summaryrefslogtreecommitdiffstats
path: root/module/zfs/metaslab.c
blob: a8a5f45db9f0fd0f392a2425cd7e1d5c3b54f555 (plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
1001
1002
1003
1004
1005
1006
1007
1008
1009
1010
1011
1012
1013
1014
1015
1016
1017
1018
1019
1020
1021
1022
1023
1024
1025
1026
1027
1028
1029
1030
1031
1032
1033
1034
1035
1036
1037
1038
1039
1040
1041
1042
1043
1044
1045
1046
1047
1048
1049
1050
1051
1052
1053
1054
1055
1056
1057
1058
1059
1060
1061
1062
1063
1064
1065
1066
1067
1068
1069
1070
1071
1072
1073
1074
1075
1076
1077
1078
1079
1080
1081
1082
1083
1084
1085
1086
1087
1088
1089
1090
1091
1092
1093
1094
1095
1096
1097
1098
1099
1100
1101
1102
1103
1104
1105
1106
1107
1108
1109
1110
1111
1112
1113
1114
1115
1116
1117
1118
1119
1120
1121
1122
1123
1124
1125
1126
1127
1128
1129
1130
1131
1132
1133
1134
1135
1136
1137
1138
1139
1140
1141
1142
1143
1144
1145
1146
1147
1148
1149
1150
1151
1152
1153
1154
1155
1156
1157
1158
1159
1160
1161
1162
1163
1164
1165
1166
1167
1168
1169
1170
1171
1172
1173
1174
1175
1176
1177
1178
1179
1180
1181
1182
1183
1184
1185
1186
1187
1188
1189
1190
1191
1192
1193
1194
1195
1196
1197
1198
1199
1200
1201
1202
1203
1204
1205
1206
1207
1208
1209
1210
1211
1212
1213
1214
1215
1216
1217
1218
1219
1220
1221
1222
1223
1224
1225
1226
1227
1228
1229
1230
1231
1232
1233
1234
1235
1236
1237
1238
1239
1240
1241
1242
1243
1244
1245
1246
1247
1248
1249
1250
1251
1252
1253
1254
1255
1256
1257
1258
1259
1260
1261
1262
1263
1264
1265
1266
1267
1268
1269
1270
1271
1272
1273
1274
1275
1276
1277
1278
1279
1280
1281
1282
1283
1284
1285
1286
1287
1288
1289
1290
1291
1292
1293
1294
1295
1296
1297
1298
1299
1300
1301
1302
1303
1304
1305
1306
1307
1308
1309
1310
1311
1312
1313
1314
1315
1316
1317
1318
1319
1320
1321
1322
1323
1324
1325
1326
1327
1328
1329
1330
1331
1332
1333
1334
1335
1336
1337
1338
1339
1340
1341
1342
1343
1344
1345
1346
1347
1348
1349
1350
1351
1352
1353
1354
1355
1356
1357
1358
1359
1360
1361
1362
1363
1364
1365
1366
1367
1368
1369
1370
1371
1372
1373
1374
1375
1376
1377
1378
1379
1380
1381
1382
1383
1384
1385
1386
1387
1388
1389
1390
1391
1392
1393
1394
1395
1396
1397
1398
1399
1400
1401
1402
1403
1404
1405
1406
1407
1408
1409
1410
1411
1412
1413
1414
1415
1416
1417
1418
1419
1420
1421
1422
1423
1424
1425
1426
1427
1428
1429
1430
1431
1432
1433
1434
1435
1436
1437
1438
1439
1440
1441
1442
1443
1444
1445
1446
1447
1448
1449
1450
1451
1452
1453
1454
1455
1456
1457
1458
1459
1460
1461
1462
1463
1464
1465
1466
1467
1468
1469
1470
1471
1472
1473
1474
1475
1476
1477
1478
1479
1480
1481
1482
1483
1484
1485
1486
1487
1488
1489
1490
1491
1492
1493
1494
1495
1496
1497
1498
1499
1500
1501
1502
1503
1504
1505
1506
1507
1508
1509
1510
1511
1512
1513
1514
1515
1516
1517
1518
1519
1520
1521
1522
1523
1524
1525
1526
1527
1528
1529
1530
1531
1532
1533
1534
1535
1536
1537
1538
1539
1540
1541
1542
1543
1544
1545
1546
1547
1548
1549
1550
1551
1552
1553
1554
1555
1556
1557
1558
1559
1560
1561
1562
1563
1564
1565
1566
1567
1568
1569
1570
1571
1572
1573
1574
1575
1576
1577
1578
1579
1580
1581
1582
1583
1584
1585
1586
1587
1588
1589
1590
1591
1592
1593
1594
1595
1596
1597
1598
1599
1600
1601
1602
1603
1604
1605
1606
1607
1608
1609
1610
1611
1612
1613
1614
1615
1616
1617
1618
1619
1620
1621
1622
1623
1624
1625
1626
1627
1628
1629
1630
1631
1632
1633
1634
1635
1636
1637
1638
1639
1640
1641
1642
1643
1644
1645
1646
1647
1648
1649
1650
1651
1652
1653
1654
1655
1656
1657
1658
1659
1660
1661
1662
1663
1664
1665
1666
1667
1668
1669
1670
1671
1672
1673
1674
1675
1676
1677
1678
1679
1680
1681
1682
1683
1684
1685
1686
1687
1688
1689
1690
1691
1692
1693
1694
1695
1696
1697
1698
1699
1700
1701
1702
1703
1704
1705
1706
1707
1708
1709
1710
1711
1712
1713
1714
1715
1716
1717
1718
1719
1720
1721
1722
1723
1724
1725
1726
1727
1728
1729
1730
1731
1732
1733
1734
1735
1736
1737
1738
1739
1740
1741
1742
1743
1744
1745
1746
1747
1748
1749
1750
1751
1752
1753
1754
1755
1756
1757
1758
1759
1760
1761
1762
1763
1764
1765
1766
1767
1768
1769
1770
1771
1772
1773
1774
1775
1776
1777
1778
1779
1780
1781
1782
1783
1784
1785
1786
1787
1788
1789
1790
1791
1792
1793
1794
1795
1796
1797
1798
1799
1800
1801
1802
1803
1804
1805
1806
1807
1808
1809
1810
1811
1812
1813
1814
1815
1816
1817
1818
1819
1820
1821
1822
1823
1824
1825
1826
1827
1828
1829
1830
1831
1832
1833
1834
1835
1836
1837
1838
1839
1840
1841
1842
1843
1844
1845
1846
1847
1848
1849
1850
1851
1852
1853
1854
1855
1856
1857
1858
1859
1860
1861
1862
1863
1864
1865
1866
1867
1868
1869
1870
1871
1872
1873
1874
1875
1876
1877
1878
1879
1880
1881
1882
1883
1884
1885
1886
1887
1888
1889
1890
1891
1892
1893
1894
1895
1896
1897
1898
1899
1900
1901
1902
1903
1904
1905
1906
1907
1908
1909
1910
1911
1912
1913
1914
1915
1916
1917
1918
1919
1920
1921
1922
1923
1924
1925
1926
1927
1928
1929
1930
1931
1932
1933
1934
1935
1936
1937
1938
1939
1940
1941
1942
1943
1944
1945
1946
1947
1948
1949
1950
1951
1952
1953
1954
1955
1956
1957
1958
1959
1960
1961
1962
1963
1964
1965
1966
1967
1968
1969
1970
1971
1972
1973
1974
1975
1976
1977
1978
1979
1980
1981
1982
1983
1984
1985
1986
1987
1988
1989
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
2011
2012
2013
2014
2015
2016
2017
2018
2019
2020
2021
2022
2023
2024
2025
2026
2027
2028
2029
2030
2031
2032
2033
2034
2035
2036
2037
2038
2039
2040
2041
2042
2043
2044
2045
2046
2047
2048
2049
2050
2051
2052
2053
2054
2055
2056
2057
2058
2059
2060
2061
2062
2063
2064
2065
2066
2067
2068
2069
2070
2071
2072
2073
2074
2075
2076
2077
2078
2079
2080
2081
2082
2083
2084
2085
2086
2087
2088
2089
2090
2091
2092
2093
2094
2095
2096
2097
2098
2099
2100
2101
2102
2103
2104
2105
2106
2107
2108
2109
2110
2111
2112
2113
2114
2115
2116
2117
2118
2119
2120
2121
2122
2123
2124
2125
2126
2127
2128
2129
2130
2131
2132
2133
2134
2135
2136
2137
2138
2139
2140
2141
2142
2143
2144
2145
2146
2147
2148
2149
2150
2151
2152
2153
2154
2155
2156
2157
2158
2159
2160
2161
2162
2163
2164
2165
2166
2167
2168
2169
2170
2171
2172
2173
2174
2175
2176
2177
2178
2179
2180
2181
2182
2183
2184
2185
2186
2187
2188
2189
2190
2191
2192
2193
2194
2195
2196
2197
2198
2199
2200
2201
2202
2203
2204
2205
2206
2207
2208
2209
2210
2211
2212
2213
2214
2215
2216
2217
2218
2219
2220
2221
2222
2223
2224
2225
2226
2227
2228
2229
2230
2231
2232
2233
2234
2235
2236
2237
2238
2239
2240
2241
2242
2243
2244
2245
2246
2247
2248
2249
2250
2251
2252
2253
2254
2255
2256
2257
2258
2259
2260
2261
2262
2263
2264
2265
2266
2267
2268
2269
2270
2271
2272
2273
2274
2275
2276
2277
2278
2279
2280
2281
2282
2283
2284
2285
2286
2287
2288
2289
2290
2291
2292
2293
2294
2295
2296
2297
2298
2299
2300
2301
2302
2303
2304
2305
2306
2307
2308
2309
2310
2311
2312
2313
2314
2315
2316
2317
2318
2319
2320
2321
2322
2323
2324
2325
2326
2327
2328
2329
2330
2331
2332
2333
2334
2335
2336
2337
2338
2339
2340
2341
2342
2343
2344
2345
2346
2347
2348
2349
2350
2351
2352
2353
2354
2355
2356
2357
2358
2359
2360
2361
2362
2363
2364
2365
2366
2367
2368
2369
2370
2371
2372
2373
2374
2375
2376
2377
2378
2379
2380
2381
2382
2383
2384
2385
2386
2387
2388
2389
2390
2391
2392
2393
2394
2395
2396
2397
2398
2399
2400
2401
2402
2403
2404
2405
2406
2407
2408
2409
2410
2411
2412
2413
2414
2415
2416
2417
2418
2419
2420
2421
2422
2423
2424
2425
2426
2427
2428
2429
2430
2431
2432
2433
2434
2435
2436
2437
2438
2439
2440
2441
2442
2443
2444
2445
2446
2447
2448
2449
2450
2451
2452
2453
2454
2455
2456
2457
2458
2459
2460
2461
2462
2463
2464
2465
2466
2467
2468
2469
2470
2471
2472
2473
2474
2475
2476
2477
2478
2479
2480
2481
2482
2483
2484
2485
2486
2487
2488
2489
2490
2491
2492
2493
2494
2495
2496
2497
2498
2499
2500
2501
2502
2503
2504
2505
2506
2507
2508
2509
2510
2511
2512
2513
2514
2515
2516
2517
2518
2519
2520
2521
2522
2523
2524
2525
2526
2527
2528
2529
2530
2531
2532
2533
2534
2535
2536
2537
2538
2539
2540
2541
2542
2543
2544
2545
2546
2547
2548
2549
2550
2551
2552
2553
2554
2555
2556
2557
2558
2559
2560
2561
2562
2563
2564
2565
2566
2567
2568
2569
2570
2571
2572
2573
2574
2575
2576
2577
2578
2579
2580
2581
2582
2583
2584
2585
2586
2587
2588
2589
2590
2591
2592
2593
2594
2595
2596
2597
2598
2599
2600
2601
2602
2603
2604
2605
2606
2607
2608
2609
2610
2611
2612
2613
2614
2615
2616
2617
2618
2619
2620
2621
2622
2623
2624
2625
2626
2627
2628
2629
2630
2631
2632
2633
2634
2635
2636
2637
2638
2639
2640
2641
2642
2643
2644
2645
2646
2647
2648
2649
2650
2651
2652
2653
2654
2655
2656
2657
2658
2659
2660
2661
2662
2663
2664
2665
2666
2667
2668
2669
2670
2671
2672
2673
2674
2675
2676
2677
2678
2679
2680
2681
2682
2683
2684
2685
2686
2687
2688
2689
2690
2691
2692
2693
2694
2695
2696
2697
2698
2699
2700
2701
2702
2703
2704
2705
2706
2707
2708
2709
2710
2711
2712
2713
2714
2715
2716
2717
2718
2719
2720
2721
2722
2723
2724
2725
2726
2727
2728
2729
2730
2731
2732
2733
2734
2735
2736
2737
2738
2739
2740
2741
2742
2743
2744
2745
2746
2747
2748
2749
2750
2751
2752
2753
2754
2755
2756
2757
2758
2759
2760
2761
2762
2763
2764
2765
2766
2767
2768
2769
2770
2771
2772
2773
2774
2775
2776
2777
2778
2779
2780
2781
2782
2783
2784
2785
2786
2787
2788
2789
2790
2791
2792
2793
2794
2795
2796
2797
2798
2799
2800
2801
2802
2803
2804
2805
2806
2807
2808
2809
2810
2811
2812
2813
2814
2815
2816
2817
2818
2819
2820
2821
2822
2823
2824
2825
2826
2827
2828
2829
2830
2831
2832
2833
2834
2835
2836
2837
2838
2839
2840
2841
2842
2843
2844
2845
2846
2847
2848
2849
2850
2851
2852
2853
2854
2855
2856
2857
2858
2859
2860
2861
2862
2863
2864
2865
2866
2867
2868
2869
2870
2871
2872
2873
2874
2875
2876
2877
2878
2879
2880
2881
2882
2883
2884
2885
2886
2887
2888
2889
2890
2891
2892
2893
2894
2895
2896
2897
2898
2899
2900
2901
2902
2903
2904
2905
2906
2907
2908
2909
2910
2911
2912
2913
2914
2915
2916
2917
2918
2919
2920
2921
2922
2923
2924
2925
2926
2927
2928
2929
2930
2931
2932
2933
2934
2935
2936
2937
2938
2939
2940
2941
2942
2943
2944
2945
2946
2947
2948
2949
2950
2951
2952
2953
2954
2955
2956
2957
2958
2959
2960
2961
2962
2963
2964
2965
2966
2967
/*
 * CDDL HEADER START
 *
 * The contents of this file are subject to the terms of the
 * Common Development and Distribution License (the "License").
 * You may not use this file except in compliance with the License.
 *
 * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
 * or http://www.opensolaris.org/os/licensing.
 * See the License for the specific language governing permissions
 * and limitations under the License.
 *
 * When distributing Covered Code, include this CDDL HEADER in each
 * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
 * If applicable, add the following below this CDDL HEADER, with the
 * fields enclosed by brackets "[]" replaced with your own identifying
 * information: Portions Copyright [yyyy] [name of copyright owner]
 *
 * CDDL HEADER END
 */
/*
 * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
 * Copyright (c) 2011, 2015 by Delphix. All rights reserved.
 * Copyright (c) 2013 by Saso Kiselkov. All rights reserved.
 */

