summaryrefslogtreecommitdiffstats
path: root/module/zfs/zio.c
blob: e36a7893d711f731dd2943fb4bdb8aa6949a7f1d (plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
1001
1002
1003
1004
1005
1006
1007
1008
1009
1010
1011
1012
1013
1014
1015
1016
1017
1018
1019
1020
1021
1022
1023
1024
1025
1026
1027
1028
1029
1030
1031
1032
1033
1034
1035
1036
1037
1038
1039
1040
1041
1042
1043
1044
1045
1046
1047
1048
1049
1050
1051
1052
1053
1054
1055
1056
1057
1058
1059
1060
1061
1062
1063
1064
1065
1066
1067
1068
1069
1070
1071
1072
1073
1074
1075
1076
1077
1078
1079
1080
1081
1082
1083
1084
1085
1086
1087
1088
1089
1090
1091
1092
1093
1094
1095
1096
1097
1098
1099
1100
1101
1102
1103
1104
1105
1106
1107
1108
1109
1110
1111
1112
1113
1114
1115
1116
1117
1118
1119
1120
1121
1122
1123
1124
1125
1126
1127
1128
1129
1130
1131
1132
1133
1134
1135
1136
1137
1138
1139
1140
1141
1142
1143
1144
1145
1146
1147
1148
1149
1150
1151
1152
1153
1154
1155
1156
1157
1158
1159
1160
1161
1162
1163
1164
1165
1166
1167
1168
1169
1170
1171
1172
1173
1174
1175
1176
1177
1178
1179
1180
1181
1182
1183
1184
1185
1186
1187
1188
1189
1190
1191
1192
1193
1194
1195
1196
1197
1198
1199
1200
1201
1202
1203
1204
1205
1206
1207
1208
1209
1210
1211
1212
1213
1214
1215
1216
1217
1218
1219
1220
1221
1222
1223
1224
1225
1226
1227
1228
1229
1230
1231
1232
1233
1234
1235
1236
1237
1238
1239
1240
1241
1242
1243
1244
1245
1246
1247
1248
1249
1250
1251
1252
1253
1254
1255
1256
1257
1258
1259
1260
1261
1262
1263
1264
1265
1266
1267
1268
1269
1270
1271
1272
1273
1274
1275
1276
1277
1278
1279
1280
1281
1282
1283
1284
1285
1286
1287
1288
1289
1290
1291
1292
1293
1294
1295
1296
1297
1298
1299
1300
1301
1302
1303
1304
1305
1306
1307
1308
1309
1310
1311
1312
1313
1314
1315
1316
1317
1318
1319
1320
1321
1322
1323
1324
1325
1326
1327
1328
1329
1330
1331
1332
1333
1334
1335
1336
1337
1338
1339
1340
1341
1342
1343
1344
1345
1346
1347
1348
1349
1350
1351
1352
1353
1354
1355
1356
1357
1358
1359
1360
1361
1362
1363
1364
1365
1366
1367
1368
1369
1370
1371
1372
1373
1374
1375
1376
1377
1378
1379
1380
1381
1382
1383
1384
1385
1386
1387
1388
1389
1390
1391
1392
1393
1394
1395
1396
1397
1398
1399
1400
1401
1402
1403
1404
1405
1406
1407
1408
1409
1410
1411
1412
1413
1414
1415
1416
1417
1418
1419
1420
1421
1422
1423
1424
1425
1426
1427
1428
1429
1430
1431
1432
1433
1434
1435
1436
1437
1438
1439
1440
1441
1442
1443
1444
1445
1446
1447
1448
1449
1450
1451
1452
1453
1454
1455
1456
1457
1458
1459
1460
1461
1462
1463
1464
1465
1466
1467
1468
1469
1470
1471
1472
1473
1474
1475
1476
1477
1478
1479
1480
1481
1482
1483
1484
1485
1486
1487
1488
1489
1490
1491
1492
1493
1494
1495
1496
1497
1498
1499
1500
1501
1502
1503
1504
1505
1506
1507
1508
1509
1510
1511
1512
1513
1514
1515
1516
1517
1518
1519
1520
1521
1522
1523
1524
1525
1526
1527
1528
1529
1530
1531
1532
1533
1534
1535
1536
1537
1538
1539
1540
1541
1542
1543
1544
1545
1546
1547
1548
1549
1550
1551
1552
1553
1554
1555
1556
1557
1558
1559
1560
1561
1562
1563
1564
1565
1566
1567
1568
1569
1570
1571
1572
1573
1574
1575
1576
1577
1578
1579
1580
1581
1582
1583
1584
1585
1586
1587
1588
1589
1590
1591
1592
1593
1594
1595
1596
1597
1598
1599
1600
1601
1602
1603
1604
1605
1606
1607
1608
1609
1610
1611
1612
1613
1614
1615
1616
1617
1618
1619
1620
1621
1622
1623
1624
1625
1626
1627
1628
1629
1630
1631
1632
1633
1634
1635
1636
1637
1638
1639
1640
1641
1642
1643
1644
1645
1646
1647
1648
1649
1650
1651
1652
1653
1654
1655
1656
1657
1658
1659
1660
1661
1662
1663
1664
1665
1666
1667
1668
1669
1670
1671
1672
1673
1674
1675
1676
1677
1678
1679
1680
1681
1682
1683
1684
1685
1686
1687
1688
1689
1690
1691
1692
1693
1694
1695
1696
1697
1698
1699
1700
1701
1702
1703
1704
1705
1706
1707
1708
1709
1710
1711
1712
1713
1714
1715
1716
1717
1718
1719
1720
1721
1722
1723
1724
1725
1726
1727
1728
1729
1730
1731
1732
1733
1734
1735
1736
1737
1738
1739
1740
1741
1742
1743
1744
1745
1746
1747
1748
1749
1750
1751
1752
1753
1754
1755
1756
1757
1758
1759
1760
1761
1762
1763
1764
1765
1766
1767
1768
1769
1770
1771
1772
1773
1774
1775
1776
1777
1778
1779
1780
1781
1782
1783
1784
1785
1786
1787
1788
1789
1790
1791
1792
1793
1794
1795
1796
1797
1798
1799
1800
1801
1802
1803
1804
1805
1806
1807
1808
1809
1810
1811
1812
1813
1814
1815
1816
1817
1818
1819
1820
1821
1822
1823
1824
1825
1826
1827
1828
1829
1830
1831
1832
1833
1834
1835
1836
1837
1838
1839
1840
1841
1842
1843
1844
1845
1846
1847
1848
1849
1850
1851
1852
1853
1854
1855
1856
1857
1858
1859
1860
1861
1862
1863
1864
1865
1866
1867
1868
1869
1870
1871
1872
1873
1874
1875
1876
1877
1878
1879
1880
1881
1882
1883
1884
1885
1886
1887
1888
1889
1890
1891
1892
1893
1894
1895
1896
1897
1898
1899
1900
1901
1902
1903
1904
1905
1906
1907
1908
1909
1910
1911
1912
1913
1914
1915
1916
1917
1918
1919
1920
1921
1922
1923
1924
1925
1926
1927
1928
1929
1930
1931
1932
1933
1934
1935
1936
1937
1938
1939
1940
1941
1942
1943
1944
1945
1946
1947
1948
1949
1950
1951
1952
1953
1954
1955
1956
1957
1958
1959
1960
1961
1962
1963
1964
1965
1966
1967
1968
1969
1970
1971
1972
1973
1974
1975
1976
1977
1978
1979
1980
1981
1982
1983
1984
1985
1986
1987
1988
1989
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
2011
2012
2013
2014
2015
2016
2017
2018
2019
2020
2021
2022
2023
2024
2025
2026
2027
2028
2029
2030
2031
2032
2033
2034
2035
2036
2037
2038
2039
2040
2041
2042
2043
2044
2045
2046
2047
2048
2049
2050
2051
2052
2053
2054
2055
2056
2057
2058
2059
2060
2061
2062
2063
2064
2065
2066
2067
2068
2069
2070
2071
2072
2073
2074
2075
2076
2077
2078
2079
2080
2081
2082
2083
2084
2085
2086
2087
2088
2089
2090
2091
2092
2093
2094
2095
2096
2097
2098
2099
2100
2101
2102
2103
2104
2105
2106
2107
2108
2109
2110
2111
2112
2113
2114
2115
2116
2117
2118
2119
2120
2121
2122
2123
2124
2125
2126
2127
2128
2129
2130
2131
2132
2133
2134
2135
2136
2137
2138
2139
2140
2141
2142
2143
2144
2145
2146
2147
2148
2149
2150
2151
2152
2153
2154
2155
2156
2157
2158
2159
2160
2161
2162
2163
2164
2165
2166
2167
2168
2169
2170
2171
2172
2173
2174
2175
2176
2177
2178
2179
2180
2181
2182
2183
2184
2185
2186
2187
2188
2189
2190
2191
2192
2193
2194
2195
2196
2197
2198
2199
2200
2201
2202
2203
2204
2205
2206
2207
2208
2209
2210
2211
2212
2213
2214
2215
2216
2217
2218
2219
2220
2221
2222
2223
2224
2225
2226
2227
2228
2229
2230
2231
2232
2233
2234
2235
2236
2237
2238
2239
2240
2241
2242
2243
2244
2245
2246
2247
2248
2249
2250
2251
2252
2253
2254
2255
2256
2257
2258
2259
2260
2261
2262
2263
2264
2265
2266
2267
2268
2269
2270
2271
2272
2273
2274
2275
2276
2277
2278
2279
2280
2281
2282
