIZE; size > 0 && i < bp->b_npages; ++i) {
m = bp->b_pages[i];
if (n > size)
n = size;
vm_page_set_valid(m, base & PAGE_MASK, n);
base += n;
size -= n;
n = PAGE_SIZE;
}
VM_OBJECT_UNLOCK(bp->b_bufobj->bo_object);
}
/*
* vfs_bio_clrbuf:
*
* If the specified buffer is a non-VMIO buffer, clear the entire
* buffer. If the specified buffer is a VMIO buffer, clear and
* validate only the previously invalid portions of the buffer.
* This routine essentially fakes an I/O, so we need to clear
* BIO_ERROR and B_INVAL.
*
* Note that while we only theoretically need to clear through b_bcount,
* we go ahead and clear through b_bufsize.
*/
void
vfs_bio_clrbuf(struct buf *bp)
{
int i, j, mask;
caddr_t sa, ea;
if ((bp->b_flags & (B_VMIO | B_MALLOC)) != B_VMIO) {
clrbuf(bp);
return;
}
bp->b_flags &= ~B_INVAL;
bp->b_ioflags &= ~BIO_ERROR;
VM_OBJECT_LOCK(bp->b_bufobj->bo_object);
if ((bp->b_npages == 1) && (bp->b_bufsize < PAGE_SIZE) &&
(bp->b_offset & PAGE_MASK) == 0) {
if (bp->b_pages[0] == bogus_page)
goto unlock;
mask = (1 << (bp->b_bufsize / DEV_BSIZE)) - 1;
VM_OBJECT_LOCK_ASSERT(bp->b_pages[0]->object, MA_OWNED);
if ((bp->b_pages[0]->valid & mask) == mask)
goto unlock;
if ((bp->b_pages[0]->valid & mask) == 0) {
bzero(bp->b_data, bp->b_bufsize);
bp->b_pages[0]->valid |= mask;
goto unlock;
}
}
ea = sa = bp->b_data;
for(i = 0; i < bp->b_npages; i++, sa = ea) {
ea = (caddr_t)trunc_page((vm_offset_t)sa + PAGE_SIZE);
ea = (caddr_t)(vm_offset_t)ulmin(
(u_long)(vm_offset_t)ea,
(u_long)(vm_offset_t)bp->b_data + bp->b_bufsize);
if (bp->b_pages[i] == bogus_page)
continue;
j = ((vm_offset_t)sa & PAGE_MASK) / DEV_BSIZE;
mask = ((1 << ((ea - sa) / DEV_BSIZE)) - 1) << j;
VM_OBJECT_LOCK_ASSERT(bp->b_pages[i]->object, MA_OWNED);
if ((bp->b_pages[i]->valid & mask) == mask)
continue;
if ((bp->b_pages[i]->valid & mask) == 0)
bzero(sa, ea - sa);
else {
for (; sa < ea; sa += DEV_BSIZE, j++) {
if ((bp->b_pages[i]->valid & (1 << j)) == 0)
bzero(sa, DEV_BSIZE);
}
}
bp->b_pages[i]->valid |= mask;
}
unlock:
VM_OBJECT_UNLOCK(bp->b_bufobj->bo_object);
bp->b_resid = 0;
}
/*
* vm_hold_load_pages and vm_hold_free_pages get pages into
* a buffers address space. The pages are anonymous and are
* not associated with a file object.
*/
static void
vm_hold_load_pages(struct buf *bp, vm_offset_t from, vm_offset_t to)
{
vm_offset_t pg;
vm_page_t p;
int index;
to = round_page(to);
from = round_page(from);
index = (from - trunc_page((vm_offset_t)bp->b_data)) >> PAGE_SHIFT;
for (pg = from; pg < to; pg += PAGE_SIZE, index++) {
tryagain:
/*
* note: must allocate system pages since blocking here
* could interfere with paging I/O, no matter which
* process we are.
*/
p = vm_page_alloc(NULL, pg >> PAGE_SHIFT, VM_ALLOC_NOOBJ |
VM_ALLOC_SYSTEM | VM_ALLOC_WIRED |
VM_ALLOC_COUNT((to - pg) >> PAGE_SHIFT));
if (!p) {
VM_WAIT;
goto tryagain;
}
pmap_qenter(pg, &p, 1);
bp->b_pages[index] = p;
}
bp->b_npages = index;
}
/* Return pages associated with this buf to the vm system */
static void
vm_hold_free_pages(struct buf *bp, int newbsize)
{
vm_offset_t from;
vm_page_t p;
int index, newnpages;
from = round_page((vm_offset_t)bp->b_data + newbsize);
newnpages = (from - trunc_page((vm_offset_t)bp->b_data)) >> PAGE_SHIFT;
if (bp->b_npages > newnpages)
pmap_qremove(from, bp->b_npages - newnpages);
for (index = newnpages; index < bp->b_npages; index++) {
p = bp->b_pages[index];
bp->b_pages[index] = NULL;
if (p->busy != 0)
printf("vm_hold_free_pages: blkno: %jd, lblkno: %jd\n",
(intmax_t)bp->b_blkno, (intmax_t)bp->b_lblkno);
p->wire_count--;
vm_page_free(p);
atomic_subtract_int(&cnt.v_wire_count, 1);
}
bp->b_npages = newnpages;
}
/*
* Map an IO request into kernel virtual address space.