#include <sys/zfs_context.h>
#include <sys/dmu.h>
#include <sys/dmu_tx.h>
#include <sys/space_map.h>
#include <sys/metaslab_impl.h>
#include <sys/vdev_impl.h>
#include <sys/zio.h>
#include <sys/spa_impl.h>
#include <sys/zfeature.h>

#define	WITH_DF_BLOCK_ALLOCATOR

#define	GANG_ALLOCATION(flags) \
	((flags) & (METASLAB_GANG_CHILD | METASLAB_GANG_HEADER))

#define	METASLAB_WEIGHT_PRIMARY		(1ULL << 63)
#define	METASLAB_WEIGHT_SECONDARY	(1ULL << 62)
#define	METASLAB_ACTIVE_MASK		\
	(METASLAB_WEIGHT_PRIMARY | METASLAB_WEIGHT_SECONDARY)

/*
 * Metaslab granularity, in bytes. This is roughly similar to what would be
 * referred to as the "stripe size" in traditional RAID arrays. In normal
 * operation, we will try to write this amount of data to a top-level vdev
 * before moving on to the next one.
 */
unsigned long metaslab_aliquot = 512 << 10;

uint64_t metaslab_gang_bang = SPA_MAXBLOCKSIZE + 1;	/* force gang blocks */

/*
 * The in-core space map representation is more compact than its on-disk form.
 * The zfs_condense_pct determines how much more compact the in-core
 * space_map representation must be before we compact it on-disk.
 * Values should be greater than or equal to 100.
 */
int zfs_condense_pct = 200;

/*
 * Condensing a metaslab is not guaranteed to actually reduce the amount of
 * space used on disk. In particular, a space map uses data in increments of
 * MAX(1 << ashift, space_map_blksz), so a metaslab might use the
 * same number of blocks after condensing. Since the goal of condensing is to
 * reduce the number of IOPs required to read the space map, we only want to
 * condense when we can be sure we will reduce the number of blocks used by the
 * space map. Unfortunately, we cannot precisely compute whether or not this is
 * the case in metaslab_should_condense since we are holding ms_lock. Instead,
 * we apply the following heuristic: do not condense a spacemap unless the
 * uncondensed size consumes greater than zfs_metaslab_condense_block_threshold
 * blocks.
 */
int zfs_metaslab_condense_block_threshold = 4;

/*
 * The zfs_mg_noalloc_threshold defines which metaslab groups should
 * be eligible for allocation. The value is defined as a percentage of
 * free space. Metaslab groups that have more free space than
 * zfs_mg_noalloc_threshold are always eligible for allocations. Once
 * a metaslab group's free space is less than or equal to the
 * zfs_mg_noalloc_threshold the allocator will avoid allocating to that
 * group unless all groups in the pool have reached zfs_mg_noalloc_threshold.
 * Once all groups in the pool reach zfs_mg_noalloc_threshold then all
 * groups are allowed to accept allocations. Gang blocks are always
 * eligible to allocate on any metaslab group. The default value of 0 means
 * no metaslab group will be excluded based on this criterion.
 */
int zfs_mg_noalloc_threshold = 0;

/*
 * Metaslab groups are considered eligible for allocations if their
 * fragmenation metric (measured as a percentage) is less than or equal to
 * zfs_mg_fragmentation_threshold. If a metaslab group exceeds this threshold
 * then it will be skipped unless all metaslab groups within the metaslab
 * class have also crossed this threshold.
 */
int zfs_mg_fragmentation_threshold = 85;

/*
 * Allow metaslabs to keep their active state as long as their fragmentation
 * percentage is less than or equal to zfs_metaslab_fragmentation_threshold. An
 * active metaslab that exceeds this threshold will no longer keep its active
 * status allowing better metaslabs to be selected.
 */
int zfs_metaslab_fragmentation_threshold = 70;

/*
 * When set will load all metaslabs when pool is first opened.
 */
int metaslab_debug_load = 0;

/*
 * When set will prevent metaslabs from being unloaded.
 */
int metaslab_debug_unload = 0;

/*
 * Minimum size which forces the dynamic allocator to change
 * it's allocation strategy.  Once the space map cannot satisfy
 * an allocation of this size then it switches to using more
 * aggressive strategy (i.e search by size rather than offset).
 */
uint64_t metaslab_df_alloc_threshold = SPA_MAXBLOCKSIZE;

/*
 * The minimum free space, in percent, which must be available
 * in a space map to continue allocations in a first-fit fashion.
 * Once the space_map's free space drops below this level we dynamically
 * switch to using best-fit allocations.
 */
int metaslab_df_free_pct = 4;

/*
 * Percentage of all cpus that can be used by the metaslab taskq.
 */
int metaslab_load_pct = 50;

/*
 * Determines how many txgs a metaslab may remain loaded without having any
 * allocations from it. As long as a metaslab continues to be used we will
 * keep it loaded.
 */
int metaslab_unload_delay = TXG_SIZE * 2;

/*
 * Max number of metaslabs per group to preload.
 */
int metaslab_preload_limit = SPA_DVAS_PER_BP;

/*
 * Enable/disable preloading of metaslab.
 */
int metaslab_preload_enabled = B_TRUE;

/*
 * Enable/disable fragmentation weighting on metaslabs.
 */
int metaslab_fragmentation_factor_enabled = B_TRUE;

/*
 * Enable/disable lba weighting (i.e. outer tracks are given preference).
 */
int metaslab_lba_weighting_enabled = B_TRUE;

/*
 * Enable/disable metaslab group biasing.
 */
int metaslab_bias_enabled = B_TRUE;

static uint64_t metaslab_fragmentation(metaslab_t *);

/*
 * ==========================================================================
 * Metaslab classes
 * ==========================================================================
 */
metaslab_class_t *
metaslab_class_create(spa_t *spa, metaslab_ops_t *ops)
{
	metaslab_class_t *mc;

	mc = kmem_zalloc(sizeof (metaslab_class_t), KM_SLEEP);

	mc->mc_spa = spa;
	mc->mc_rotor = NULL;
	mc->mc_ops = ops;
	mutex_init(&mc->mc_lock, NULL, MUTEX_DEFAULT, NULL);
	refcount_create_tracked(&mc->mc_alloc_slots);

	return (mc);
}

void
metaslab_class_destroy(metaslab_class_t *mc)
{
	ASSERT(mc->mc_rotor == NULL);
	ASSERT(mc->mc_alloc == 0);
	ASSERT(mc->mc_deferred == 0);
	ASSERT(mc->mc_space == 0);
	ASSERT(mc->mc_dspace == 0);

	refcount_destroy(&mc->mc_alloc_slots);
	mutex_destroy(&mc->mc_lock);
	kmem_free(mc, sizeof (metaslab_class_t));
}

int
metaslab_class_validate(metaslab_class_t *mc)
{
	metaslab_group_t *mg;
	vdev_t *vd;

	/*
	 * Must hold one of the spa_config locks.
	 */
	ASSERT(spa_config_held(mc->mc_spa, SCL_ALL, RW_READER) ||
	    spa_config_held(mc->mc_spa, SCL_ALL, RW_WRITER));

	if ((mg = mc->mc_rotor) == NULL)
		return (0);

	do {
		vd = mg->mg_vd;
		ASSERT(vd->vdev_mg != NULL);
		ASSERT3P(vd->vdev_top, ==, vd);
		ASSERT3P(mg->mg_class, ==, mc);
		ASSERT3P(vd->vdev_ops, !=, &vdev_hole_ops);
	} while ((mg = mg->mg_next) != mc->mc_rotor);

	return (0);
}

void
metaslab_class_space_update(metaslab_class_t *mc, int64_t alloc_delta,
    int64_t defer_delta, int64_t space_delta, int64_t dspace_delta)
{
	atomic_add_64(&mc->mc_alloc, alloc_delta);
	atomic_add_64(&mc->mc_deferred, defer_delta);
	atomic_add_64(&mc->mc_space, space_delta);
	atomic_add_64(&mc->mc_dspace, dspace_delta);
}

uint64_t
metaslab_class_get_alloc(metaslab_class_t *mc)
{
	return (mc->mc_alloc);
}

uint64_t
metaslab_class_get_deferred(metaslab_class_t *mc)
{
	return (mc->mc_deferred);
}

uint64_t
metaslab_class_get_space(metaslab_class_t *mc)
{
	return (mc->mc_space);
}

uint64_t
metaslab_class_get_dspace(metaslab_class_t *mc)
{
	return (spa_deflate(mc->mc_spa) ? mc->mc_dspace : mc->mc_space);
}

void
metaslab_class_histogram_verify(metaslab_class_t *mc)
{
	vdev_t *rvd = mc->mc_spa->spa_root_vdev;
	uint64_t *mc_hist;
	int i, c;

	if ((zfs_flags & ZFS_DEBUG_HISTOGRAM_VERIFY) == 0)
		return;

	mc_hist = kmem_zalloc(sizeof (uint64_t) * RANGE_TREE_HISTOGRAM_SIZE,
	    KM_SLEEP);

	for (c = 0; c < rvd->vdev_children; c++) {
		vdev_t *tvd = rvd->vdev_child[c];
		metaslab_group_t *mg = tvd->vdev_mg;

		/*
		 * Skip any holes, uninitialized top-levels, or
		 * vdevs that are not in this metalab class.
		 */
		if (tvd->vdev_ishole || tvd->vdev_ms_shift == 0 ||
		    mg->mg_class != mc) {
			continue;
		}

		for (i = 0; i < RANGE_TREE_HISTOGRAM_SIZE; i++)
			mc_hist[i] += mg->mg_histogram[i];
	}

	for (i = 0; i < RANGE_TREE_HISTOGRAM_SIZE; i++)
		VERIFY3U(mc_hist[i], ==, mc->mc_histogram[i]);

	kmem_free(mc_hist, sizeof (uint64_t) * RANGE_TREE_HISTOGRAM_SIZE);
}

/*
 * Calculate the metaslab class's fragmentation metric. The metric
 * is weighted based on the space contribution of each metaslab group.
 * The return value will be a number between 0 and 100 (inclusive), or
 * ZFS_FRAG_INVALID if the metric has not been set. See comment above the
 * zfs_frag_table for more information about the metric.
 */
uint64_t
metaslab_class_fragmentation(metaslab_class_t *mc)
{
	vdev_t *rvd = mc->mc_spa->spa_root_vdev;
	uint64_t fragmentation = 0;
	int c;

	spa_config_enter(mc->mc_spa, SCL_VDEV, FTAG, RW_READER);

	for (c = 0; c < rvd->vdev_children; c++) {
		vdev_t *tvd = rvd->vdev_child[c];
		metaslab_group_t *mg = tvd->vdev_mg;

		/*
		 * Skip any holes, uninitialized top-levels, or
		 * vdevs that are not in this metalab class.
		 */
		if (tvd->vdev_ishole || tvd->vdev_ms_shift == 0 ||
		    mg->mg_class != mc) {
			continue;
		}

		/*
		 * If a metaslab group does not contain a fragmentation
		 * metric then just bail out.
		 */
		if (mg->mg_fragmentation == ZFS_FRAG_INVALID) {
			spa_config_exit(mc->mc_spa, SCL_VDEV, FTAG);
			return (ZFS_FRAG_INVALID);
		}

		/*
		 * Determine how much this metaslab_group is contributing
		 * to the overall pool fragmentation metric.
		 */
		fragmentation += mg->mg_fragmentation *
		    metaslab_group_get_space(mg);
	}
	fragmentation /= metaslab_class_get_space(mc);

	ASSERT3U(fragmentation, <=, 100);
	spa_config_exit(mc->mc_spa, SCL_VDEV, FTAG);
	return (fragmentation);
}

/*
 * Calculate the amount of expandable space that is available in
 * this metaslab class. If a device is expanded then its expandable
 * space will be the amount of allocatable space that is currently not
 * part of this metaslab class.
 */
uint64_t
metaslab_class_expandable_space(metaslab_class_t *mc)
{
	vdev_t *rvd = mc->mc_spa->spa_root_vdev;
	uint64_t space = 0;
	int c;

	spa_config_enter(mc->mc_spa, SCL_VDEV, FTAG, RW_READER);
	for (c = 0; c < rvd->vdev_children; c++) {
		vdev_t *tvd = rvd->vdev_child[c];
		metaslab_group_t *mg = tvd->vdev_mg;

		if (tvd->vdev_ishole || tvd->vdev_ms_shift == 0 ||
		    mg->mg_class != mc) {
			continue;
		}

		space += tvd->vdev_max_asize - tvd->vdev_asize;
	}
	spa_config_exit(mc->mc_spa, SCL_VDEV, FTAG);
	return (space);
}

/*
 * ==========================================================================
 * Metaslab groups
 * ==========================================================================
 */
static int
metaslab_compare(const void *x1, const void *x2)
{
	const metaslab_t *m1 = (const metaslab_t *)x1;
	const metaslab_t *m2 = (const metaslab_t *)x2;

	int cmp = AVL_CMP(m2->ms_weight, m1->ms_weight);
	if (likely(cmp))
		return (cmp);

	IMPLY(AVL_CMP(m1->ms_start, m2->ms_start) == 0, m1 == m2);

	return (AVL_CMP(m1->ms_start, m2->ms_start));
}

/*
 * Update the allocatable flag and the metaslab group's capacity.
 * The allocatable flag is set to true if the capacity is below
 * the zfs_mg_noalloc_threshold or has a fragmentation value that is
 * greater than zfs_mg_fragmentation_threshold. If a metaslab group
 * transitions from allocatable to non-allocatable or vice versa then the
 * metaslab group's class is updated to reflect the transition.
 */
static void
metaslab_group_alloc_update(metaslab_group_t *mg)
{
	vdev_t *vd = mg->mg_vd;
	metaslab_class_t *mc = mg->mg_class;
	vdev_stat_t *vs = &vd->vdev_stat;
	boolean_t was_allocatable;
	boolean_t was_initialized;

	ASSERT(vd == vd->vdev_top);

	mutex_enter(&mg->mg_lock);
	was_allocatable = mg->mg_allocatable;
	was_initialized = mg->mg_initialized;

	mg->mg_free_capacity = ((vs->vs_space - vs->vs_alloc) * 100) /
	    (vs->vs_space + 1);

	mutex_enter(&mc->mc_lock);

	/*
	 * If the metaslab group was just added then it won't
	 * have any space until we finish syncing out this txg.
	 * At that point we will consider it initialized and available
	 * for allocations.  We also don't consider non-activated
	 * metaslab groups (e.g. vdevs that are in the middle of being removed)
	 * to be initialized, because they can't be used for allocation.
	 */
	mg->mg_initialized = metaslab_group_initialized(mg);
	if (!was_initialized && mg->mg_initialized) {
		mc->mc_groups++;
	} else if (was_initialized && !mg->mg_initialized) {
		ASSERT3U(mc->mc_groups, >, 0);
		mc->mc_groups--;
	}
	if (mg->mg_initialized)
		mg->mg_no_free_space = B_FALSE;