2283
2284
2285
2286
2287
2288
2289
2290
2291
2292
2293
2294
2295
2296
2297
2298
2299
2300
2301
2302
2303
2304
2305
2306
2307
2308
2309
2310
2311
2312
2313
2314
2315
2316
2317
2318
2319
2320
2321
2322
2323
2324
2325
2326
2327
2328
2329
2330
2331
2332
2333
2334
2335
2336
2337
2338
2339
2340
2341
2342
2343
2344
2345
2346
2347
2348
2349
2350
2351
2352
2353
2354
2355
2356
2357
2358
2359
2360
2361
2362
2363
2364
2365
2366
2367
2368
2369
2370
2371
2372
2373
2374
2375
2376
2377
2378
2379
2380
2381
2382
2383
2384
2385
2386
2387
2388
2389
2390
2391
2392
2393
2394
2395
2396
2397
2398
2399
2400
2401
2402
2403
2404
2405
2406
2407
2408
2409
2410
2411
2412
2413
2414
2415
2416
2417
2418
2419
2420
2421
2422
2423
2424
2425
2426
2427
2428
2429
2430
2431
2432
2433
2434
2435
2436
2437
2438
2439
2440
2441
2442
2443
2444
2445
2446
2447
2448
2449
2450
2451
2452
2453
2454
2455
2456
2457
2458
2459
2460
2461
2462
2463
2464
2465
2466
2467
2468
2469
2470
2471
2472
2473
2474
2475
2476
2477
2478
2479
2480
2481
2482
2483
2484
2485
2486
2487
2488
2489
2490
2491
2492
2493
2494
2495
2496
2497
2498
2499
2500
2501
2502
2503
2504
2505
2506
2507
2508
2509
2510
2511
2512
2513
2514
2515
2516
2517
2518
2519
2520
2521
2522
2523
2524
2525
2526
2527
2528
2529
2530
2531
2532
2533
2534
2535
2536
2537
2538
2539
2540
2541
2542
2543
2544
2545
2546
2547
2548
2549
2550
2551
2552
2553
2554
2555
2556
2557
2558
2559
2560
2561
2562
2563
2564
2565
2566
2567
2568
2569
2570
2571
2572
2573
2574
2575
2576
2577
2578
2579
2580
2581
2582
2583
2584
2585
2586
2587
2588
2589
2590
2591
2592
2593
2594
2595
2596
2597
2598
2599
2600
2601
2602
2603
2604
2605
2606
2607
2608
2609
2610
2611
2612
2613
2614
2615
2616
2617
2618
2619
2620
2621
2622
2623
2624
2625
2626
2627
2628
2629
2630
2631
2632
2633
2634
2635
2636
2637
2638
2639
2640
2641
2642
2643
2644
2645
2646
2647
2648
2649
2650
2651
2652
2653
2654
2655
2656
2657
2658
2659
2660
2661
2662
2663
2664
2665
2666
2667
2668
2669
2670
2671
2672
2673
2674
2675
2676
2677
2678
2679
2680
2681
2682
2683
2684
2685
2686
2687
2688
2689
2690
2691
2692
2693
2694
2695
2696
2697
2698
2699
2700
2701
2702
2703
2704
2705
2706
2707
2708
2709
2710
2711
2712
2713
2714
2715
2716
2717
2718
2719
2720
2721
2722
2723
2724
2725
2726
2727
2728
2729
2730
2731
2732
2733
2734
2735
2736
2737
2738
2739
2740
2741
2742
2743
2744
2745
2746
2747
2748
2749
2750
2751
2752
2753
2754
2755
2756
2757
2758
2759
2760
2761
2762
2763
2764
2765
2766
2767
2768
2769
2770
2771
2772
2773
2774
2775
2776
2777
2778
2779
2780
2781
2782
2783
2784
2785
2786
2787
2788
2789
2790
2791
2792
2793
2794
2795
2796
2797
2798
2799
2800
2801
2802
2803
2804
2805
2806
2807
2808
2809
2810
2811
2812
2813
2814
2815
2816
2817
2818
2819
2820
2821
2822
2823
2824
2825
2826
2827
2828
2829
2830
2831
2832
2833
2834
2835
2836
2837
2838
2839
2840
2841
2842
2843
2844
2845
2846
2847
2848
2849
2850
2851
2852
2853
2854
2855
2856
2857
2858
2859
2860
2861
2862
2863
2864
2865
2866
2867
2868
2869
2870
2871
2872
2873
2874
2875
2876
2877
2878
2879
2880
2881
2882
2883
2884
2885
2886
2887
2888
2889
2890
2891
2892
2893
2894
2895
2896
2897
2898
2899
2900
2901
2902
2903
2904
2905
2906
2907
2908
2909
2910
2911
2912
2913
2914
2915
2916
2917
2918
2919
2920
2921
2922
2923
2924
2925
2926
2927
2928
2929
2930
2931
2932
2933
2934
2935
2936
2937
2938
2939
2940
2941
2942
2943
2944
2945
2946
2947
2948
2949
2950
2951
2952
2953
2954
2955
2956
2957
2958
2959
2960
2961
2962
2963
2964
2965
2966
2967
2968
2969
2970
2971
2972
2973
2974
2975
2976
2977
2978
2979
2980
2981
2982
2983
2984
2985
2986
2987
2988
2989
2990
2991
2992
2993
2994
2995
2996
2997
2998
2999
3000
3001
3002
3003
3004
3005
3006
3007
3008
3009
3010
3011
3012
3013
3014
3015
3016
3017
3018
3019
3020
3021
3022
3023
3024
3025
3026
3027
3028
3029
3030
3031
3032
3033
3034
3035
3036
3037
3038
3039
3040
3041
3042
3043
3044
3045
3046
3047
3048
3049
3050
3051
3052
3053
3054
3055
3056
3057
3058
3059
3060
3061
3062
3063
3064
3065
3066
3067
3068
3069
3070
3071
3072
3073
3074
3075
3076
3077
3078
3079
3080
3081
3082
3083
3084
3085
3086
3087
3088
3089
3090
3091
3092
3093
3094
3095
3096
3097
3098
3099
3100
3101
3102
3103
3104
3105
3106
3107
3108
3109
3110
3111
3112
3113
3114
3115
3116
3117
3118
3119
3120
3121
3122
3123
3124
3125
3126
3127
3128
3129
3130
3131
3132
3133
3134
3135
3136
3137
3138
3139
3140
3141
3142
3143
3144
3145
3146
3147
3148
3149
3150
3151
3152
3153
3154
3155
3156
3157
3158
3159
3160
3161
3162
3163
3164
3165
3166
3167
3168
3169
3170
3171
3172
3173
3174
3175
3176
3177
3178
3179
3180
3181
3182
3183
3184
3185
3186
3187
3188
3189
3190
3191
3192
3193
3194
3195
3196
3197
3198
3199
3200
3201
3202
3203
3204
3205
3206
3207
3208
3209
3210
3211
3212
3213
3214
3215
3216
3217
3218
3219
3220
3221
3222
3223
3224
3225
3226
3227
3228
3229
3230
3231
3232
3233
3234
3235
3236
3237
3238
3239
3240
3241
3242
3243
3244
3245
3246
3247
3248
3249
3250
3251
3252
3253
3254
3255
3256
3257
3258
3259
3260
3261
3262
3263
3264
3265
3266
3267
3268
3269
3270
3271
3272
3273
3274
3275
3276
3277
3278
3279
3280
3281
3282
3283
3284
3285
3286
3287
3288
3289
3290
3291
3292
3293
3294
3295
3296
3297
3298
3299
3300
3301
3302
3303
3304
3305
3306
3307
3308
3309
3310
3311
3312
3313
3314
3315
3316
3317
3318
3319
3320
3321
3322
3323
3324
3325
3326
3327
3328
3329
3330
3331
3332
3333
3334
3335
3336
3337
3338
3339
3340
3341
3342
3343
3344
3345
3346
3347
3348
3349
3350
3351
3352
3353
3354
3355
3356
3357
3358
3359
3360
3361
3362
3363
3364
3365
3366
3367
3368
3369
3370
3371
3372
3373
3374
3375
3376
3377
3378
3379
3380
3381
3382
3383
3384
3385
3386
3387
3388
3389
3390
3391
3392
3393
3394
3395
3396
3397
3398
3399
3400
3401
3402
3403
3404
3405
3406
3407
3408
3409
3410
3411
3412
3413
3414
3415
3416
3417
3418
3419
3420
3421
3422
3423
3424
3425
3426
3427
3428
3429
3430
3431
3432
3433
3434
3435
3436
3437
3438
3439
3440
3441
3442
3443
3444
3445
3446
3447
3448
3449
3450
3451
3452
3453
3454
3455
3456
3457
3458
3459
3460
3461
3462
3463
3464
3465
3466
3467
3468
3469
3470
3471
3472
3473
3474
3475
3476
3477
3478
3479
3480
3481
3482
3483
3484
3485
3486
3487
3488
3489
3490
3491
3492
3493
3494
3495
3496
3497
3498
3499
3500
3501
3502
3503
3504
3505
3506
3507
3508
3509
3510
3511
3512
3513
3514
3515
3516
3517
3518
3519
3520
3521
3522
3523
3524
3525
3526
3527
3528
3529
3530
3531
3532
3533
3534
3535
3536
3537
3538
3539
3540
3541
3542
3543
3544
3545
3546
3547
3548
3549
3550
3551
3552
3553
3554
3555
3556
3557
3558
3559
3560
3561
3562
3563
3564
3565
3566
3567
3568
3569
3570
3571
3572
3573
3574
3575
3576
3577
3578
3579
3580
3581
3582
3583
3584
3585
3586
3587
3588