*
* All requests are (re)mapped into kernel VA space.
* Notice that we use b_bufsize for the size of the buffer
* to be mapped. b_bcount might be modified by the driver.
*
* Note that even if the caller determines that the address space should
* be valid, a race or a smaller-file mapped into a larger space may
* actually cause vmapbuf() to fail, so all callers of vmapbuf() MUST
* check the return value.
*/
int
vmapbuf(struct buf *bp)
{
caddr_t kva;
vm_prot_t prot;
int pidx;
if (bp->b_bufsize < 0)
return (-1);
prot = VM_PROT_READ;
if (bp->b_iocmd == BIO_READ)
prot |= VM_PROT_WRITE; /* Less backwards than it looks */
if ((pidx = vm_fault_quick_hold_pages(&curproc->p_vmspace->vm_map,
(vm_offset_t)bp->b_data, bp->b_bufsize, prot, bp->b_pages,
btoc(MAXPHYS))) < 0)
return (-1);
pmap_qenter((vm_offset_t)bp->b_saveaddr, bp->b_pages, pidx);
kva = bp->b_saveaddr;
bp->b_npages = pidx;
bp->b_saveaddr = bp->b_data;
bp->b_data = kva + (((vm_offset_t) bp->b_data) & PAGE_MASK);
return(0);
}
/*
* Free the io map PTEs associated with this IO operation.
* We also invalidate the TLB entries and restore the original b_addr.
*/
void
vunmapbuf(struct buf *bp)
{
int npages;
npages = bp->b_npages;
pmap_qremove(trunc_page((vm_offset_t)bp->b_data), npages);
vm_page_unhold_pages(bp->b_pages, npages);
bp->b_data = bp->b_saveaddr;
}
void
bdone(struct buf *bp)
{
struct mtx *mtxp;
mtxp = mtx_pool_find(mtxpool_sleep, bp);
mtx_lock(mtxp);
bp->b_flags |= B_DONE;
wakeup(bp);
mtx_unlock(mtxp);
}
void
bwait(struct buf *bp, u_char pri, const char *wchan)
{
struct mtx *mtxp;
mtxp = mtx_pool_find(mtxpool_sleep, bp);
mtx_lock(mtxp);
while ((bp->b_flags & B_DONE) == 0)
msleep(bp, mtxp, pri, wchan, 0);
mtx_unlock(mtxp);
}
int
bufsync(struct bufobj *bo, int waitfor)
{
return (VOP_FSYNC(bo->__bo_vnode, waitfor, curthread));
}
void
bufstrategy(struct bufobj *bo, struct buf *bp)
{
int i = 0;
struct vnode *vp;
vp = bp->b_vp;
KASSERT(vp == bo->bo_private, ("Inconsistent vnode bufstrategy"));
KASSERT(vp->v_type != VCHR && vp->v_type != VBLK,
("Wrong vnode in bufstrategy(bp=%p, vp=%p)", bp, vp));
i = VOP_STRATEGY(vp, bp);
KASSERT(i == 0, ("VOP_STRATEGY failed bp=%p vp=%p", bp, bp->b_vp));
}
void
bufobj_wrefl(struct bufobj *bo)
{
KASSERT(bo != NULL, ("NULL bo in bufobj_wref"));
ASSERT_BO_LOCKED(bo);
bo->bo_numoutput++;
}
void
bufobj_wref(struct bufobj *bo)
{
KASSERT(bo != NULL, ("NULL bo in bufobj_wref"));
BO_LOCK(bo);
bo->bo_numoutput++;
BO_UNLOCK(bo);
}
void
bufobj_wdrop(struct bufobj *bo)
{
KASSERT(bo != NULL, ("NULL bo in bufobj_wdrop"));
BO_LOCK(bo);
KASSERT(bo->bo_numoutput > 0, ("bufobj_wdrop non-positive count"));
if ((--bo->bo_numoutput == 0) && (bo->bo_flag & BO_WWAIT)) {
bo->bo_flag &= ~BO_WWAIT;
wakeup(&bo->bo_numoutput);
}
BO_UNLOCK(bo);
}
int
bufobj_wwait(struct bufobj *bo, int slpflag, int timeo)
{
int error;