	/*
	 * A metaslab group is considered allocatable if it has plenty
	 * of free space or is not heavily fragmented. We only take
	 * fragmentation into account if the metaslab group has a valid
	 * fragmentation metric (i.e. a value between 0 and 100).
	 */
	mg->mg_allocatable = (mg->mg_activation_count > 0 &&
	    mg->mg_free_capacity > zfs_mg_noalloc_threshold &&
	    (mg->mg_fragmentation == ZFS_FRAG_INVALID ||
	    mg->mg_fragmentation <= zfs_mg_fragmentation_threshold));

	/*
	 * The mc_alloc_groups maintains a count of the number of
	 * groups in this metaslab class that are still above the
	 * zfs_mg_noalloc_threshold. This is used by the allocating
	 * threads to determine if they should avoid allocations to
	 * a given group. The allocator will avoid allocations to a group
	 * if that group has reached or is below the zfs_mg_noalloc_threshold
	 * and there are still other groups that are above the threshold.
	 * When a group transitions from allocatable to non-allocatable or
	 * vice versa we update the metaslab class to reflect that change.
	 * When the mc_alloc_groups value drops to 0 that means that all
	 * groups have reached the zfs_mg_noalloc_threshold making all groups
	 * eligible for allocations. This effectively means that all devices
	 * are balanced again.
	 */
	if (was_allocatable && !mg->mg_allocatable)
		mc->mc_alloc_groups--;
	else if (!was_allocatable && mg->mg_allocatable)
		mc->mc_alloc_groups++;
	mutex_exit(&mc->mc_lock);

	mutex_exit(&mg->mg_lock);
}

metaslab_group_t *
metaslab_group_create(metaslab_class_t *mc, vdev_t *vd)
{
	metaslab_group_t *mg;

	mg = kmem_zalloc(sizeof (metaslab_group_t), KM_SLEEP);
	mutex_init(&mg->mg_lock, NULL, MUTEX_DEFAULT, NULL);
	avl_create(&mg->mg_metaslab_tree, metaslab_compare,
	    sizeof (metaslab_t), offsetof(struct metaslab, ms_group_node));
	mg->mg_vd = vd;
	mg->mg_class = mc;
	mg->mg_activation_count = 0;
	mg->mg_initialized = B_FALSE;
	mg->mg_no_free_space = B_TRUE;
	refcount_create_tracked(&mg->mg_alloc_queue_depth);

	mg->mg_taskq = taskq_create("metaslab_group_taskq", metaslab_load_pct,
	    maxclsyspri, 10, INT_MAX, TASKQ_THREADS_CPU_PCT | TASKQ_DYNAMIC);

	return (mg);
}

void
metaslab_group_destroy(metaslab_group_t *mg)
{
	ASSERT(mg->mg_prev == NULL);
	ASSERT(mg->mg_next == NULL);
	/*
	 * We may have gone below zero with the activation count
	 * either because we never activated in the first place or
	 * because we're done, and possibly removing the vdev.
	 */
	ASSERT(mg->mg_activation_count <= 0);

	taskq_destroy(mg->mg_taskq);
	avl_destroy(&mg->mg_metaslab_tree);
	mutex_destroy(&mg->mg_lock);
	refcount_destroy(&mg->mg_alloc_queue_depth);
	kmem_free(mg, sizeof (metaslab_group_t));
}

void
metaslab_group_activate(metaslab_group_t *mg)
{
	metaslab_class_t *mc = mg->mg_class;
	metaslab_group_t *mgprev, *mgnext;

	ASSERT(spa_config_held(mc->mc_spa, SCL_ALLOC, RW_WRITER));

	ASSERT(mc->mc_rotor != mg);
	ASSERT(mg->mg_prev == NULL);
	ASSERT(mg->mg_next == NULL);
	ASSERT(mg->mg_activation_count <= 0);

	if (++mg->mg_activation_count <= 0)
		return;

	mg->mg_aliquot = metaslab_aliquot * MAX(1, mg->mg_vd->vdev_children);
	metaslab_group_alloc_update(mg);

	if ((mgprev = mc->mc_rotor) == NULL) {
		mg->mg_prev = mg;
		mg->mg_next = mg;
	} else {
		mgnext = mgprev->mg_next;
		mg->mg_prev = mgprev;
		mg->mg_next = mgnext;
		mgprev->mg_next = mg;
		mgnext->mg_prev = mg;
	}
	mc->mc_rotor = mg;
}

void
metaslab_group_passivate(metaslab_group_t *mg)
{
	metaslab_class_t *mc = mg->mg_class;
	metaslab_group_t *mgprev, *mgnext;

	ASSERT(spa_config_held(mc->mc_spa, SCL_ALLOC, RW_WRITER));

	if (--mg->mg_activation_count != 0) {
		ASSERT(mc->mc_rotor != mg);
		ASSERT(mg->mg_prev == NULL);
		ASSERT(mg->mg_next == NULL);
		ASSERT(mg->mg_activation_count < 0);
		return;
	}

	taskq_wait_outstanding(mg->mg_taskq, 0);
	metaslab_group_alloc_update(mg);

	mgprev = mg->mg_prev;
	mgnext = mg->mg_next;

	if (mg == mgnext) {
		mc->mc_rotor = NULL;
	} else {
		mc->mc_rotor = mgnext;
		mgprev->mg_next = mgnext;
		mgnext->mg_prev = mgprev;
	}

	mg->mg_prev = NULL;
	mg->mg_next = NULL;
}

boolean_t
metaslab_group_initialized(metaslab_group_t *mg)
{
	vdev_t *vd = mg->mg_vd;
	vdev_stat_t *vs = &vd->vdev_stat;

	return (vs->vs_space != 0 && mg->mg_activation_count > 0);
}

uint64_t
metaslab_group_get_space(metaslab_group_t *mg)
{
	return ((1ULL << mg->mg_vd->vdev_ms_shift) * mg->mg_vd->vdev_ms_count);
}

void
metaslab_group_histogram_verify(metaslab_group_t *mg)
{
	uint64_t *mg_hist;
	vdev_t *vd = mg->mg_vd;
	uint64_t ashift = vd->vdev_ashift;
	int i, m;

	if ((zfs_flags & ZFS_DEBUG_HISTOGRAM_VERIFY) == 0)
		return;

	mg_hist = kmem_zalloc(sizeof (uint64_t) * RANGE_TREE_HISTOGRAM_SIZE,
	    KM_SLEEP);

	ASSERT3U(RANGE_TREE_HISTOGRAM_SIZE, >=,
	    SPACE_MAP_HISTOGRAM_SIZE + ashift);

	for (m = 0; m < vd->vdev_ms_count; m++) {
		metaslab_t *msp = vd->vdev_ms[m];

		if (msp->ms_sm == NULL)
			continue;

		for (i = 0; i < SPACE_MAP_HISTOGRAM_SIZE; i++)
			mg_hist[i + ashift] +=
			    msp->ms_sm->sm_phys->smp_histogram[i];
	}

	for (i = 0; i < RANGE_TREE_HISTOGRAM_SIZE; i ++)
		VERIFY3U(mg_hist[i], ==, mg->mg_histogram[i]);

	kmem_free(mg_hist, sizeof (uint64_t) * RANGE_TREE_HISTOGRAM_SIZE);
}

static void
metaslab_group_histogram_add(metaslab_group_t *mg, metaslab_t *msp)
{
	metaslab_class_t *mc = mg->mg_class;
	uint64_t ashift = mg->mg_vd->vdev_ashift;
	int i;

	ASSERT(MUTEX_HELD(&msp->ms_lock));
	if (msp->ms_sm == NULL)
		return;

	mutex_enter(&mg->mg_lock);
	for (i = 0; i < SPACE_MAP_HISTOGRAM_SIZE; i++) {
		mg->mg_histogram[i + ashift] +=
		    msp->ms_sm->sm_phys->smp_histogram[i];
		mc->mc_histogram[i + ashift] +=
		    msp->ms_sm->sm_phys->smp_histogram[i];
	}
	mutex_exit(&mg->mg_lock);
}

void
metaslab_group_histogram_remove(metaslab_group_t *mg, metaslab_t *msp)
{
	metaslab_class_t *mc = mg->mg_class;
	uint64_t ashift = mg->mg_vd->vdev_ashift;
	int i;

	ASSERT(MUTEX_HELD(&msp->ms_lock));
	if (msp->ms_sm == NULL)
		return;

	mutex_enter(&mg->mg_lock);
	for (i = 0; i < SPACE_MAP_HISTOGRAM_SIZE; i++) {
		ASSERT3U(mg->mg_histogram[i + ashift], >=,
		    msp->ms_sm->sm_phys->smp_histogram[i]);
		ASSERT3U(mc->mc_histogram[i + ashift], >=,
		    msp->ms_sm->sm_phys->smp_histogram[i]);

		mg->mg_histogram[i + ashift] -=
		    msp->ms_sm->sm_phys->smp_histogram[i];
		mc->mc_histogram[i + ashift] -=
		    msp->ms_sm->sm_phys->smp_histogram[i];
	}
	mutex_exit(&mg->mg_lock);
}

static void
metaslab_group_add(metaslab_group_t *mg, metaslab_t *msp)
{
	ASSERT(msp->ms_group == NULL);
	mutex_enter(&mg->mg_lock);
	msp->ms_group = mg;
	msp->ms_weight = 0;
	avl_add(&mg->mg_metaslab_tree, msp);
	mutex_exit(&mg->mg_lock);

	mutex_enter(&msp->ms_lock);
	metaslab_group_histogram_add(mg, msp);
	mutex_exit(&msp->ms_lock);
}

static void
metaslab_group_remove(metaslab_group_t *mg, metaslab_t *msp)
{
	mutex_enter(&msp->ms_lock);
	metaslab_group_histogram_remove(mg, msp);
	mutex_exit(&msp->ms_lock);

	mutex_enter(&mg->mg_lock);
	ASSERT(msp->ms_group == mg);
	avl_remove(&mg->mg_metaslab_tree, msp);
	msp->ms_group = NULL;
	mutex_exit(&mg->mg_lock);
}

static void
metaslab_group_sort(metaslab_group_t *mg, metaslab_t *msp, uint64_t weight)
{
	/*
	 * Although in principle the weight can be any value, in
	 * practice we do not use values in the range [1, 511].
	 */
	ASSERT(weight >= SPA_MINBLOCKSIZE || weight == 0);
	ASSERT(MUTEX_HELD(&msp->ms_lock));

	mutex_enter(&mg->mg_lock);
	ASSERT(msp->ms_group == mg);
	avl_remove(&mg->mg_metaslab_tree, msp);
	msp->ms_weight = weight;
	avl_add(&mg->mg_metaslab_tree, msp);
	mutex_exit(&mg->mg_lock);
}

/*
 * Calculate the fragmentation for a given metaslab group. We can use
 * a simple average here since all metaslabs within the group must have
 * the same size. The return value will be a value between 0 and 100
 * (inclusive), or ZFS_FRAG_INVALID if less than half of the metaslab in this
 * group have a fragmentation metric.
 */
uint64_t
metaslab_group_fragmentation(metaslab_group_t *mg)
{
	vdev_t *vd = mg->mg_vd;
	uint64_t fragmentation = 0;
	uint64_t valid_ms = 0;
	int m;

	for (m = 0; m < vd->vdev_ms_count; m++) {
		metaslab_t *msp = vd->vdev_ms[m];

		if (msp->ms_fragmentation == ZFS_FRAG_INVALID)
			continue;

		valid_ms++;
		fragmentation += msp->ms_fragmentation;
	}

	if (valid_ms <= vd->vdev_ms_count / 2)
		return (ZFS_FRAG_INVALID);

	fragmentation /= valid_ms;
	ASSERT3U(fragmentation, <=, 100);
	return (fragmentation);
}

/*
 * Determine if a given metaslab group should skip allocations. A metaslab
 * group should avoid allocations if its free capacity is less than the
 * zfs_mg_noalloc_threshold or its fragmentation metric is greater than
 * zfs_mg_fragmentation_threshold and there is at least one metaslab group
 * that can still handle allocations. If the allocation throttle is enabled
 * then we skip allocations to devices that have reached their maximum
 * allocation queue depth unless the selected metaslab group is the only
 * eligible group remaining.
 */
static boolean_t
metaslab_group_allocatable(metaslab_group_t *mg, metaslab_group_t *rotor,
    uint64_t psize)
{
	spa_t *spa = mg->mg_vd->vdev_spa;
	metaslab_class_t *mc = mg->mg_class;

	/*
	 * We can only consider skipping this metaslab group if it's
	 * in the normal metaslab class and there are other metaslab
	 * groups to select from. Otherwise, we always consider it eligible
	 * for allocations.
	 */
	if (mc != spa_normal_class(spa) || mc->mc_groups <= 1)
		return (B_TRUE);

	/*
	 * If the metaslab group's mg_allocatable flag is set (see comments
	 * in metaslab_group_alloc_update() for more information) and
	 * the allocation throttle is disabled then allow allocations to this
	 * device. However, if the allocation throttle is enabled then
	 * check if we have reached our allocation limit (mg_alloc_queue_depth)
	 * to determine if we should allow allocations to this metaslab group.
	 * If all metaslab groups are no longer considered allocatable
	 * (mc_alloc_groups == 0) or we're trying to allocate the smallest
	 * gang block size then we allow allocations on this metaslab group
	 * regardless of the mg_allocatable or throttle settings.
	 */
	if (mg->mg_allocatable) {
		metaslab_group_t *mgp;
		int64_t qdepth;
		uint64_t qmax = mg->mg_max_alloc_queue_depth;

		if (!mc->mc_alloc_throttle_enabled)
			return (B_TRUE);

		/*
		 * If this metaslab group does not have any free space, then
		 * there is no point in looking further.
		 */
		if (mg->mg_no_free_space)
			return (B_FALSE);

		qdepth = refcount_count(&mg->mg_alloc_queue_depth);

		/*
		 * If this metaslab group is below its qmax or it's
		 * the only allocatable metasable group, then attempt
		 * to allocate from it.
		 */
		if (qdepth < qmax || mc->mc_alloc_groups == 1)
			return (B_TRUE);
		ASSERT3U(mc->mc_alloc_groups, >, 1);

		/*
		 * Since this metaslab group is at or over its qmax, we
		 * need to determine if there are metaslab groups after this
		 * one that might be able to handle this allocation. This is
		 * racy since we can't hold the locks for all metaslab
		 * groups at the same time when we make this check.
		 */
		for (mgp = mg->mg_next; mgp != rotor; mgp = mgp->mg_next) {
			qmax = mgp->mg_max_alloc_queue_depth;

			qdepth = refcount_count(&mgp->mg_alloc_queue_depth);