3589
3590
3591
3592
3593
3594
3595
3596
3597
3598
3599
3600
3601
3602
3603
3604
3605
3606
3607
3608
3609
3610
3611
3612
3613
3614
3615
3616
3617
3618
3619
3620
3621
3622
3623
3624
3625
3626
3627
3628
3629
3630
3631
3632
3633
3634
3635
3636
3637
3638
3639
3640
3641
3642
3643
3644
3645
3646
3647
3648
3649
3650
3651
3652
3653
3654
3655
3656
3657
3658
3659
3660
3661
3662
3663
3664
3665
3666
3667
3668
3669
3670
3671
3672
3673
3674
3675
3676
3677
3678
3679
3680
3681
3682
3683
3684
3685
3686
3687
3688
3689
3690
3691
3692
3693
3694
3695
3696
3697
3698
3699
3700
3701
3702
3703
3704
3705
3706
3707
3708
3709
3710
3711
3712
3713
3714
3715
3716
3717
3718
3719
3720
3721
3722
3723
3724
3725
3726
3727
3728
3729
3730
3731
3732
3733
3734
3735
3736
3737
3738
3739
3740
3741
3742
3743
3744
3745
3746
3747
3748
3749
3750
3751
3752
3753
3754
3755
3756
3757
3758
3759
3760
3761
3762
3763
3764
3765
3766
3767
3768
3769
3770
3771
3772
3773
3774
3775
3776
3777
3778
3779
3780
3781
3782
3783
3784
3785
3786
3787
3788
3789
3790
3791
3792
3793
3794
3795
3796
3797
3798
3799
3800
3801
3802
3803
3804
3805
3806
3807
3808
3809
3810
3811
3812
3813
3814
3815
3816
3817
3818
3819
3820
3821
3822
3823
3824
3825
3826
3827
3828
3829
3830
3831
3832
3833
3834
3835
3836
3837
3838
3839
3840
3841
3842
3843
3844
3845
3846
3847
3848
3849
3850
3851
3852
3853
3854
3855
3856
3857
3858
3859
3860
3861
3862
3863
3864
3865
3866
3867
3868
3869
3870
3871
3872
3873
3874
3875
3876
3877
3878
3879
3880
3881
3882
3883
3884
3885
3886
3887
3888
3889
3890
3891
3892
3893
3894
3895
3896
3897
3898
3899
3900
3901
3902
3903
3904
3905
3906
3907
3908
3909
3910
3911
3912
3913
3914
3915
3916
3917
3918
3919
3920
3921
3922
3923
3924
3925
3926
3927
3928
3929
3930
3931
3932
3933
3934
3935
3936
3937
3938
3939
3940
3941
3942
3943
3944
3945
3946
3947
3948
3949
3950
3951
3952
3953
3954
3955
3956
3957
3958
3959
3960
3961
3962
3963
3964
3965
3966
3967
3968
3969
3970
3971
3972
3973
3974
3975
3976
3977
3978
3979
3980
3981
3982
3983
3984
3985
3986
3987
3988
3989
3990
3991
3992
3993
3994
3995
3996
3997
3998
3999
4000
4001
4002
4003
4004
4005
4006
4007
4008
4009
4010
4011
4012
4013
4014
4015
4016
4017
4018
4019
4020
4021
4022
4023
4024
4025
4026
4027
4028
4029
4030
4031
4032
4033
4034
4035
4036
4037
4038
4039
4040
4041
4042
4043
4044
4045
4046
4047
4048
4049
4050
4051
4052
4053
4054
4055
4056
4057
4058
4059
4060
4061
4062
4063
4064
4065
4066
4067
4068
4069
4070
4071
4072
4073
4074
4075
4076
4077
4078
4079
4080
4081
4082
4083
4084
4085
4086
4087
4088
4089
4090
4091
4092
4093
4094
4095
4096
4097
4098
4099
4100
4101
4102
4103
4104
4105
4106
4107
4108
4109
4110
4111
4112
4113
4114
4115
4116
4117
4118
4119
4120
4121
4122
4123
4124
4125
4126
4127
4128
4129
4130
4131
4132
4133
4134
4135
4136
4137
4138
4139
4140
4141
4142
4143
4144
4145
4146
4147
4148
4149
4150
4151
4152
4153
4154
4155
4156
4157
4158
4159
4160
4161
4162
4163
4164
4165
4166
4167
4168
4169
4170
4171
4172
4173
4174
4175
4176
4177
4178
4179
4180
4181
4182
4183
4184
4185
4186
4187
4188
4189
4190
4191
4192
4193
4194
4195
4196
4197
4198
4199
4200
4201
4202
4203
4204
4205
4206
4207
4208
4209
4210
4211
4212
4213
4214
4215
4216
4217
4218
4219
4220
4221
4222
4223
4224
4225
4226
4227
4228
4229
4230
4231
4232
4233
4234
4235
4236
4237
4238
4239
4240
4241
4242
4243
4244
4245
4246
4247
4248
4249
4250
4251
4252
4253
4254
4255
4256
4257
4258
4259
4260
4261
4262
4263
4264
4265
4266
4267
4268
4269
4270
4271
4272
4273
4274
4275
4276
4277
4278
4279
4280
4281
4282
4283
4284
4285
4286
4287
4288
4289
4290
4291
4292
4293
4294
4295
4296
4297
4298
4299
4300
4301
4302
4303
4304
4305
4306
4307
4308
4309
4310
4311
4312
4313
4314
4315
4316
4317
4318
4319
4320
4321
4322
4323
4324
4325
4326
4327
4328
4329
4330
4331
4332
4333
4334
4335
4336
4337
4338
4339
4340
4341
4342
4343
4344
4345
4346
4347
4348
4349
4350
4351
4352
4353
4354
4355
4356
4357
4358
4359
4360
4361
4362
4363
4364
4365
4366
4367
4368
4369
4370
4371
4372
4373
4374
4375
4376
4377
4378
4379
4380
4381
4382
4383
4384
4385
4386
4387
4388
4389
4390
4391
4392
4393
4394
4395
4396
4397
4398
4399
4400
4401
4402
4403
4404
4405
4406
4407
4408
4409
4410
4411
4412
4413
4414
4415
4416
4417
4418
4419
4420
4421
4422
4423
4424
4425
4426
4427
4428
4429
4430
4431
4432
4433
4434
4435
4436
4437
4438
4439
4440
4441
4442
4443
4444
4445
4446
4447
4448
4449
4450
4451
4452
4453
4454
4455
4456
4457
4458
4459
4460
4461
4462
4463
4464
4465
4466
4467
4468
4469
4470
4471
4472
4473
4474
4475
4476
4477
4478
4479
4480
4481
4482
4483
4484
4485
4486
4487
4488
4489
4490
4491
4492
4493
4494
4495
4496
4497
4498
4499
4500
4501
4502
4503
4504
4505
4506
4507
4508
4509
4510
4511
4512
4513
4514
4515
4516
4517
4518
4519
4520
4521
4522
4523
4524
4525
4526
4527
4528
4529
4530
4531
4532
4533
4534
4535
4536
4537
4538
4539
4540
4541
4542
4543
4544
4545
4546
4547
4548
4549
4550
4551
4552
4553
4554
4555
4556
4557
4558
4559
4560
4561
4562
4563
4564
4565
4566
4567
4568
4569
4570
4571
4572
4573
4574
4575
4576
4577
4578
4579
4580
4581
4582
4583
4584
4585
4586
4587
4588
4589
4590
4591
4592
4593
4594
4595
4596
4597
4598
4599
4600
4601
4602
4603
4604
4605
4606
4607
4608
4609
4610
4611
4612
4613
4614
4615
4616
4617
4618
4619
4620
4621
4622
4623
4624
4625
4626
4627
4628
4629
4630
4631
4632
4633
4634
4635
4636
4637
4638
4639
4640
4641
4642
4643
4644
4645
4646
4647
4648
4649
4650
4651
4652
4653
4654
4655
4656
4657
4658
4659
4660
4661
4662
4663
4664
4665
4666
4667
4668
4669
4670
4671
4672
4673
4674
4675
4676
4677
4678
4679
4680
4681
4682
4683
4684
4685
4686
4687
4688
4689
4690
4691
4692
4693
4694
4695
4696
4697
4698
4699
4700
4701
4702
4703
4704
4705
4706
4707
4708
4709
4710
4711
4712
4713
4714
4715
4716
4717
4718
4719
4720
4721
4722
4723
4724
4725
4726
4727
4728
4729
4730
4731
4732
4733
4734
4735
4736
4737
4738
4739
4740
4741
4742
4743
4744
4745
4746
4747
4748
4749
4750
4751
4752
4753
4754
4755
4756
4757
4758
4759
4760
4761
4762
4763
4764
4765
4766
4767
4768
4769
4770
4771
4772
4773
4774
4775
4776
4777
4778
4779
4780
4781
4782
4783
4784
4785
4786
4787
4788
4789
4790
4791
4792
4793
4794
4795
4796
4797
4798
4799
4800
4801
4802
4803
4804
4805
4806
4807
4808
4809
4810
4811
4812
4813
4814
4815
4816
4817
4818
4819
4820
4821
4822
4823
4824
4825
4826
4827
4828
4829
4830
4831
4832
4833
4834
4835
4836
4837
4838
4839
4840
4841
4842
4843
4844
4845
4846
4847
4848
4849
4850
4851
/*
 * 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, 2019 by Delphix. All rights reserved.
 * Copyright (c) 2011 Nexenta Systems, Inc. All rights reserved.
 * Copyright (c) 2017, Intel Corporation.
 */