KASSERT(bo != NULL, ("NULL bo in bufobj_wwait"));
ASSERT_BO_LOCKED(bo);
error = 0;
while (bo->bo_numoutput) {
bo->bo_flag |= BO_WWAIT;
error = msleep(&bo->bo_numoutput, BO_MTX(bo),
slpflag | (PRIBIO + 1), "bo_wwait", timeo);
if (error)
break;
}
return (error);
}
void
bpin(struct buf *bp)
{
struct mtx *mtxp;
mtxp = mtx_pool_find(mtxpool_sleep, bp);
mtx_lock(mtxp);
bp->b_pin_count++;
mtx_unlock(mtxp);
}
void
bunpin(struct buf *bp)
{
struct mtx *mtxp;
mtxp = mtx_pool_find(mtxpool_sleep, bp);
mtx_lock(mtxp);
if (--bp->b_pin_count == 0)
wakeup(bp);
mtx_unlock(mtxp);
}
void
bunpin_wait(struct buf *bp)
{
struct mtx *mtxp;
mtxp = mtx_pool_find(mtxpool_sleep, bp);
mtx_lock(mtxp);
while (bp->b_pin_count > 0)
msleep(bp, mtxp, PRIBIO, "bwunpin", 0);
mtx_unlock(mtxp);
}
#include "opt_ddb.h"
#ifdef DDB
#include <ddb/ddb.h>
/* DDB command to show buffer data */
DB_SHOW_COMMAND(buffer, db_show_buffer)
{
/* get args */
struct buf *bp = (struct buf *)addr;
if (!have_addr) {
db_printf("usage: show buffer <addr>\n");
return;
}
db_printf("buf at %p\n", bp);
db_printf("b_flags = 0x%b\n", (u_int)bp->b_flags, PRINT_BUF_FLAGS);
db_printf(
"b_error = %d, b_bufsize = %ld, b_bcount = %ld, b_resid = %ld\n"
"b_bufobj = (%p), b_data = %p, b_blkno = %jd, b_dep = %p\n",
bp->b_error, bp->b_bufsize, bp->b_bcount, bp->b_resid,
bp->b_bufobj, bp->b_data, (intmax_t)bp->b_blkno,
bp->b_dep.lh_first);
if (bp->b_npages) {
int i;
db_printf("b_npages = %d, pages(OBJ, IDX, PA): ", bp->b_npages);
for (i = 0; i < bp->b_npages; i++) {
vm_page_t m;
m = bp->b_pages[i];
db_printf("(%p, 0x%lx, 0x%lx)", (void *)m->object,
(u_long)m->pindex, (u_long)VM_PAGE_TO_PHYS(m));
if ((i + 1) < bp->b_npages)
db_printf(",");
}
db_printf("\n");
}
db_printf(" ");
lockmgr_printinfo(&bp->b_lock);
}
DB_SHOW_COMMAND(lockedbufs, lockedbufs)
{
struct buf *bp;
int i;
for (i = 0; i < nbuf; i++) {
bp = &buf[i];
if (BUF_ISLOCKED(bp)) {
db_show_buffer((uintptr_t)bp, 1, 0, NULL);
db_printf("\n");
}
}
}
DB_SHOW_COMMAND(vnodebufs, db_show_vnodebufs)
{
struct vnode *vp;
struct buf *bp;
if (!have_addr) {
db_printf("usage: show vnodebufs <addr>\n");
return;
}
vp = (struct vnode *)addr;
db_printf("Clean buffers:\n");
TAILQ_FOREACH(bp, &vp->v_bufobj.bo_clean.bv_hd, b_bobufs) {
db_show_buffer((uintptr_t)bp, 1, 0, NULL);
db_printf("\n");
}
db_printf("Dirty buffers:\n");
TAILQ_FOREACH(bp, &vp->v_bufobj.bo_dirty.bv_hd, b_bobufs) {
db_show_buffer((uintptr_t)bp, 1, 0, NULL);
db_printf("\n");
}
}
DB_COMMAND(countfreebufs, db_coundfreebufs)
{
struct buf *bp;
int i, used = 0, nfree = 0;
if (have_addr) {
db_printf("usage: countfreebufs\n");
return;
}
for (i = 0; i < nbuf; i++) {
bp = &buf[i];
if ((bp->b_vflags & BV_INFREECNT) != 0)
nfree++;
else
used++;
}
db_printf("Counted %d free, %d used (%d tot)\n", nfree, used,
nfree + used);
db_printf("numfreebuffers is %d\n", numfreebuffers);
}
#endif /* DDB */
|