			/*
			 * If there is another metaslab group that
			 * might be able to handle the allocation, then
			 * we return false so that we skip this group.
			 */
			if (qdepth < qmax && !mgp->mg_no_free_space)
				return (B_FALSE);
		}

		/*
		 * We didn't find another group to handle the allocation
		 * so we can't skip this metaslab group even though
		 * we are at or over our qmax.
		 */
		return (B_TRUE);

	} else if (mc->mc_alloc_groups == 0 || psize == SPA_MINBLOCKSIZE) {
		return (B_TRUE);
	}
	return (B_FALSE);
}

/*
 * ==========================================================================
 * Range tree callbacks
 * ==========================================================================
 */

/*
 * Comparison function for the private size-ordered tree. Tree is sorted
 * by size, larger sizes at the end of the tree.
 */
static int
metaslab_rangesize_compare(const void *x1, const void *x2)
{
	const range_seg_t *r1 = x1;
	const range_seg_t *r2 = x2;
	uint64_t rs_size1 = r1->rs_end - r1->rs_start;
	uint64_t rs_size2 = r2->rs_end - r2->rs_start;

	int cmp = AVL_CMP(rs_size1, rs_size2);
	if (likely(cmp))
		return (cmp);

	return (AVL_CMP(r1->rs_start, r2->rs_start));
}

/*
 * Create any block allocator specific components. The current allocators
 * rely on using both a size-ordered range_tree_t and an array of uint64_t's.
 */
static void
metaslab_rt_create(range_tree_t *rt, void *arg)
{
	metaslab_t *msp = arg;

	ASSERT3P(rt->rt_arg, ==, msp);
	ASSERT(msp->ms_tree == NULL);

	avl_create(&msp->ms_size_tree, metaslab_rangesize_compare,
	    sizeof (range_seg_t), offsetof(range_seg_t, rs_pp_node));
}

/*
 * Destroy the block allocator specific components.
 */
static void
metaslab_rt_destroy(range_tree_t *rt, void *arg)
{
	metaslab_t *msp = arg;

	ASSERT3P(rt->rt_arg, ==, msp);
	ASSERT3P(msp->ms_tree, ==, rt);
	ASSERT0(avl_numnodes(&msp->ms_size_tree));

	avl_destroy(&msp->ms_size_tree);
}

static void
metaslab_rt_add(range_tree_t *rt, range_seg_t *rs, void *arg)
{
	metaslab_t *msp = arg;

	ASSERT3P(rt->rt_arg, ==, msp);
	ASSERT3P(msp->ms_tree, ==, rt);
	VERIFY(!msp->ms_condensing);
	avl_add(&msp->ms_size_tree, rs);
}

static void
metaslab_rt_remove(range_tree_t *rt, range_seg_t *rs, void *arg)
{
	metaslab_t *msp = arg;

	ASSERT3P(rt->rt_arg, ==, msp);
	ASSERT3P(msp->ms_tree, ==, rt);
	VERIFY(!msp->ms_condensing);
	avl_remove(&msp->ms_size_tree, rs);
}

static void
metaslab_rt_vacate(range_tree_t *rt, void *arg)
{
	metaslab_t *msp = arg;

	ASSERT3P(rt->rt_arg, ==, msp);
	ASSERT3P(msp->ms_tree, ==, rt);

	/*
	 * Normally one would walk the tree freeing nodes along the way.
	 * Since the nodes are shared with the range trees we can avoid
	 * walking all nodes and just reinitialize the avl tree. The nodes
	 * will be freed by the range tree, so we don't want to free them here.
	 */
	avl_create(&msp->ms_size_tree, metaslab_rangesize_compare,
	    sizeof (range_seg_t), offsetof(range_seg_t, rs_pp_node));
}

static range_tree_ops_t metaslab_rt_ops = {
	metaslab_rt_create,
	metaslab_rt_destroy,
	metaslab_rt_add,
	metaslab_rt_remove,
	metaslab_rt_vacate
};

/*
 * ==========================================================================
 * Metaslab block operations
 * ==========================================================================
 */

/*
 * Return the maximum contiguous segment within the metaslab.
 */
uint64_t
metaslab_block_maxsize(metaslab_t *msp)
{
	avl_tree_t *t = &msp->ms_size_tree;
	range_seg_t *rs;

	if (t == NULL || (rs = avl_last(t)) == NULL)
		return (0ULL);

	return (rs->rs_end - rs->rs_start);
}

uint64_t
metaslab_block_alloc(metaslab_t *msp, uint64_t size)
{
	uint64_t start;
	range_tree_t *rt = msp->ms_tree;

	VERIFY(!msp->ms_condensing);

	start = msp->ms_ops->msop_alloc(msp, size);
	if (start != -1ULL) {
		vdev_t *vd = msp->ms_group->mg_vd;

		VERIFY0(P2PHASE(start, 1ULL << vd->vdev_ashift));
		VERIFY0(P2PHASE(size, 1ULL << vd->vdev_ashift));
		VERIFY3U(range_tree_space(rt) - size, <=, msp->ms_size);
		range_tree_remove(rt, start, size);
	}
	return (start);
}

/*
 * ==========================================================================
 * Common allocator routines
 * ==========================================================================
 */

#if defined(WITH_FF_BLOCK_ALLOCATOR) || \
    defined(WITH_DF_BLOCK_ALLOCATOR) || \
    defined(WITH_CF_BLOCK_ALLOCATOR)
/*
 * This is a helper function that can be used by the allocator to find
 * a suitable block to allocate. This will search the specified AVL
 * tree looking for a block that matches the specified criteria.
 */
static uint64_t
metaslab_block_picker(avl_tree_t *t, uint64_t *cursor, uint64_t size,
    uint64_t align)
{
	range_seg_t *rs, rsearch;
	avl_index_t where;

	rsearch.rs_start = *cursor;
	rsearch.rs_end = *cursor + size;

	rs = avl_find(t, &rsearch, &where);
	if (rs == NULL)
		rs = avl_nearest(t, where, AVL_AFTER);

	while (rs != NULL) {
		uint64_t offset = P2ROUNDUP(rs->rs_start, align);

		if (offset + size <= rs->rs_end) {
			*cursor = offset + size;
			return (offset);
		}
		rs = AVL_NEXT(t, rs);
	}

	/*
	 * If we know we've searched the whole map (*cursor == 0), give up.
	 * Otherwise, reset the cursor to the beginning and try again.
	 */
	if (*cursor == 0)
		return (-1ULL);

	*cursor = 0;
	return (metaslab_block_picker(t, cursor, size, align));
}
#endif /* WITH_FF/DF/CF_BLOCK_ALLOCATOR */

#if defined(WITH_FF_BLOCK_ALLOCATOR)
/*
 * ==========================================================================
 * The first-fit block allocator
 * ==========================================================================
 */
static uint64_t
metaslab_ff_alloc(metaslab_t *msp, uint64_t size)
{
	/*
	 * Find the largest power of 2 block size that evenly divides the
	 * requested size. This is used to try to allocate blocks with similar
	 * alignment from the same area of the metaslab (i.e. same cursor
	 * bucket) but it does not guarantee that other allocations sizes
	 * may exist in the same region.
	 */
	uint64_t align = size & -size;
	uint64_t *cursor = &msp->ms_lbas[highbit64(align) - 1];
	avl_tree_t *t = &msp->ms_tree->rt_root;

	return (metaslab_block_picker(t, cursor, size, align));
}

static metaslab_ops_t metaslab_ff_ops = {
	metaslab_ff_alloc
};

metaslab_ops_t *zfs_metaslab_ops = &metaslab_ff_ops;
#endif /* WITH_FF_BLOCK_ALLOCATOR */

#if defined(WITH_DF_BLOCK_ALLOCATOR)
/*
 * ==========================================================================
 * Dynamic block allocator -
 * Uses the first fit allocation scheme until space get low and then
 * adjusts to a best fit allocation method. Uses metaslab_df_alloc_threshold
 * and metaslab_df_free_pct to determine when to switch the allocation scheme.
 * ==========================================================================
 */
static uint64_t
metaslab_df_alloc(metaslab_t *msp, uint64_t size)
{
	/*
	 * Find the largest power of 2 block size that evenly divides the
	 * requested size. This is used to try to allocate blocks with similar
	 * alignment from the same area of the metaslab (i.e. same cursor
	 * bucket) but it does not guarantee that other allocations sizes
	 * may exist in the same region.
	 */
	uint64_t align = size & -size;
	uint64_t *cursor = &msp->ms_lbas[highbit64(align) - 1];
	range_tree_t *rt = msp->ms_tree;
	avl_tree_t *t = &rt->rt_root;
	uint64_t max_size = metaslab_block_maxsize(msp);
	int free_pct = range_tree_space(rt) * 100 / msp->ms_size;

	ASSERT(MUTEX_HELD(&msp->ms_lock));
	ASSERT3U(avl_numnodes(t), ==, avl_numnodes(&msp->ms_size_tree));

	if (max_size < size)
		return (-1ULL);

	/*
	 * If we're running low on space switch to using the size
	 * sorted AVL tree (best-fit).
	 */
	if (max_size < metaslab_df_alloc_threshold ||
	    free_pct < metaslab_df_free_pct) {
		t = &msp->ms_size_tree;
		*cursor = 0;
	}

	return (metaslab_block_picker(t, cursor, size, 1ULL));
}

static metaslab_ops_t metaslab_df_ops = {
	metaslab_df_alloc
};

metaslab_ops_t *zfs_metaslab_ops = &metaslab_df_ops;
#endif /* WITH_DF_BLOCK_ALLOCATOR */

#if defined(WITH_CF_BLOCK_ALLOCATOR)
/*
 * ==========================================================================
 * Cursor fit block allocator -
 * Select the largest region in the metaslab, set the cursor to the beginning
 * of the range and the cursor_end to the end of the range. As allocations
 * are made advance the cursor. Continue allocating from the cursor until
 * the range is exhausted and then find a new range.
 * ==========================================================================
 */
static uint64_t
metaslab_cf_alloc(metaslab_t *msp, uint64_t size)
{
	range_tree_t *rt = msp->ms_tree;
	avl_tree_t *t = &msp->ms_size_tree;
	uint64_t *cursor = &msp->ms_lbas[0];
	uint64_t *cursor_end = &msp->ms_lbas[1];
	uint64_t offset = 0;

	ASSERT(MUTEX_HELD(&msp->ms_lock));
	ASSERT3U(avl_numnodes(t), ==, avl_numnodes(&rt->rt_root));

	ASSERT3U(*cursor_end, >=, *cursor);

	if ((*cursor + size) > *cursor_end) {
		range_seg_t *rs;

		rs = avl_last(&msp->ms_size_tree);
		if (rs == NULL || (rs->rs_end - rs->rs_start) < size)
			return (-1ULL);

		*cursor = rs->rs_start;
		*cursor_end = rs->rs_end;
	}

	offset = *cursor;
	*cursor += size;

	return (offset);
}

static metaslab_ops_t metaslab_cf_ops = {
	metaslab_cf_alloc
};

metaslab_ops_t *zfs_metaslab_ops = &metaslab_cf_ops;
#endif /* WITH_CF_BLOCK_ALLOCATOR */

#if defined(WITH_NDF_BLOCK_ALLOCATOR)
/*
 * ==========================================================================
 * New dynamic fit allocator -
 * Select a region that is large enough to allocate 2^metaslab_ndf_clump_shift
 * contiguous blocks. If no region is found then just use the largest segment
 * that remains.
 * ==========================================================================
 */

/*
 * Determines desired number of contiguous blocks (2^metaslab_ndf_clump_shift)
 * to request from the allocator.
 */
uint64_t metaslab_ndf_clump_shift = 4;

static uint64_t
metaslab_ndf_alloc(metaslab_t *msp, uint64_t size)
{
	avl_tree_t *t = &msp->ms_tree->rt_root;
	avl_index_t where;
	range_seg_t *rs, rsearch;
	uint64_t hbit = highbit64(size);
	uint64_t *cursor = &msp->ms_lbas[hbit - 1];
	uint64_t max_size = metaslab_block_maxsize(msp);

	ASSERT(MUTEX_HELD(&msp->ms_lock));
	ASSERT3U(avl_numnodes(t), ==, avl_numnodes(&msp->ms_size_tree));

	if (max_size < size)
		return (-1ULL);

	rsearch.rs_start = *cursor;
	rsearch.rs_end = *cursor + size;

	rs = avl_find(t, &rsearch, &where);
	if (rs == NULL || (rs->rs_end - rs->rs_start) < size) {
		t = &msp->ms_size_tree;

		rsearch.rs_start = 0;
		rsearch.rs_end = MIN(max_size,
		    1ULL << (hbit + metaslab_ndf_clump_shift));
		rs = avl_find(t, &rsearch, &where);
		if (rs == NULL)
			rs = avl_nearest(t, where, AVL_AFTER);
		ASSERT(rs != NULL);
	}

	if ((rs->rs_end - rs->rs_start) >= size) {
		*cursor = rs->rs_start + size;
		return (rs->rs_start);
	}
	return (-1ULL);
}

static metaslab_ops_t metaslab_ndf_ops = {
	metaslab_ndf_alloc
};

metaslab_ops_t *zfs_metaslab_ops = &metaslab_ndf_ops;
#endif /* WITH_NDF_BLOCK_ALLOCATOR */


/*
 * ==========================================================================
 * Metaslabs
 * ==========================================================================
 */

/*
 * Wait for any in-progress metaslab loads to complete.
 */
void
metaslab_load_wait(metaslab_t *msp)
{
	ASSERT(MUTEX_HELD(&msp->ms_lock));

	while (msp->ms_loading) {
		ASSERT(!msp->ms_loaded);
		cv_wait(&msp->ms_load_cv, &msp->ms_lock);
	}
}

int
metaslab_load(metaslab_t *msp)
{
	int error = 0;
	int t;

	ASSERT(MUTEX_HELD(&msp->ms_lock));
	ASSERT(!msp->ms_loaded);
	ASSERT(!msp->ms_loading);

	msp->ms_loading = B_TRUE;

	/*
	 * If the space map has not been allocated yet, then treat
	 * all the space in the metaslab as free and add it to the
	 * ms_tree.
	 */
	if (msp->ms_sm != NULL)
		error = space_map_load(msp->ms_sm, msp->ms_tree, SM_FREE);
	else
		range_tree_add(msp->ms_tree, msp->ms_start, msp->ms_size);

	msp->ms_loaded = (error == 0);
	msp->ms_loading = B_FALSE;

	if (msp->ms_loaded) {
		for (t = 0; t < TXG_DEFER_SIZE; t++) {
			range_tree_walk(msp->ms_defertree[t],
			    range_tree_remove, msp->ms_tree);
		}
	}
	cv_broadcast(&msp->ms_load_cv);
	return (error);
}

void
metaslab_unload(metaslab_t *msp)
{
	ASSERT(MUTEX_HELD(&msp->ms_lock));
	range_tree_vacate(msp->ms_tree, NULL, NULL);
	msp->ms_loaded = B_FALSE;
	msp->ms_weight &= ~METASLAB_ACTIVE_MASK;
}

int
metaslab_init(metaslab_group_t *mg, uint64_t id, uint64_t object, uint64_t txg,
    metaslab_t **msp)
{
	vdev_t *vd = mg->mg_vd;
	objset_t *mos = vd->vdev_spa->spa_meta_objset;
	metaslab_t *ms;
	int error;

	ms = kmem_zalloc(sizeof (metaslab_t), KM_SLEEP);
	mutex_init(&ms->ms_lock, NULL, MUTEX_DEFAULT, NULL);
	cv_init(&ms->ms_load_cv, NULL, CV_DEFAULT, NULL);
	ms->ms_id = id;
	ms->ms_start = id << vd->vdev_ms_shift;
	ms->ms_size = 1ULL << vd->vdev_ms_shift;