#include <sys/sysmacros.h>
#include <sys/zfs_context.h>
#include <sys/fm/fs/zfs.h>
#include <sys/spa.h>
#include <sys/txg.h>
#include <sys/spa_impl.h>
#include <sys/vdev_impl.h>
#include <sys/vdev_trim.h>
#include <sys/zio_impl.h>
#include <sys/zio_compress.h>
#include <sys/zio_checksum.h>
#include <sys/dmu_objset.h>
#include <sys/arc.h>
#include <sys/ddt.h>
#include <sys/blkptr.h>
#include <sys/zfeature.h>
#include <sys/dsl_scan.h>
#include <sys/metaslab_impl.h>
#include <sys/time.h>
#include <sys/trace_zio.h>
#include <sys/abd.h>
#include <sys/dsl_crypt.h>
#include <sys/cityhash.h>

/*
 * ==========================================================================
 * I/O type descriptions
 * ==========================================================================
 */
const char *zio_type_name[ZIO_TYPES] = {
	/*
	 * Note: Linux kernel thread name length is limited
	 * so these names will differ from upstream open zfs.
	 */
	"z_null", "z_rd", "z_wr", "z_fr", "z_cl", "z_ioctl", "z_trim"
};

int zio_dva_throttle_enabled = B_TRUE;
int zio_deadman_log_all = B_FALSE;