	/*
	 * We only open space map objects that already exist. All others
	 * will be opened when we finally allocate an object for it.
	 */
	if (object != 0) {
		error = space_map_open(&ms->ms_sm, mos, object, ms->ms_start,
		    ms->ms_size, vd->vdev_ashift, &ms->ms_lock);

		if (error != 0) {
			kmem_free(ms, sizeof (metaslab_t));
			return (error);
		}

		ASSERT(ms->ms_sm != NULL);
	}

	/*
	 * We create the main range tree here, but we don't create the
	 * alloctree and freetree until metaslab_sync_done().  This serves
	 * two purposes: it allows metaslab_sync_done() to detect the
	 * addition of new space; and for debugging, it ensures that we'd
	 * data fault on any attempt to use this metaslab before it's ready.
	 */
	ms->ms_tree = range_tree_create(&metaslab_rt_ops, ms, &ms->ms_lock);
	metaslab_group_add(mg, ms);

	ms->ms_fragmentation = metaslab_fragmentation(ms);
	ms->ms_ops = mg->mg_class->mc_ops;

	/*
	 * If we're opening an existing pool (txg == 0) or creating
	 * a new one (txg == TXG_INITIAL), all space is available now.
	 * If we're adding space to an existing pool, the new space
	 * does not become available until after this txg has synced.
	 */
	if (txg <= TXG_INITIAL)
		metaslab_sync_done(ms, 0);

	/*
	 * If metaslab_debug_load is set and we're initializing a metaslab
	 * that has an allocated space_map object then load the its space
	 * map so that can verify frees.
	 */
	if (metaslab_debug_load && ms->ms_sm != NULL) {
		mutex_enter(&ms->ms_lock);
		VERIFY0(metaslab_load(ms));
		mutex_exit(&ms->ms_lock);
	}

	if (txg != 0) {
		vdev_dirty(vd, 0, NULL, txg);
		vdev_dirty(vd, VDD_METASLAB, ms, txg);
	}

	*msp = ms;

	return (0);
}

void
metaslab_fini(metaslab_t *msp)
{
	int t;

	metaslab_group_t *mg = msp->ms_group;

	metaslab_group_remove(mg, msp);

	mutex_enter(&msp->ms_lock);

	VERIFY(msp->ms_group == NULL);
	vdev_space_update(mg->mg_vd, -space_map_allocated(msp->ms_sm),
	    0, -msp->ms_size);
	space_map_close(msp->ms_sm);

	metaslab_unload(msp);
	range_tree_destroy(msp->ms_tree);

	for (t = 0; t < TXG_SIZE; t++) {
		range_tree_destroy(msp->ms_alloctree[t]);
		range_tree_destroy(msp->ms_freetree[t]);
	}

	for (t = 0; t < TXG_DEFER_SIZE; t++) {
		range_tree_destroy(msp->ms_defertree[t]);
	}

	ASSERT0(msp->ms_deferspace);

	mutex_exit(&msp->ms_lock);
	cv_destroy(&msp->ms_load_cv);
	mutex_destroy(&msp->ms_lock);

	kmem_free(msp, sizeof (metaslab_t));
}

#define	FRAGMENTATION_TABLE_SIZE	17

/*
 * This table defines a segment size based fragmentation metric that will
 * allow each metaslab to derive its own fragmentation value. This is done
 * by calculating the space in each bucket of the spacemap histogram and
 * multiplying that by the fragmetation metric in this table. Doing
 * this for all buckets and dividing it by the total amount of free
 * space in this metaslab (i.e. the total free space in all buckets) gives
 * us the fragmentation metric. This means that a high fragmentation metric
 * equates to most of the free space being comprised of small segments.
 * Conversely, if the metric is low, then most of the free space is in
 * large segments. A 10% change in fragmentation equates to approximately
 * double the number of segments.
 *
 * This table defines 0% fragmented space using 16MB segments. Testing has
 * shown that segments that are greater than or equal to 16MB do not suffer
 * from drastic performance problems. Using this value, we derive the rest
 * of the table. Since the fragmentation value is never stored on disk, it
 * is possible to change these calculations in the future.
 */
int zfs_frag_table[FRAGMENTATION_TABLE_SIZE] = {
	100,	/* 512B	*/
	100,	/* 1K	*/
	98,	/* 2K	*/
	95,	/* 4K	*/
	90,	/* 8K	*/
	80,	/* 16K	*/
	70,	/* 32K	*/
	60,	/* 64K	*/
	50,	/* 128K	*/
	40,	/* 256K	*/
	30,	/* 512K	*/
	20,	/* 1M	*/
	15,	/* 2M	*/
	10,	/* 4M	*/
	5,	/* 8M	*/
	0	/* 16M	*/
};

/*
 * Calclate the metaslab's fragmentation metric. A return value
 * of ZFS_FRAG_INVALID means that the metaslab has not been upgraded and does
 * not support this metric. Otherwise, the return value should be in the
 * range [0, 100].
 */
static uint64_t
metaslab_fragmentation(metaslab_t *msp)
{
	spa_t *spa = msp->ms_group->mg_vd->vdev_spa;
	uint64_t fragmentation = 0;
	uint64_t total = 0;
	boolean_t feature_enabled = spa_feature_is_enabled(spa,
	    SPA_FEATURE_SPACEMAP_HISTOGRAM);
	int i;

	if (!feature_enabled)
		return (ZFS_FRAG_INVALID);

	/*
	 * A null space map means that the entire metaslab is free
	 * and thus is not fragmented.
	 */
	if (msp->ms_sm == NULL)
		return (0);

	/*
	 * If this metaslab's space_map has not been upgraded, flag it
	 * so that we upgrade next time we encounter it.
	 */
	if (msp->ms_sm->sm_dbuf->db_size != sizeof (space_map_phys_t)) {
		vdev_t *vd = msp->ms_group->mg_vd;

		if (spa_writeable(vd->vdev_spa)) {
			uint64_t txg = spa_syncing_txg(spa);

			msp->ms_condense_wanted = B_TRUE;
			vdev_dirty(vd, VDD_METASLAB, msp, txg + 1);
			spa_dbgmsg(spa, "txg %llu, requesting force condense: "
			    "msp %p, vd %p", txg, msp, vd);
		}
		return (ZFS_FRAG_INVALID);
	}

	for (i = 0; i < SPACE_MAP_HISTOGRAM_SIZE; i++) {
		uint64_t space = 0;
		uint8_t shift = msp->ms_sm->sm_shift;
		int idx = MIN(shift - SPA_MINBLOCKSHIFT + i,
		    FRAGMENTATION_TABLE_SIZE - 1);

		if (msp->ms_sm->sm_phys->smp_histogram[i] == 0)
			continue;

		space = msp->ms_sm->sm_phys->smp_histogram[i] << (i + shift);
		total += space;

		ASSERT3U(idx, <, FRAGMENTATION_TABLE_SIZE);
		fragmentation += space * zfs_frag_table[idx];
	}

	if (total > 0)
		fragmentation /= total;
	ASSERT3U(fragmentation, <=, 100);
	return (fragmentation);
}

/*
 * Compute a weight -- a selection preference value -- for the given metaslab.
 * This is based on the amount of free space, the level of fragmentation,
 * the LBA range, and whether the metaslab is loaded.
 */
static uint64_t
metaslab_weight(metaslab_t *msp)
{
	metaslab_group_t *mg = msp->ms_group;
	vdev_t *vd = mg->mg_vd;
	uint64_t weight, space;

	ASSERT(MUTEX_HELD(&msp->ms_lock));

	/*
	 * This vdev is in the process of being removed so there is nothing
	 * for us to do here.
	 */
	if (vd->vdev_removing) {
		ASSERT0(space_map_allocated(msp->ms_sm));
		ASSERT0(vd->vdev_ms_shift);
		return (0);
	}

	/*
	 * The baseline weight is the metaslab's free space.
	 */
	space = msp->ms_size - space_map_allocated(msp->ms_sm);

	msp->ms_fragmentation = metaslab_fragmentation(msp);
	if (metaslab_fragmentation_factor_enabled &&
	    msp->ms_fragmentation != ZFS_FRAG_INVALID) {
		/*
		 * Use the fragmentation information to inversely scale
		 * down the baseline weight. We need to ensure that we
		 * don't exclude this metaslab completely when it's 100%
		 * fragmented. To avoid this we reduce the fragmented value
		 * by 1.
		 */
		space = (space * (100 - (msp->ms_fragmentation - 1))) / 100;

		/*
		 * If space < SPA_MINBLOCKSIZE, then we will not allocate from
		 * this metaslab again. The fragmentation metric may have
		 * decreased the space to something smaller than
		 * SPA_MINBLOCKSIZE, so reset the space to SPA_MINBLOCKSIZE
		 * so that we can consume any remaining space.
		 */
		if (space > 0 && space < SPA_MINBLOCKSIZE)
			space = SPA_MINBLOCKSIZE;
	}
	weight = space;

	/*
	 * Modern disks have uniform bit density and constant angular velocity.
	 * Therefore, the outer recording zones are faster (higher bandwidth)
	 * than the inner zones by the ratio of outer to inner track diameter,
	 * which is typically around 2:1.  We account for this by assigning
	 * higher weight to lower metaslabs (multiplier ranging from 2x to 1x).
	 * In effect, this means that we'll select the metaslab with the most
	 * free bandwidth rather than simply the one with the most free space.
	 */
	if (!vd->vdev_nonrot && metaslab_lba_weighting_enabled) {
		weight = 2 * weight - (msp->ms_id * weight) / vd->vdev_ms_count;
		ASSERT(weight >= space && weight <= 2 * space);
	}

	/*
	 * If this metaslab is one we're actively using, adjust its
	 * weight to make it preferable to any inactive metaslab so
	 * we'll polish it off. If the fragmentation on this metaslab
	 * has exceed our threshold, then don't mark it active.
	 */
	if (msp->ms_loaded && msp->ms_fragmentation != ZFS_FRAG_INVALID &&
	    msp->ms_fragmentation <= zfs_metaslab_fragmentation_threshold) {
		weight |= (msp->ms_weight & METASLAB_ACTIVE_MASK);
	}

	return (weight);
}

static int
metaslab_activate(metaslab_t *msp, uint64_t activation_weight)
{
	ASSERT(MUTEX_HELD(&msp->ms_lock));

	if ((msp->ms_weight & METASLAB_ACTIVE_MASK) == 0) {
		metaslab_load_wait(msp);
		if (!msp->ms_loaded) {
			int error = metaslab_load(msp);
			if (error) {
				metaslab_group_sort(msp->ms_group, msp, 0);
				return (error);
			}
		}

		metaslab_group_sort(msp->ms_group, msp,
		    msp->ms_weight | activation_weight);
	}
	ASSERT(msp->ms_loaded);
	ASSERT(msp->ms_weight & METASLAB_ACTIVE_MASK);

	return (0);
}

static void
metaslab_passivate(metaslab_t *msp, uint64_t size)
{
	/*
	 * If size < SPA_MINBLOCKSIZE, then we will not allocate from
	 * this metaslab again.  In that case, it had better be empty,
	 * or we would be leaving space on the table.
	 */
	ASSERT(size >= SPA_MINBLOCKSIZE || range_tree_space(msp->ms_tree) == 0);
	metaslab_group_sort(msp->ms_group, msp, MIN(msp->ms_weight, size));
	ASSERT((msp->ms_weight & METASLAB_ACTIVE_MASK) == 0);
}

static void
metaslab_preload(void *arg)
{
	metaslab_t *msp = arg;
	spa_t *spa = msp->ms_group->mg_vd->vdev_spa;
	fstrans_cookie_t cookie = spl_fstrans_mark();

	ASSERT(!MUTEX_HELD(&msp->ms_group->mg_lock));

	mutex_enter(&msp->ms_lock);
	metaslab_load_wait(msp);
	if (!msp->ms_loaded)
		(void) metaslab_load(msp);

	/*
	 * Set the ms_access_txg value so that we don't unload it right away.
	 */
	msp->ms_access_txg = spa_syncing_txg(spa) + metaslab_unload_delay + 1;
	mutex_exit(&msp->ms_lock);
	spl_fstrans_unmark(cookie);
}

static void
metaslab_group_preload(metaslab_group_t *mg)
{
	spa_t *spa = mg->mg_vd->vdev_spa;
	metaslab_t *msp;
	avl_tree_t *t = &mg->mg_metaslab_tree;
	int m = 0;

	if (spa_shutting_down(spa) || !metaslab_preload_enabled) {
		taskq_wait_outstanding(mg->mg_taskq, 0);
		return;
	}

	mutex_enter(&mg->mg_lock);
	/*
	 * Load the next potential metaslabs
	 */
	msp = avl_first(t);
	while (msp != NULL) {
		metaslab_t *msp_next = AVL_NEXT(t, msp);

		/*
		 * We preload only the maximum number of metaslabs specified
		 * by metaslab_preload_limit. If a metaslab is being forced
		 * to condense then we preload it too. This will ensure
		 * that force condensing happens in the next txg.
		 */
		if (++m > metaslab_preload_limit && !msp->ms_condense_wanted) {
			msp = msp_next;
			continue;
		}

		/*
		 * We must drop the metaslab group lock here to preserve
		 * lock ordering with the ms_lock (when grabbing both
		 * the mg_lock and the ms_lock, the ms_lock must be taken
		 * first).  As a result, it is possible that the ordering
		 * of the metaslabs within the avl tree may change before
		 * we reacquire the lock. The metaslab cannot be removed from
		 * the tree while we're in syncing context so it is safe to
		 * drop the mg_lock here. If the metaslabs are reordered
		 * nothing will break -- we just may end up loading a
		 * less than optimal one.
		 */
		mutex_exit(&mg->mg_lock);
		VERIFY(taskq_dispatch(mg->mg_taskq, metaslab_preload,
		    msp, TQ_SLEEP) != TASKQID_INVALID);
		mutex_enter(&mg->mg_lock);
		msp = msp_next;
	}
	mutex_exit(&mg->mg_lock);
}

/*
 * Determine if the space map's on-disk footprint is past our tolerance
 * for inefficiency. We would like to use the following criteria to make
 * our decision:
 *
 * 1. The size of the space map object should not dramatically increase as a
 * result of writing out the free space range tree.
 *
 * 2. The minimal on-disk space map representation is zfs_condense_pct/100
 * times the size than the free space range tree representation
 * (i.e. zfs_condense_pct = 110 and in-core = 1MB, minimal = 1.1.MB).
 *
 * 3. The on-disk size of the space map should actually decrease.
 *
 * Checking the first condition is tricky since we don't want to walk
 * the entire AVL tree calculating the estimated on-disk size. Instead we
 * use the size-ordered range tree in the metaslab and calculate the
 * size required to write out the largest segment in our free tree. If the
 * size required to represent that segment on disk is larger than the space
 * map object then we avoid condensing this map.
 *
 * To determine the second criterion we use a best-case estimate and assume
 * each segment can be represented on-disk as a single 64-bit entry. We refer
 * to this best-case estimate as the space map's minimal form.
 *
 * Unfortunately, we cannot compute the on-disk size of the space map in this
 * context because we cannot accurately compute the effects of compression, etc.
 * Instead, we apply the heuristic described in the block comment for
 * zfs_metaslab_condense_block_threshold - we only condense if the space used
 * is greater than a threshold number of blocks.
 */
static boolean_t
metaslab_should_condense(metaslab_t *msp)
{
	space_map_t *sm = msp->ms_sm;
	range_seg_t *rs;
	uint64_t size, entries, segsz, object_size, optimal_size, record_size;
	dmu_object_info_t doi;
	uint64_t vdev_blocksize = 1ULL << msp->ms_group->mg_vd->vdev_ashift;

	ASSERT(MUTEX_HELD(&msp->ms_lock));
	ASSERT(msp->ms_loaded);

	/*
	 * Use the ms_size_tree range tree, which is ordered by size, to
	 * obtain the largest segment in the free tree. We always condense
	 * metaslabs that are empty and metaslabs for which a condense
	 * request has been made.
	 */
	rs = avl_last(&msp->ms_size_tree);
	if (rs == NULL || msp->ms_condense_wanted)
		return (B_TRUE);

	/*
	 * Calculate the number of 64-bit entries this segment would
	 * require when written to disk. If this single segment would be
	 * larger on-disk than the entire current on-disk structure, then
	 * clearly condensing will increase the on-disk structure size.
	 */
	size = (rs->rs_end - rs->rs_start) >> sm->sm_shift;
	entries = size / (MIN(size, SM_RUN_MAX));
	segsz = entries * sizeof (uint64_t);

	optimal_size = sizeof (uint64_t) * avl_numnodes(&msp->ms_tree->rt_root);
	object_size = space_map_length(msp->ms_sm);

	dmu_object_info_from_db(sm->sm_dbuf, &doi);
	record_size = MAX(doi.doi_data_block_size, vdev_blocksize);

	return (segsz <= object_size &&
	    object_size >= (optimal_size * zfs_condense_pct / 100) &&
	    object_size > zfs_metaslab_condense_block_threshold * record_size);
}

/*
 * Condense the on-disk space map representation to its minimized form.
 * The minimized form consists of a small number of allocations followed by
 * the entries of the free range tree.
 */
static void
metaslab_condense(metaslab_t *msp, uint64_t txg, dmu_tx_t *tx)
{
	spa_t *spa = msp->ms_group->mg_vd->vdev_spa;
	range_tree_t *freetree = msp->ms_freetree[txg & TXG_MASK];
	range_tree_t *condense_tree;
	space_map_t *sm = msp->ms_sm;
	int t;

	ASSERT(MUTEX_HELD(&msp->ms_lock));
	ASSERT3U(spa_sync_pass(spa), ==, 1);
	ASSERT(msp->ms_loaded);


	spa_dbgmsg(spa, "condensing: txg %llu, msp[%llu] %p, vdev id %llu, "
	    "spa %s, smp size %llu, segments %lu, forcing condense=%s", txg,
	    msp->ms_id, msp, msp->ms_group->mg_vd->vdev_id,
	    msp->ms_group->mg_vd->vdev_spa->spa_name,
	    space_map_length(msp->ms_sm), avl_numnodes(&msp->ms_tree->rt_root),
	    msp->ms_condense_wanted ? "TRUE" : "FALSE");

	msp->ms_condense_wanted = B_FALSE;

	/*
	 * Create an range tree that is 100% allocated. We remove segments
	 * that have been freed in this txg, any deferred frees that exist,
	 * and any allocation in the future. Removing segments should be
	 * a relatively inexpensive operation since we expect these trees to
	 * have a small number of nodes.
	 */
	condense_tree = range_tree_create(NULL, NULL, &msp->ms_lock);
	range_tree_add(condense_tree, msp->ms_start, msp->ms_size);

	/*
	 * Remove what's been freed in this txg from the condense_tree.
	 * Since we're in sync_pass 1, we know that all the frees from
	 * this txg are in the freetree.
	 */
	range_tree_walk(freetree, range_tree_remove, condense_tree);

	for (t = 0; t < TXG_DEFER_SIZE; t++) {
		range_tree_walk(msp->ms_defertree[t],
		    range_tree_remove, condense_tree);
	}

	for (t = 1; t < TXG_CONCURRENT_STATES; t++) {
		range_tree_walk(msp->ms_alloctree[(txg + t) & TXG_MASK],
		    range_tree_remove, condense_tree);
	}

	/*
	 * We're about to drop the metaslab's lock thus allowing
	 * other consumers to change it's content. Set the
	 * metaslab's ms_condensing flag to ensure that
	 * allocations on this metaslab do not occur while we're
	 * in the middle of committing it to disk. This is only critical
	 * for the ms_tree as all other range trees use per txg
	 * views of their content.
	 */
	msp->ms_condensing = B_TRUE;

	mutex_exit(&msp->ms_lock);
	space_map_truncate(sm, tx);
	mutex_enter(&msp->ms_lock);

	/*
	 * While we would ideally like to create a space_map representation
	 * that consists only of allocation records, doing so can be
	 * prohibitively expensive because the in-core free tree can be
	 * large, and therefore computationally expensive to subtract
	 * from the condense_tree. Instead we sync out two trees, a cheap
	 * allocation only tree followed by the in-core free tree. While not
	 * optimal, this is typically close to optimal, and much cheaper to
	 * compute.
	 */
	space_map_write(sm, condense_tree, SM_ALLOC, tx);
	range_tree_vacate(condense_tree, NULL, NULL);
	range_tree_destroy(condense_tree);

	space_map_write(sm, msp->ms_tree, SM_FREE, tx);
	msp->ms_condensing = B_FALSE;
}

/*
 * Write a metaslab to disk in the context of the specified transaction group.
 */
void
metaslab_sync(metaslab_t *msp, uint64_t txg)
{
	metaslab_group_t *mg = msp->ms_group;
	vdev_t *vd = mg->mg_vd;
	spa_t *spa = vd->vdev_spa;
	objset_t *mos = spa_meta_objset(spa);
	range_tree_t *alloctree = msp->ms_alloctree[txg & TXG_MASK];
	range_tree_t **freetree = &msp->ms_freetree[txg & TXG_MASK];
	range_tree_t **freed_tree =
	    &msp->ms_freetree[TXG_CLEAN(txg) & TXG_MASK];
	dmu_tx_t *tx;
	uint64_t object = space_map_object(msp->ms_sm);

	ASSERT(!vd->vdev_ishole);

	/*
	 * This metaslab has just been added so there's no work to do now.
	 */
	if (*freetree == NULL) {
		ASSERT3P(alloctree, ==, NULL);
		return;
	}

	ASSERT3P(alloctree, !=, NULL);
	ASSERT3P(*freetree, !=, NULL);
	ASSERT3P(*freed_tree, !=, NULL);

	/*
	 * Normally, we don't want to process a metaslab if there
	 * are no allocations or frees to perform. However, if the metaslab
	 * is being forced to condense we need to let it through.
	 */
	if (range_tree_space(alloctree) == 0 &&
	    range_tree_space(*freetree) == 0 &&
	    !msp->ms_condense_wanted)
		return;

	/*
	 * The only state that can actually be changing concurrently with
	 * metaslab_sync() is the metaslab's ms_tree.  No other thread can
	 * be modifying this txg's alloctree, freetree, freed_tree, or
	 * space_map_phys_t. Therefore, we only hold ms_lock to satify
	 * space_map ASSERTs. We drop it whenever we call into the DMU,
	 * because the DMU can call down to us (e.g. via zio_free()) at
	 * any time.
	 */

	tx = dmu_tx_create_assigned(spa_get_dsl(spa), txg);

	if (msp->ms_sm == NULL) {
		uint64_t new_object;

		new_object = space_map_alloc(mos, tx);
		VERIFY3U(new_object, !=, 0);

		VERIFY0(space_map_open(&msp->ms_sm, mos, new_object,
		    msp->ms_start, msp->ms_size, vd->vdev_ashift,
		    &msp->ms_lock));
		ASSERT(msp->ms_sm != NULL);
	}

	mutex_enter(&msp->ms_lock);

	/*
	 * Note: metaslab_condense() clears the space_map's histogram.
	 * Therefore we muse verify and remove this histogram before
	 * condensing.
	 */
	metaslab_group_histogram_verify(mg);
	metaslab_class_histogram_verify(mg->mg_class);
	metaslab_group_histogram_remove(mg, msp);

	if (msp->ms_loaded && spa_sync_pass(spa) == 1 &&
	    metaslab_should_condense(msp)) {
		metaslab_condense(msp, txg, tx);
	} else {
		space_map_write(msp->ms_sm, alloctree, SM_ALLOC, tx);
		space_map_write(msp->ms_sm, *freetree, SM_FREE, tx);
	}

	if (msp->ms_loaded) {
		/*
		 * When the space map is loaded, we have an accruate
		 * histogram in the range tree. This gives us an opportunity
		 * to bring the space map's histogram up-to-date so we clear
		 * it first before updating it.
		 */
		space_map_histogram_clear(msp->ms_sm);
		space_map_histogram_add(msp->ms_sm, msp->ms_tree, tx);
	} else {
		/*
		 * Since the space map is not loaded we simply update the
		 * existing histogram with what was freed in this txg. This
		 * means that the on-disk histogram may not have an accurate
		 * view of the free space but it's close enough to allow
		 * us to make allocation decisions.
		 */
		space_map_histogram_add(msp->ms_sm, *freetree, tx);
	}
	metaslab_group_histogram_add(mg, msp);
	metaslab_group_histogram_verify(mg);
	metaslab_class_histogram_verify(mg->mg_class);

	/*
	 * For sync pass 1, we avoid traversing this txg's free range tree
	 * and instead will just swap the pointers for freetree and
	 * freed_tree. We can safely do this since the freed_tree is
	 * guaranteed to be empty on the initial pass.
	 */
	if (spa_sync_pass(spa) == 1) {
		range_tree_swap(freetree, freed_tree);
	} else {
		range_tree_vacate(*freetree, range_tree_add, *freed_tree);
	}
	range_tree_vacate(alloctree, NULL, NULL);

	ASSERT0(range_tree_space(msp->ms_alloctree[txg & TXG_MASK]));
	ASSERT0(range_tree_space(msp->ms_freetree[txg & TXG_MASK]));

	mutex_exit(&msp->ms_lock);

	if (object != space_map_object(msp->ms_sm)) {
		object = space_map_object(msp->ms_sm);
		dmu_write(mos, vd->vdev_ms_array, sizeof (uint64_t) *
		    msp->ms_id, sizeof (uint64_t), &object, tx);
	}
	dmu_tx_commit(tx);
}

/*
 * Called after a transaction group has completely synced to mark
 * all of the metaslab's free space as usable.
 */
void
metaslab_sync_done(metaslab_t *msp, uint64_t txg)
{
	metaslab_group_t *mg = msp->ms_group;
	vdev_t *vd = mg->mg_vd;
	range_tree_t **freed_tree;
	range_tree_t **defer_tree;
	int64_t alloc_delta, defer_delta;
	int t;

	ASSERT(!vd->vdev_ishole);

	mutex_enter(&msp->ms_lock);

	/*
	 * If this metaslab is just becoming available, initialize its
	 * alloctrees, freetrees, and defertree and add its capacity to
	 * the vdev.
	 */
	if (msp->ms_freetree[TXG_CLEAN(txg) & TXG_MASK] == NULL) {
		for (t = 0; t < TXG_SIZE; t++) {
			ASSERT(msp->ms_alloctree[t] == NULL);
			ASSERT(msp->ms_freetree[t] == NULL);

			msp->ms_alloctree[t] = range_tree_create(NULL, msp,
			    &msp->ms_lock);
			msp->ms_freetree[t] = range_tree_create(NULL, msp,
			    &msp->ms_lock);
		}

		for (t = 0; t < TXG_DEFER_SIZE; t++) {
			ASSERT(msp->ms_defertree[t] == NULL);

			msp->ms_defertree[t] = range_tree_create(NULL, msp,
			    &msp->ms_lock);
		}

		vdev_space_update(vd, 0, 0, msp->ms_size);
	}

	freed_tree = &msp->ms_freetree[TXG_CLEAN(txg) & TXG_MASK];
	defer_tree = &msp->ms_defertree[txg % TXG_DEFER_SIZE];

	alloc_delta = space_map_alloc_delta(msp->ms_sm);
	defer_delta = range_tree_space(*freed_tree) -
	    range_tree_space(*defer_tree);

	vdev_space_update(vd, alloc_delta + defer_delta, defer_delta, 0);

	ASSERT0(range_tree_space(msp->ms_alloctree[txg & TXG_MASK]));
	ASSERT0(range_tree_space(msp->ms_freetree[txg & TXG_MASK]));

	/*
	 * If there's a metaslab_load() in progress, wait for it to complete
	 * so that we have a consistent view of the in-core space map.
	 */
	metaslab_load_wait(msp);

	/*
	 * Move the frees from the defer_tree back to the free
	 * range tree (if it's loaded). Swap the freed_tree and the
	 * defer_tree -- this is safe to do because we've just emptied out
	 * the defer_tree.
	 */
	range_tree_vacate(*defer_tree,
	    msp->ms_loaded ? range_tree_add : NULL, msp->ms_tree);
	range_tree_swap(freed_tree, defer_tree);

	space_map_update(msp->ms_sm);

	msp->ms_deferspace += defer_delta;
	ASSERT3S(msp->ms_deferspace, >=, 0);
	ASSERT3S(msp->ms_deferspace, <=, msp->ms_size);
	if (msp->ms_deferspace != 0) {
		/*
		 * Keep syncing this metaslab until all deferred frees
		 * are back in circulation.
		 */
		vdev_dirty(vd, VDD_METASLAB, msp, txg + 1);
	}

	if (msp->ms_loaded && msp->ms_access_txg < txg) {
		for (t = 1; t < TXG_CONCURRENT_STATES; t++) {
			VERIFY0(range_tree_space(
			    msp->ms_alloctree[(txg + t) & TXG_MASK]));
		}

		if (!metaslab_debug_unload)
			metaslab_unload(msp);
	}

	metaslab_group_sort(mg, msp, metaslab_weight(msp));
	mutex_exit(&msp->ms_lock);
}

void
metaslab_sync_reassess(metaslab_group_t *mg)
{
	metaslab_group_alloc_update(mg);
	mg->mg_fragmentation = metaslab_group_fragmentation(mg);