/*
 * ==========================================================================
 * I/O kmem caches
 * ==========================================================================
 */
kmem_cache_t *zio_cache;
kmem_cache_t *zio_link_cache;
kmem_cache_t *zio_buf_cache[SPA_MAXBLOCKSIZE >> SPA_MINBLOCKSHIFT];
kmem_cache_t *zio_data_buf_cache[SPA_MAXBLOCKSIZE >> SPA_MINBLOCKSHIFT];
#if defined(ZFS_DEBUG) && !defined(_KERNEL)
uint64_t zio_buf_cache_allocs[SPA_MAXBLOCKSIZE >> SPA_MINBLOCKSHIFT];
uint64_t zio_buf_cache_frees[SPA_MAXBLOCKSIZE >> SPA_MINBLOCKSHIFT];
#endif

/* Mark IOs as "slow" if they take longer than 30 seconds */
int zio_slow_io_ms = (30 * MILLISEC);

#define	BP_SPANB(indblkshift, level) \
	(((uint64_t)1) << ((level) * ((indblkshift) - SPA_BLKPTRSHIFT)))
#define	COMPARE_META_LEVEL	0x80000000ul
/*
 * The following actions directly effect the spa's sync-to-convergence logic.
 * The values below define the sync pass when we start performing the action.
 * Care should be taken when changing these values as they directly impact
 * spa_sync() performance. Tuning these values may introduce subtle performance
 * pathologies and should only be done in the context of performance analysis.
 * These tunables will eventually be removed and replaced with #defines once
 * enough analysis has been done to determine optimal values.
 *
 * The 'zfs_sync_pass_deferred_free' pass must be greater than 1 to ensure that
 * regular blocks are not deferred.
 *
 * Starting in sync pass 8 (zfs_sync_pass_dont_compress), we disable
 * compression (including of metadata).  In practice, we don't have this
 * many sync passes, so this has no effect.
 *
 * The original intent was that disabling compression would help the sync
 * passes to converge. However, in practice disabling compression increases
 * the average number of sync passes, because when we turn compression off, a
 * lot of block's size will change and thus we have to re-allocate (not
 * overwrite) them. It also increases the number of 128KB allocations (e.g.
 * for indirect blocks and spacemaps) because these will not be compressed.
 * The 128K allocations are especially detrimental to performance on highly
 * fragmented systems, which may have very few free segments of this size,
 * and may need to load new metaslabs to satisfy 128K allocations.
 */
int zfs_sync_pass_deferred_free = 2; /* defer frees starting in this pass */
int zfs_sync_pass_dont_compress = 8; /* don't compress starting in this pass */
int zfs_sync_pass_rewrite = 2; /* rewrite new bps starting in this pass */

/*
 * An allocating zio is one that either currently has the DVA allocate
 * stage set or will have it later in its lifetime.
 */
#define	IO_IS_ALLOCATING(zio) ((zio)->io_orig_pipeline & ZIO_STAGE_DVA_ALLOCATE)

int zio_requeue_io_start_cut_in_line = 1;

#ifdef ZFS_DEBUG
int zio_buf_debug_limit = 16384;
#else
int zio_buf_debug_limit = 0;
#endif

static inline void __zio_execute(zio_t *zio);

static void zio_taskq_dispatch(zio_t *, zio_taskq_type_t, boolean_t);

void
zio_init(void)
{
	size_t c;
	vmem_t *data_alloc_arena = NULL;

	zio_cache = kmem_cache_create("zio_cache",
	    sizeof (zio_t), 0, NULL, NULL, NULL, NULL, NULL, 0);
	zio_link_cache = kmem_cache_create("zio_link_cache",
	    sizeof (zio_link_t), 0, NULL, NULL, NULL, NULL, NULL, 0);

	/*
	 * For small buffers, we want a cache for each multiple of
	 * SPA_MINBLOCKSIZE.  For larger buffers, we want a cache
	 * for each quarter-power of 2.
	 */
	for (c = 0; c < SPA_MAXBLOCKSIZE >> SPA_MINBLOCKSHIFT; c++) {
		size_t size = (c + 1) << SPA_MINBLOCKSHIFT;
		size_t p2 = size;
		size_t align = 0;
		size_t cflags = (size > zio_buf_debug_limit) ? KMC_NODEBUG : 0;

#if defined(_ILP32) && defined(_KERNEL)
		/*
		 * Cache size limited to 1M on 32-bit platforms until ARC
		 * buffers no longer require virtual address space.
		 */
		if (size > zfs_max_recordsize)
			break;
#endif

		while (!ISP2(p2))
			p2 &= p2 - 1;

#ifndef _KERNEL
		/*
		 * If we are using watchpoints, put each buffer on its own page,
		 * to eliminate the performance overhead of trapping to the
		 * kernel when modifying a non-watched buffer that shares the
		 * page with a watched buffer.
		 */
		if (arc_watch && !IS_P2ALIGNED(size, PAGESIZE))
			continue;
		/*
		 * Here's the problem - on 4K native devices in userland on
		 * Linux using O_DIRECT, buffers must be 4K aligned or I/O
		 * will fail with EINVAL, causing zdb (and others) to coredump.
		 * Since userland probably doesn't need optimized buffer caches,
		 * we just force 4K alignment on everything.
		 */
		align = 8 * SPA_MINBLOCKSIZE;
#else
		if (size < PAGESIZE) {
			align = SPA_MINBLOCKSIZE;
		} else if (IS_P2ALIGNED(size, p2 >> 2)) {
			align = PAGESIZE;
		}
#endif

		if (align != 0) {
			char name[36];
			(void) sprintf(name, "zio_buf_%lu", (ulong_t)size);
			zio_buf_cache[c] = kmem_cache_create(name, size,
			    align, NULL, NULL, NULL, NULL, NULL, cflags);

			(void) sprintf(name, "zio_data_buf_%lu", (ulong_t)size);
			zio_data_buf_cache[c] = kmem_cache_create(name, size,
			    align, NULL, NULL, NULL, NULL,
			    data_alloc_arena, cflags);
		}
	}

	while (--c != 0) {
		ASSERT(zio_buf_cache[c] != NULL);
		if (zio_buf_cache[c - 1] == NULL)
			zio_buf_cache[c - 1] = zio_buf_cache[c];

		ASSERT(zio_data_buf_cache[c] != NULL);
		if (zio_data_buf_cache[c - 1] == NULL)
			zio_data_buf_cache[c - 1] = zio_data_buf_cache[c];
	}

	zio_inject_init();

	lz4_init();
}

void
zio_fini(void)
{
	size_t c;
	kmem_cache_t *last_cache = NULL;
	kmem_cache_t *last_data_cache = NULL;

	for (c = 0; c < SPA_MAXBLOCKSIZE >> SPA_MINBLOCKSHIFT; c++) {
#ifdef _ILP32
		/*
		 * Cache size limited to 1M on 32-bit platforms until ARC
		 * buffers no longer require virtual address space.
		 */
		if (((c + 1) << SPA_MINBLOCKSHIFT) > zfs_max_recordsize)
			break;
#endif
#if defined(ZFS_DEBUG) && !defined(_KERNEL)
		if (zio_buf_cache_allocs[c] != zio_buf_cache_frees[c])
			(void) printf("zio_fini: [%d] %llu != %llu\n",
			    (int)((c + 1) << SPA_MINBLOCKSHIFT),
			    (long long unsigned)zio_buf_cache_allocs[c],
			    (long long unsigned)zio_buf_cache_frees[c]);
#endif
		if (zio_buf_cache[c] != last_cache) {
			last_cache = zio_buf_cache[c];
			kmem_cache_destroy(zio_buf_cache[c]);
		}
		zio_buf_cache[c] = NULL;

		if (zio_data_buf_cache[c] != last_data_cache) {
			last_data_cache = zio_data_buf_cache[c];
			kmem_cache_destroy(zio_data_buf_cache[c]);
		}
		zio_data_buf_cache[c] = NULL;
	}

	kmem_cache_destroy(zio_link_cache);
	kmem_cache_destroy(zio_cache);

	zio_inject_fini();

	lz4_fini();
}

/*
 * ==========================================================================
 * Allocate and free I/O buffers
 * ==========================================================================
 */

/*
 * Use zio_buf_alloc to allocate ZFS metadata.  This data will appear in a
 * crashdump if the kernel panics, so use it judiciously.  Obviously, it's
 * useful to inspect ZFS metadata, but if possible, we should avoid keeping
 * excess / transient data in-core during a crashdump.
 */
void *
zio_buf_alloc(size_t size)
{
	size_t c = (size - 1) >> SPA_MINBLOCKSHIFT;

	VERIFY3U(c, <, SPA_MAXBLOCKSIZE >> SPA_MINBLOCKSHIFT);
#if defined(ZFS_DEBUG) && !defined(_KERNEL)
	atomic_add_64(&zio_buf_cache_allocs[c], 1);
#endif

	return (kmem_cache_alloc(zio_buf_cache[c], KM_PUSHPAGE));
}

/*
 * Use zio_data_buf_alloc to allocate data.  The data will not appear in a
 * crashdump if the kernel panics.  This exists so that we will limit the amount
 * of ZFS data that shows up in a kernel crashdump.  (Thus reducing the amount
 * of kernel heap dumped to disk when the kernel panics)
 */
void *
zio_data_buf_alloc(size_t size)
{
	size_t c = (size - 1) >> SPA_MINBLOCKSHIFT;

	VERIFY3U(c, <, SPA_MAXBLOCKSIZE >> SPA_MINBLOCKSHIFT);

	return (kmem_cache_alloc(zio_data_buf_cache[c], KM_PUSHPAGE));
}

void
zio_buf_free(void *buf, size_t size)
{
	size_t c = (size - 1) >> SPA_MINBLOCKSHIFT;

	VERIFY3U(c, <, SPA_MAXBLOCKSIZE >> SPA_MINBLOCKSHIFT);
#if defined(ZFS_DEBUG) && !defined(_KERNEL)
	atomic_add_64(&zio_buf_cache_frees[c], 1);
#endif

	kmem_cache_free(zio_buf_cache[c], buf);
}

void
zio_data_buf_free(void *buf, size_t size)
{
	size_t c = (size - 1) >> SPA_MINBLOCKSHIFT;