	/*
	 * Preload the next potential metaslabs
	 */
	metaslab_group_preload(mg);
}

static uint64_t
metaslab_distance(metaslab_t *msp, dva_t *dva)
{
	uint64_t ms_shift = msp->ms_group->mg_vd->vdev_ms_shift;
	uint64_t offset = DVA_GET_OFFSET(dva) >> ms_shift;
	uint64_t start = msp->ms_id;

	if (msp->ms_group->mg_vd->vdev_id != DVA_GET_VDEV(dva))
		return (1ULL << 63);

	if (offset < start)
		return ((start - offset) << ms_shift);
	if (offset > start)
		return ((offset - start) << ms_shift);
	return (0);
}

/*
 * ==========================================================================
 * Metaslab block operations
 * ==========================================================================
 */

static void
metaslab_group_alloc_increment(spa_t *spa, uint64_t vdev, void *tag, int flags)
{
	metaslab_group_t *mg;

	if (!(flags & METASLAB_ASYNC_ALLOC) ||
	    flags & METASLAB_DONT_THROTTLE)
		return;

	mg = vdev_lookup_top(spa, vdev)->vdev_mg;
	if (!mg->mg_class->mc_alloc_throttle_enabled)
		return;

	(void) refcount_add(&mg->mg_alloc_queue_depth, tag);
}

void
metaslab_group_alloc_decrement(spa_t *spa, uint64_t vdev, void *tag, int flags)
{
	metaslab_group_t *mg;

	if (!(flags & METASLAB_ASYNC_ALLOC) ||
	    flags & METASLAB_DONT_THROTTLE)
		return;

	mg = vdev_lookup_top(spa, vdev)->vdev_mg;
	if (!mg->mg_class->mc_alloc_throttle_enabled)
		return;

	(void) refcount_remove(&mg->mg_alloc_queue_depth, tag);
}

void
metaslab_group_alloc_verify(spa_t *spa, const blkptr_t *bp, void *tag)
{
#ifdef ZFS_DEBUG
	const dva_t *dva = bp->blk_dva;
	int ndvas = BP_GET_NDVAS(bp);
	int d;

	for (d = 0; d < ndvas; d++) {
		uint64_t vdev = DVA_GET_VDEV(&dva[d]);
		metaslab_group_t *mg = vdev_lookup_top(spa, vdev)->vdev_mg;
		VERIFY(refcount_not_held(&mg->mg_alloc_queue_depth, tag));
	}
#endif
}

static uint64_t
metaslab_group_alloc(metaslab_group_t *mg, uint64_t asize,
    uint64_t txg, uint64_t min_distance, dva_t *dva, int d)
{
	spa_t *spa = mg->mg_vd->vdev_spa;
	metaslab_t *msp = NULL;
	uint64_t offset = -1ULL;
	avl_tree_t *t = &mg->mg_metaslab_tree;
	uint64_t activation_weight;
	uint64_t target_distance;
	int i;

	activation_weight = METASLAB_WEIGHT_PRIMARY;
	for (i = 0; i < d; i++) {
		if (DVA_GET_VDEV(&dva[i]) == mg->mg_vd->vdev_id) {
			activation_weight = METASLAB_WEIGHT_SECONDARY;
			break;
		}
	}

	for (;;) {
		boolean_t was_active;

		mutex_enter(&mg->mg_lock);
		for (msp = avl_first(t); msp; msp = AVL_NEXT(t, msp)) {
			if (msp->ms_weight < asize) {
				spa_dbgmsg(spa, "%s: failed to meet weight "
				    "requirement: vdev %llu, txg %llu, mg %p, "
				    "msp %p, asize %llu, "
				    "weight %llu", spa_name(spa),
				    mg->mg_vd->vdev_id, txg,
				    mg, msp, asize, msp->ms_weight);
				mutex_exit(&mg->mg_lock);
				return (-1ULL);
			}

			/*
			 * If the selected metaslab is condensing, skip it.
			 */
			if (msp->ms_condensing)
				continue;

			was_active = msp->ms_weight & METASLAB_ACTIVE_MASK;
			if (activation_weight == METASLAB_WEIGHT_PRIMARY)
				break;

			target_distance = min_distance +
			    (space_map_allocated(msp->ms_sm) != 0 ? 0 :
			    min_distance >> 1);

			for (i = 0; i < d; i++)
				if (metaslab_distance(msp, &dva[i]) <
				    target_distance)
					break;
			if (i == d)
				break;
		}
		mutex_exit(&mg->mg_lock);
		if (msp == NULL)
			return (-1ULL);

		mutex_enter(&msp->ms_lock);

		/*
		 * Ensure that the metaslab we have selected is still
		 * capable of handling our request. It's possible that
		 * another thread may have changed the weight while we
		 * were blocked on the metaslab lock.
		 */
		if (msp->ms_weight < asize || (was_active &&
		    !(msp->ms_weight & METASLAB_ACTIVE_MASK) &&
		    activation_weight == METASLAB_WEIGHT_PRIMARY)) {
			mutex_exit(&msp->ms_lock);
			continue;
		}

		if ((msp->ms_weight & METASLAB_WEIGHT_SECONDARY) &&
		    activation_weight == METASLAB_WEIGHT_PRIMARY) {
			metaslab_passivate(msp,
			    msp->ms_weight & ~METASLAB_ACTIVE_MASK);
			mutex_exit(&msp->ms_lock);
			continue;
		}

		if (metaslab_activate(msp, activation_weight) != 0) {
			mutex_exit(&msp->ms_lock);
			continue;
		}

		/*
		 * If this metaslab is currently condensing then pick again as
		 * we can't manipulate this metaslab until it's committed
		 * to disk.
		 */
		if (msp->ms_condensing) {
			mutex_exit(&msp->ms_lock);
			continue;
		}

		if ((offset = metaslab_block_alloc(msp, asize)) != -1ULL)
			break;

		metaslab_passivate(msp, metaslab_block_maxsize(msp));
		mutex_exit(&msp->ms_lock);
	}

	if (range_tree_space(msp->ms_alloctree[txg & TXG_MASK]) == 0)
		vdev_dirty(mg->mg_vd, VDD_METASLAB, msp, txg);

	range_tree_add(msp->ms_alloctree[txg & TXG_MASK], offset, asize);
	msp->ms_access_txg = txg + metaslab_unload_delay;

	mutex_exit(&msp->ms_lock);
	return (offset);
}

/*
 * Allocate a block for the specified i/o.
 */
static int
metaslab_alloc_dva(spa_t *spa, metaslab_class_t *mc, uint64_t psize,
    dva_t *dva, int d, dva_t *hintdva, uint64_t txg, int flags)
{
	metaslab_group_t *mg, *fast_mg, *rotor;
	vdev_t *vd;
	int dshift = 3;
	int all_zero;
	int zio_lock = B_FALSE;
	boolean_t allocatable;
	uint64_t asize;
	uint64_t distance;

	ASSERT(!DVA_IS_VALID(&dva[d]));

	/*
	 * For testing, make some blocks above a certain size be gang blocks.
	 */
	if (psize >= metaslab_gang_bang && (ddi_get_lbolt() & 3) == 0)
		return (SET_ERROR(ENOSPC));

	/*
	 * Start at the rotor and loop through all mgs until we find something.
	 * Note that there's no locking on mc_rotor or mc_aliquot because
	 * nothing actually breaks if we miss a few updates -- we just won't
	 * allocate quite as evenly.  It all balances out over time.
	 *
	 * If we are doing ditto or log blocks, try to spread them across
	 * consecutive vdevs.  If we're forced to reuse a vdev before we've
	 * allocated all of our ditto blocks, then try and spread them out on
	 * that vdev as much as possible.  If it turns out to not be possible,
	 * gradually lower our standards until anything becomes acceptable.
	 * Also, allocating on consecutive vdevs (as opposed to random vdevs)
	 * gives us hope of containing our fault domains to something we're
	 * able to reason about.  Otherwise, any two top-level vdev failures
	 * will guarantee the loss of data.  With consecutive allocation,
	 * only two adjacent top-level vdev failures will result in data loss.
	 *
	 * If we are doing gang blocks (hintdva is non-NULL), try to keep
	 * ourselves on the same vdev as our gang block header.  That
	 * way, we can hope for locality in vdev_cache, plus it makes our
	 * fault domains something tractable.
	 */
	if (hintdva) {
		vd = vdev_lookup_top(spa, DVA_GET_VDEV(&hintdva[d]));

		/*
		 * It's possible the vdev we're using as the hint no
		 * longer exists (i.e. removed). Consult the rotor when
		 * all else fails.
		 */
		if (vd != NULL) {
			mg = vd->vdev_mg;

			if (flags & METASLAB_HINTBP_AVOID &&
			    mg->mg_next != NULL)
				mg = mg->mg_next;
		} else {
			mg = mc->mc_rotor;
		}
	} else if (d != 0) {
		vd = vdev_lookup_top(spa, DVA_GET_VDEV(&dva[d - 1]));
		mg = vd->vdev_mg->mg_next;
	} else if (flags & METASLAB_FASTWRITE) {
		mg = fast_mg = mc->mc_rotor;

		do {
			if (fast_mg->mg_vd->vdev_pending_fastwrite <
			    mg->mg_vd->vdev_pending_fastwrite)
				mg = fast_mg;
		} while ((fast_mg = fast_mg->mg_next) != mc->mc_rotor);

	} else {
		mg = mc->mc_rotor;
	}

	/*
	 * If the hint put us into the wrong metaslab class, or into a
	 * metaslab group that has been passivated, just follow the rotor.
	 */
	if (mg->mg_class != mc || mg->mg_activation_count <= 0)
		mg = mc->mc_rotor;

	rotor = mg;
top:
	all_zero = B_TRUE;
	do {
		uint64_t offset;

		ASSERT(mg->mg_activation_count == 1);
		vd = mg->mg_vd;

		/*
		 * Don't allocate from faulted devices.
		 */
		if (zio_lock) {
			spa_config_enter(spa, SCL_ZIO, FTAG, RW_READER);
			allocatable = vdev_allocatable(vd);
			spa_config_exit(spa, SCL_ZIO, FTAG);
		} else {
			allocatable = vdev_allocatable(vd);
		}

		/*
		 * Determine if the selected metaslab group is eligible
		 * for allocations. If we're ganging then don't allow
		 * this metaslab group to skip allocations since that would
		 * inadvertently return ENOSPC and suspend the pool
		 * even though space is still available.
		 */
		if (allocatable && !GANG_ALLOCATION(flags) && !zio_lock) {
			allocatable = metaslab_group_allocatable(mg, rotor,
			    psize);
		}

		if (!allocatable)
			goto next;

		ASSERT(mg->mg_initialized);

		/*
		 * Avoid writing single-copy data to a failing vdev.
		 */
		if ((vd->vdev_stat.vs_write_errors > 0 ||
		    vd->vdev_state < VDEV_STATE_HEALTHY) &&
		    d == 0 && dshift == 3 && vd->vdev_children == 0) {
			all_zero = B_FALSE;
			goto next;
		}

		ASSERT(mg->mg_class == mc);

		distance = vd->vdev_asize >> dshift;
		if (distance <= (1ULL << vd->vdev_ms_shift))
			distance = 0;
		else
			all_zero = B_FALSE;

		asize = vdev_psize_to_asize(vd, psize);
		ASSERT(P2PHASE(asize, 1ULL << vd->vdev_ashift) == 0);

		offset = metaslab_group_alloc(mg, asize, txg, distance, dva, d);

		mutex_enter(&mg->mg_lock);
		if (offset == -1ULL) {
			mg->mg_failed_allocations++;
			if (asize == SPA_GANGBLOCKSIZE) {
				/*
				 * This metaslab group was unable to allocate
				 * the minimum gang block size so it must be
				 * out of space. We must notify the allocation
				 * throttle to start skipping allocation
				 * attempts to this metaslab group until more
				 * space becomes available.
				 *
				 * Note: this failure cannot be caused by the
				 * allocation throttle since the allocation
				 * throttle is only responsible for skipping
				 * devices and not failing block allocations.
				 */
				mg->mg_no_free_space = B_TRUE;
			}
		}
		mg->mg_allocations++;
		mutex_exit(&mg->mg_lock);

		if (offset != -1ULL) {
			/*
			 * If we've just selected this metaslab group,
			 * figure out whether the corresponding vdev is
			 * over- or under-used relative to the pool,
			 * and set an allocation bias to even it out.
			 *
			 * Bias is also used to compensate for unequally
			 * sized vdevs so that space is allocated fairly.
			 */
			if (mc->mc_aliquot == 0 && metaslab_bias_enabled) {
				vdev_stat_t *vs = &vd->vdev_stat;
				int64_t vs_free = vs->vs_space - vs->vs_alloc;
				int64_t mc_free = mc->mc_space - mc->mc_alloc;
				int64_t ratio;

				/*
				 * Calculate how much more or less we should
				 * try to allocate from this device during
				 * this iteration around the rotor.
				 *
				 * This basically introduces a zero-centered
				 * bias towards the devices with the most
				 * free space, while compensating for vdev
				 * size differences.
				 *
				 * Examples:
				 *  vdev V1 = 16M/128M
				 *  vdev V2 = 16M/128M
				 *  ratio(V1) = 100% ratio(V2) = 100%
				 *
				 *  vdev V1 = 16M/128M
				 *  vdev V2 = 64M/128M
				 *  ratio(V1) = 127% ratio(V2) =  72%
				 *
				 *  vdev V1 = 16M/128M
				 *  vdev V2 = 64M/512M
				 *  ratio(V1) =  40% ratio(V2) = 160%
				 */
				ratio = (vs_free * mc->mc_alloc_groups * 100) /
				    (mc_free + 1);
				mg->mg_bias = ((ratio - 100) *
				    (int64_t)mg->mg_aliquot) / 100;
			} else if (!metaslab_bias_enabled) {
				mg->mg_bias = 0;
			}

			if ((flags & METASLAB_FASTWRITE) ||
			    atomic_add_64_nv(&mc->mc_aliquot, asize) >=
			    mg->mg_aliquot + mg->mg_bias) {
				mc->mc_rotor = mg->mg_next;
				mc->mc_aliquot = 0;
			}