	VERIFY3U(c, <, SPA_MAXBLOCKSIZE >> SPA_MINBLOCKSHIFT);

	kmem_cache_free(zio_data_buf_cache[c], buf);
}

static void
zio_abd_free(void *abd, size_t size)
{
	abd_free((abd_t *)abd);
}

/*
 * ==========================================================================
 * Push and pop I/O transform buffers
 * ==========================================================================
 */
void
zio_push_transform(zio_t *zio, abd_t *data, uint64_t size, uint64_t bufsize,
    zio_transform_func_t *transform)
{
	zio_transform_t *zt = kmem_alloc(sizeof (zio_transform_t), KM_SLEEP);

	zt->zt_orig_abd = zio->io_abd;
	zt->zt_orig_size = zio->io_size;
	zt->zt_bufsize = bufsize;
	zt->zt_transform = transform;

	zt->zt_next = zio->io_transform_stack;
	zio->io_transform_stack = zt;

	zio->io_abd = data;
	zio->io_size = size;
}

void
zio_pop_transforms(zio_t *zio)
{
	zio_transform_t *zt;

	while ((zt = zio->io_transform_stack) != NULL) {
		if (zt->zt_transform != NULL)
			zt->zt_transform(zio,
			    zt->zt_orig_abd, zt->zt_orig_size);

		if (zt->zt_bufsize != 0)
			abd_free(zio->io_abd);

		zio->io_abd = zt->zt_orig_abd;
		zio->io_size = zt->zt_orig_size;
		zio->io_transform_stack = zt->zt_next;

		kmem_free(zt, sizeof (zio_transform_t));
	}
}

/*
 * ==========================================================================
 * I/O transform callbacks for subblocks, decompression, and decryption
 * ==========================================================================
 */
static void
zio_subblock(zio_t *zio, abd_t *data, uint64_t size)
{
	ASSERT(zio->io_size > size);

	if (zio->io_type == ZIO_TYPE_READ)
		abd_copy(data, zio->io_abd, size);
}

static void
zio_decompress(zio_t *zio, abd_t *data, uint64_t size)
{
	if (zio->io_error == 0) {
		void *tmp = abd_borrow_buf(data, size);
		int ret = zio_decompress_data(BP_GET_COMPRESS(zio->io_bp),
		    zio->io_abd, tmp, zio->io_size, size);
		abd_return_buf_copy(data, tmp, size);

		if (zio_injection_enabled && ret == 0)
			ret = zio_handle_fault_injection(zio, EINVAL);

		if (ret != 0)
			zio->io_error = SET_ERROR(EIO);
	}
}

static void
zio_decrypt(zio_t *zio, abd_t *data, uint64_t size)
{
	int ret;
	void *tmp;
	blkptr_t *bp = zio->io_bp;
	spa_t *spa = zio->io_spa;
	uint64_t dsobj = zio->io_bookmark.zb_objset;
	uint64_t lsize = BP_GET_LSIZE(bp);
	dmu_object_type_t ot = BP_GET_TYPE(bp);
	uint8_t salt[ZIO_DATA_SALT_LEN];
	uint8_t iv[ZIO_DATA_IV_LEN];
	uint8_t mac[ZIO_DATA_MAC_LEN];
	boolean_t no_crypt = B_FALSE;

	ASSERT(BP_USES_CRYPT(bp));
	ASSERT3U(size, !=, 0);

	if (zio->io_error != 0)
		return;

	/*
	 * Verify the cksum of MACs stored in an indirect bp. It will always
	 * be possible to verify this since it does not require an encryption
	 * key.
	 */
	if (BP_HAS_INDIRECT_MAC_CKSUM(bp)) {
		zio_crypt_decode_mac_bp(bp, mac);

		if (BP_GET_COMPRESS(bp) != ZIO_COMPRESS_OFF) {
			/*
			 * We haven't decompressed the data yet, but
			 * zio_crypt_do_indirect_mac_checksum() requires
			 * decompressed data to be able to parse out the MACs
			 * from the indirect block. We decompress it now and
			 * throw away the result after we are finished.
			 */
			tmp = zio_buf_alloc(lsize);
			ret = zio_decompress_data(BP_GET_COMPRESS(bp),
			    zio->io_abd, tmp, zio->io_size, lsize);
			if (ret != 0) {
				ret = SET_ERROR(EIO);
				goto error;
			}
			ret = zio_crypt_do_indirect_mac_checksum(B_FALSE,
			    tmp, lsize, BP_SHOULD_BYTESWAP(bp), mac);
			zio_buf_free(tmp, lsize);
		} else {
			ret = zio_crypt_do_indirect_mac_checksum_abd(B_FALSE,
			    zio->io_abd, size, BP_SHOULD_BYTESWAP(bp), mac);
		}
		abd_copy(data, zio->io_abd, size);

		if (zio_injection_enabled && ot != DMU_OT_DNODE && ret == 0) {
			ret = zio_handle_decrypt_injection(spa,
			    &zio->io_bookmark, ot, ECKSUM);
		}
		if (ret != 0)
			goto error;

		return;
	}

	/*
	 * If this is an authenticated block, just check the MAC. It would be
	 * nice to separate this out into its own flag, but for the moment
	 * enum zio_flag is out of bits.
	 */
	if (BP_IS_AUTHENTICATED(bp)) {
		if (ot == DMU_OT_OBJSET) {
			ret = spa_do_crypt_objset_mac_abd(B_FALSE, spa,
			    dsobj, zio->io_abd, size, BP_SHOULD_BYTESWAP(bp));
		} else {
			zio_crypt_decode_mac_bp(bp, mac);
			ret = spa_do_crypt_mac_abd(B_FALSE, spa, dsobj,
			    zio->io_abd, size, mac);
			if (zio_injection_enabled && ret == 0) {
				ret = zio_handle_decrypt_injection(spa,
				    &zio->io_bookmark, ot, ECKSUM);
			}
		}
		abd_copy(data, zio->io_abd, size);

		if (ret != 0)
			goto error;

		return;
	}

	zio_crypt_decode_params_bp(bp, salt, iv);

	if (ot == DMU_OT_INTENT_LOG) {
		tmp = abd_borrow_buf_copy(zio->io_abd, sizeof (zil_chain_t));
		zio_crypt_decode_mac_zil(tmp, mac);
		abd_return_buf(zio->io_abd, tmp, sizeof (zil_chain_t));
	} else {
		zio_crypt_decode_mac_bp(bp, mac);
	}

	ret = spa_do_crypt_abd(B_FALSE, spa, &zio->io_bookmark, BP_GET_TYPE(bp),
	    BP_GET_DEDUP(bp), BP_SHOULD_BYTESWAP(bp), salt, iv, mac, size, data,
	    zio->io_abd, &no_crypt);
	if (no_crypt)
		abd_copy(data, zio->io_abd, size);

	if (ret != 0)
		goto error;

	return;

error:
	/* assert that the key was found unless this was speculative */
	ASSERT(ret != EACCES || (zio->io_flags & ZIO_FLAG_SPECULATIVE));