			DVA_SET_VDEV(&dva[d], vd->vdev_id);
			DVA_SET_OFFSET(&dva[d], offset);
			DVA_SET_GANG(&dva[d],
			    ((flags & METASLAB_GANG_HEADER) ? 1 : 0));
			DVA_SET_ASIZE(&dva[d], asize);

			if (flags & METASLAB_FASTWRITE) {
				atomic_add_64(&vd->vdev_pending_fastwrite,
				    psize);
			}

			return (0);
		}
next:
		mc->mc_rotor = mg->mg_next;
		mc->mc_aliquot = 0;
	} while ((mg = mg->mg_next) != rotor);

	if (!all_zero) {
		dshift++;
		ASSERT(dshift < 64);
		goto top;
	}

	if (!allocatable && !zio_lock) {
		dshift = 3;
		zio_lock = B_TRUE;
		goto top;
	}

	bzero(&dva[d], sizeof (dva_t));

	return (SET_ERROR(ENOSPC));
}

/*
 * Free the block represented by DVA in the context of the specified
 * transaction group.
 */
static void
metaslab_free_dva(spa_t *spa, const dva_t *dva, uint64_t txg, boolean_t now)
{
	uint64_t vdev = DVA_GET_VDEV(dva);
	uint64_t offset = DVA_GET_OFFSET(dva);
	uint64_t size = DVA_GET_ASIZE(dva);
	vdev_t *vd;
	metaslab_t *msp;

	if (txg > spa_freeze_txg(spa))
		return;

	if ((vd = vdev_lookup_top(spa, vdev)) == NULL || !DVA_IS_VALID(dva) ||
	    (offset >> vd->vdev_ms_shift) >= vd->vdev_ms_count) {
		zfs_panic_recover("metaslab_free_dva(): bad DVA %llu:%llu:%llu",
		    (u_longlong_t)vdev, (u_longlong_t)offset,
		    (u_longlong_t)size);
		return;
	}

	msp = vd->vdev_ms[offset >> vd->vdev_ms_shift];

	if (DVA_GET_GANG(dva))
		size = vdev_psize_to_asize(vd, SPA_GANGBLOCKSIZE);

	mutex_enter(&msp->ms_lock);

	if (now) {
		range_tree_remove(msp->ms_alloctree[txg & TXG_MASK],
		    offset, size);

		VERIFY(!msp->ms_condensing);
		VERIFY3U(offset, >=, msp->ms_start);
		VERIFY3U(offset + size, <=, msp->ms_start + msp->ms_size);
		VERIFY3U(range_tree_space(msp->ms_tree) + size, <=,
		    msp->ms_size);
		VERIFY0(P2PHASE(offset, 1ULL << vd->vdev_ashift));
		VERIFY0(P2PHASE(size, 1ULL << vd->vdev_ashift));
		range_tree_add(msp->ms_tree, offset, size);
	} else {
		if (range_tree_space(msp->ms_freetree[txg & TXG_MASK]) == 0)
			vdev_dirty(vd, VDD_METASLAB, msp, txg);
		range_tree_add(msp->ms_freetree[txg & TXG_MASK],
		    offset, size);
	}

	mutex_exit(&msp->ms_lock);
}

/*
 * Intent log support: upon opening the pool after a crash, notify the SPA
 * of blocks that the intent log has allocated for immediate write, but
 * which are still considered free by the SPA because the last transaction
 * group didn't commit yet.
 */
static int
metaslab_claim_dva(spa_t *spa, const dva_t *dva, uint64_t txg)
{
	uint64_t vdev = DVA_GET_VDEV(dva);
	uint64_t offset = DVA_GET_OFFSET(dva);
	uint64_t size = DVA_GET_ASIZE(dva);
	vdev_t *vd;
	metaslab_t *msp;
	int error = 0;

	ASSERT(DVA_IS_VALID(dva));

	if ((vd = vdev_lookup_top(spa, vdev)) == NULL ||
	    (offset >> vd->vdev_ms_shift) >= vd->vdev_ms_count)
		return (SET_ERROR(ENXIO));

	msp = vd->vdev_ms[offset >> vd->vdev_ms_shift];

	if (DVA_GET_GANG(dva))
		size = vdev_psize_to_asize(vd, SPA_GANGBLOCKSIZE);

	mutex_enter(&msp->ms_lock);

	if ((txg != 0 && spa_writeable(spa)) || !msp->ms_loaded)
		error = metaslab_activate(msp, METASLAB_WEIGHT_SECONDARY);

	if (error == 0 && !range_tree_contains(msp->ms_tree, offset, size))
		error = SET_ERROR(ENOENT);

	if (error || txg == 0) {	/* txg == 0 indicates dry run */
		mutex_exit(&msp->ms_lock);
		return (error);
	}

	VERIFY(!msp->ms_condensing);
	VERIFY0(P2PHASE(offset, 1ULL << vd->vdev_ashift));
	VERIFY0(P2PHASE(size, 1ULL << vd->vdev_ashift));
	VERIFY3U(range_tree_space(msp->ms_tree) - size, <=, msp->ms_size);
	range_tree_remove(msp->ms_tree, offset, size);

	if (spa_writeable(spa)) {	/* don't dirty if we're zdb(1M) */
		if (range_tree_space(msp->ms_alloctree[txg & TXG_MASK]) == 0)
			vdev_dirty(vd, VDD_METASLAB, msp, txg);
		range_tree_add(msp->ms_alloctree[txg & TXG_MASK], offset, size);
	}

	mutex_exit(&msp->ms_lock);

	return (0);
}

/*
 * Reserve some allocation slots. The reservation system must be called
 * before we call into the allocator. If there aren't any available slots
 * then the I/O will be throttled until an I/O completes and its slots are
 * freed up. The function returns true if it was successful in placing
 * the reservation.
 */
boolean_t
metaslab_class_throttle_reserve(metaslab_class_t *mc, int slots, zio_t *zio,
    int flags)
{
	uint64_t available_slots = 0;
	uint64_t reserved_slots;
	boolean_t slot_reserved = B_FALSE;

	ASSERT(mc->mc_alloc_throttle_enabled);
	mutex_enter(&mc->mc_lock);

	reserved_slots = refcount_count(&mc->mc_alloc_slots);
	if (reserved_slots < mc->mc_alloc_max_slots)
		available_slots = mc->mc_alloc_max_slots - reserved_slots;

	if (slots <= available_slots || GANG_ALLOCATION(flags)) {
		int d;

		/*
		 * We reserve the slots individually so that we can unreserve
		 * them individually when an I/O completes.
		 */
		for (d = 0; d < slots; d++) {
			reserved_slots = refcount_add(&mc->mc_alloc_slots, zio);
		}
		zio->io_flags |= ZIO_FLAG_IO_ALLOCATING;
		slot_reserved = B_TRUE;
	}

	mutex_exit(&mc->mc_lock);
	return (slot_reserved);
}

void
metaslab_class_throttle_unreserve(metaslab_class_t *mc, int slots, zio_t *zio)
{
	int d;

	ASSERT(mc->mc_alloc_throttle_enabled);
	mutex_enter(&mc->mc_lock);
	for (d = 0; d < slots; d++) {
		(void) refcount_remove(&mc->mc_alloc_slots, zio);
	}
	mutex_exit(&mc->mc_lock);
}

int
metaslab_alloc(spa_t *spa, metaslab_class_t *mc, uint64_t psize, blkptr_t *bp,
    int ndvas, uint64_t txg, blkptr_t *hintbp, int flags, zio_t *zio)
{
	dva_t *dva = bp->blk_dva;
	dva_t *hintdva = hintbp->blk_dva;
	int d, error = 0;

	ASSERT(bp->blk_birth == 0);
	ASSERT(BP_PHYSICAL_BIRTH(bp) == 0);

	spa_config_enter(spa, SCL_ALLOC, FTAG, RW_READER);

	if (mc->mc_rotor == NULL) {	/* no vdevs in this class */
		spa_config_exit(spa, SCL_ALLOC, FTAG);
		return (SET_ERROR(ENOSPC));
	}

	ASSERT(ndvas > 0 && ndvas <= spa_max_replication(spa));
	ASSERT(BP_GET_NDVAS(bp) == 0);
	ASSERT(hintbp == NULL || ndvas <= BP_GET_NDVAS(hintbp));

	for (d = 0; d < ndvas; d++) {
		error = metaslab_alloc_dva(spa, mc, psize, dva, d, hintdva,
		    txg, flags);
		if (error != 0) {
			for (d--; d >= 0; d--) {
				metaslab_free_dva(spa, &dva[d], txg, B_TRUE);
				metaslab_group_alloc_decrement(spa,
				    DVA_GET_VDEV(&dva[d]), zio, flags);
				bzero(&dva[d], sizeof (dva_t));
			}
			spa_config_exit(spa, SCL_ALLOC, FTAG);
			return (error);
		} else {
			/*
			 * Update the metaslab group's queue depth
			 * based on the newly allocated dva.
			 */
			metaslab_group_alloc_increment(spa,
			    DVA_GET_VDEV(&dva[d]), zio, flags);
		}

	}
	ASSERT(error == 0);
	ASSERT(BP_GET_NDVAS(bp) == ndvas);

	spa_config_exit(spa, SCL_ALLOC, FTAG);

	BP_SET_BIRTH(bp, txg, 0);

	return (0);
}

void
metaslab_free(spa_t *spa, const blkptr_t *bp, uint64_t txg, boolean_t now)
{
	const dva_t *dva = bp->blk_dva;
	int d, ndvas = BP_GET_NDVAS(bp);

	ASSERT(!BP_IS_HOLE(bp));
	ASSERT(!now || bp->blk_birth >= spa_syncing_txg(spa));

	spa_config_enter(spa, SCL_FREE, FTAG, RW_READER);

	for (d = 0; d < ndvas; d++)
		metaslab_free_dva(spa, &dva[d], txg, now);

	spa_config_exit(spa, SCL_FREE, FTAG);
}

int
metaslab_claim(spa_t *spa, const blkptr_t *bp, uint64_t txg)
{
	const dva_t *dva = bp->blk_dva;
	int ndvas = BP_GET_NDVAS(bp);
	int d, error = 0;

	ASSERT(!BP_IS_HOLE(bp));

	if (txg != 0) {
		/*
		 * First do a dry run to make sure all DVAs are claimable,
		 * so we don't have to unwind from partial failures below.
		 */
		if ((error = metaslab_claim(spa, bp, 0)) != 0)
			return (error);
	}

	spa_config_enter(spa, SCL_ALLOC, FTAG, RW_READER);

	for (d = 0; d < ndvas; d++)
		if ((error = metaslab_claim_dva(spa, &dva[d], txg)) != 0)
			break;

	spa_config_exit(spa, SCL_ALLOC, FTAG);

	ASSERT(error == 0 || txg == 0);

	return (error);
}

void
metaslab_fastwrite_mark(spa_t *spa, const blkptr_t *bp)
{
	const dva_t *dva = bp->blk_dva;
	int ndvas = BP_GET_NDVAS(bp);
	uint64_t psize = BP_GET_PSIZE(bp);
	int d;
	vdev_t *vd;

	ASSERT(!BP_IS_HOLE(bp));
	ASSERT(!BP_IS_EMBEDDED(bp));
	ASSERT(psize > 0);

	spa_config_enter(spa, SCL_VDEV, FTAG, RW_READER);

	for (d = 0; d < ndvas; d++) {
		if ((vd = vdev_lookup_top(spa, DVA_GET_VDEV(&dva[d]))) == NULL)
			continue;
		atomic_add_64(&vd->vdev_pending_fastwrite, psize);
	}

	spa_config_exit(spa, SCL_VDEV, FTAG);
}

void
metaslab_fastwrite_unmark(spa_t *spa, const blkptr_t *bp)
{
	const dva_t *dva = bp->blk_dva;
	int ndvas = BP_GET_NDVAS(bp);
	uint64_t psize = BP_GET_PSIZE(bp);
	int d;
	vdev_t *vd;

	ASSERT(!BP_IS_HOLE(bp));
	ASSERT(!BP_IS_EMBEDDED(bp));
	ASSERT(psize > 0);

	spa_config_enter(spa, SCL_VDEV, FTAG, RW_READER);

	for (d = 0; d < ndvas; d++) {
		if ((vd = vdev_lookup_top(spa, DVA_GET_VDEV(&dva[d]))) == NULL)
			continue;
		ASSERT3U(vd->vdev_pending_fastwrite, >=, psize);
		atomic_sub_64(&vd->vdev_pending_fastwrite, psize);
	}

	spa_config_exit(spa, SCL_VDEV, FTAG);
}

void
metaslab_check_free(spa_t *spa, const blkptr_t *bp)
{
	int i, j;

	if ((zfs_flags & ZFS_DEBUG_ZIO_FREE) == 0)
		return;

	spa_config_enter(spa, SCL_VDEV, FTAG, RW_READER);
	for (i = 0; i < BP_GET_NDVAS(bp); i++) {
		uint64_t vdev = DVA_GET_VDEV(&bp->blk_dva[i]);
		vdev_t *vd = vdev_lookup_top(spa, vdev);
		uint64_t offset = DVA_GET_OFFSET(&bp->blk_dva[i]);
		uint64_t size = DVA_GET_ASIZE(&bp->blk_dva[i]);
		metaslab_t *msp = vd->vdev_ms[offset >> vd->vdev_ms_shift];

		if (msp->ms_loaded)
			range_tree_verify(msp->ms_tree, offset, size);

		for (j = 0; j < TXG_SIZE; j++)
			range_tree_verify(msp->ms_freetree[j], offset, size);
		for (j = 0; j < TXG_DEFER_SIZE; j++)
			range_tree_verify(msp->ms_defertree[j], offset, size);
	}
	spa_config_exit(spa, SCL_VDEV, FTAG);
}

#if defined(_KERNEL) && defined(HAVE_SPL)
/* CSTYLED */
module_param(metaslab_aliquot, ulong, 0644);
MODULE_PARM_DESC(metaslab_aliquot,
	"allocation granularity (a.k.a. stripe size)");

module_param(metaslab_debug_load, int, 0644);
MODULE_PARM_DESC(metaslab_debug_load,
	"load all metaslabs when pool is first opened");

module_param(metaslab_debug_unload, int, 0644);
MODULE_PARM_DESC(metaslab_debug_unload,
	"prevent metaslabs from being unloaded");

module_param(metaslab_preload_enabled, int, 0644);
MODULE_PARM_DESC(metaslab_preload_enabled,
	"preload potential metaslabs during reassessment");

module_param(zfs_mg_noalloc_threshold, int, 0644);
MODULE_PARM_DESC(zfs_mg_noalloc_threshold,
	"percentage of free space for metaslab group to allow allocation");

module_param(zfs_mg_fragmentation_threshold, int, 0644);
MODULE_PARM_DESC(zfs_mg_fragmentation_threshold,
	"fragmentation for metaslab group to allow allocation");

module_param(zfs_metaslab_fragmentation_threshold, int, 0644);
MODULE_PARM_DESC(zfs_metaslab_fragmentation_threshold,
	"fragmentation for metaslab to allow allocation");

module_param(metaslab_fragmentation_factor_enabled, int, 0644);
MODULE_PARM_DESC(metaslab_fragmentation_factor_enabled,
	"use the fragmentation metric to prefer less fragmented metaslabs");

module_param(metaslab_lba_weighting_enabled, int, 0644);
MODULE_PARM_DESC(metaslab_lba_weighting_enabled,
	"prefer metaslabs with lower LBAs");

module_param(metaslab_bias_enabled, int, 0644);
MODULE_PARM_DESC(metaslab_bias_enabled,
	"enable metaslab group biasing");
#endif /* _KERNEL && HAVE_SPL */