	/*
	 * If there was a decryption / authentication error return EIO as
	 * the io_error. If this was not a speculative zio, create an ereport.
	 */
	if (ret == ECKSUM) {
		zio->io_error = SET_ERROR(EIO);
		if ((zio->io_flags & ZIO_FLAG_SPECULATIVE) == 0) {
			spa_log_error(spa, &zio->io_bookmark);
			zfs_ereport_post(FM_EREPORT_ZFS_AUTHENTICATION,
			    spa, NULL, &zio->io_bookmark, zio, 0, 0);
		}
	} else {
		zio->io_error = ret;
	}
}

/*
 * ==========================================================================
 * I/O parent/child relationships and pipeline interlocks
 * ==========================================================================
 */
zio_t *
zio_walk_parents(zio_t *cio, zio_link_t **zl)
{
	list_t *pl = &cio->io_parent_list;

	*zl = (*zl == NULL) ? list_head(pl) : list_next(pl, *zl);
	if (*zl == NULL)
		return (NULL);

	ASSERT((*zl)->zl_child == cio);
	return ((*zl)->zl_parent);
}

zio_t *
zio_walk_children(zio_t *pio, zio_link_t **zl)
{
	list_t *cl = &pio->io_child_list;

	ASSERT(MUTEX_HELD(&pio->io_lock));

	*zl = (*zl == NULL) ? list_head(cl) : list_next(cl, *zl);
	if (*zl == NULL)
		return (NULL);

	ASSERT((*zl)->zl_parent == pio);
	return ((*zl)->zl_child);
}

zio_t *
zio_unique_parent(zio_t *cio)
{
	zio_link_t *zl = NULL;
	zio_t *pio = zio_walk_parents(cio, &zl);

	VERIFY3P(zio_walk_parents(cio, &zl), ==, NULL);
	return (pio);
}

void
zio_add_child(zio_t *pio, zio_t *cio)
{
	zio_link_t *zl = kmem_cache_alloc(zio_link_cache, KM_SLEEP);

	/*
	 * Logical I/Os can have logical, gang, or vdev children.
	 * Gang I/Os can have gang or vdev children.
	 * Vdev I/Os can only have vdev children.
	 * The following ASSERT captures all of these constraints.
	 */
	ASSERT3S(cio->io_child_type, <=, pio->io_child_type);

	zl->zl_parent = pio;
	zl->zl_child = cio;

	mutex_enter(&pio->io_lock);
	mutex_enter(&cio->io_lock);

	ASSERT(pio->io_state[ZIO_WAIT_DONE] == 0);

	for (int w = 0; w < ZIO_WAIT_TYPES; w++)
		pio->io_children[cio->io_child_type][w] += !cio->io_state[w];

	list_insert_head(&pio->io_child_list, zl);
	list_insert_head(&cio->io_parent_list, zl);

	pio->io_child_count++;
	cio->io_parent_count++;

	mutex_exit(&cio->io_lock);
	mutex_exit(&pio->io_lock);
}

static void
zio_remove_child(zio_t *pio, zio_t *cio, zio_link_t *zl)
{
	ASSERT(zl->zl_parent == pio);
	ASSERT(zl->zl_child == cio);

	mutex_enter(&pio->io_lock);
	mutex_enter(&cio->io_lock);

	list_remove(&pio->io_child_list, zl);
	list_remove(&cio->io_parent_list, zl);

	pio->io_child_count--;
	cio->io_parent_count--;

	mutex_exit(&cio->io_lock);
	mutex_exit(&pio->io_lock);
	kmem_cache_free(zio_link_cache, zl);
}

static boolean_t
zio_wait_for_children(zio_t *zio, uint8_t childbits, enum zio_wait_type wait)
{
	boolean_t waiting = B_FALSE;

	mutex_enter(&zio->io_lock);
	ASSERT(zio->io_stall == NULL);
	for (int c = 0; c < ZIO_CHILD_TYPES; c++) {
		if (!(ZIO_CHILD_BIT_IS_SET(childbits, c)))
			continue;

		uint64_t *countp = &zio->io_children[c][wait];
		if (*countp != 0) {
			zio->io_stage >>= 1;
			ASSERT3U(zio->io_stage, !=, ZIO_STAGE_OPEN);
			zio->io_stall = countp;
			waiting = B_TRUE;
			break;
		}
	}
	mutex_exit(&zio->io_lock);
	return (waiting);
}

__attribute__((always_inline))
static inline void
zio_notify_parent(zio_t *pio, zio_t *zio, enum zio_wait_type wait,
    zio_t **next_to_executep)
{
	uint64_t *countp = &pio->io_children[zio->io_child_type][wait];
	int *errorp = &pio->io_child_error[zio->io_child_type];

	mutex_enter(&pio->io_lock);
	if (zio->io_error && !(zio->io_flags & ZIO_FLAG_DONT_PROPAGATE))
		*errorp = zio_worst_error(*errorp, zio->io_error);
	pio->io_reexecute |= zio->io_reexecute;
	ASSERT3U(*countp, >, 0);

	(*countp)--;

	if (*countp == 0 && pio->io_stall == countp) {
		zio_taskq_type_t type =
		    pio->io_stage < ZIO_STAGE_VDEV_IO_START ? ZIO_TASKQ_ISSUE :
		    ZIO_TASKQ_INTERRUPT;
		pio->io_stall = NULL;
		mutex_exit(&pio->io_lock);

		/*
		 * If we can tell the caller to execute this parent next, do
		 * so.  Otherwise dispatch the parent zio as its own task.
		 *
		 * Having the caller execute the parent when possible reduces
		 * locking on the zio taskq's, reduces context switch
		 * overhead, and has no recursion penalty.  Note that one
		 * read from disk typically causes at least 3 zio's: a
		 * zio_null(), the logical zio_read(), and then a physical
		 * zio.  When the physical ZIO completes, we are able to call
		 * zio_done() on all 3 of these zio's from one invocation of
		 * zio_execute() by returning the parent back to
		 * zio_execute().  Since the parent isn't executed until this
		 * thread returns back to zio_execute(), the caller should do
		 * so promptly.
		 *
		 * In other cases, dispatching the parent prevents
		 * overflowing the stack when we have deeply nested
		 * parent-child relationships, as we do with the "mega zio"
		 * of writes for spa_sync(), and the chain of ZIL blocks.
		 */
		if (next_to_executep != NULL && *next_to_executep == NULL) {
			*next_to_executep = pio;
		} else {
			zio_taskq_dispatch(pio, type, B_FALSE);
		}
	} else {
		mutex_exit(&pio->io_lock);
	}
}

static void
zio_inherit_child_errors(zio_t *zio, enum zio_child c)
{
	if (zio->io_child_error[c] != 0 && zio->io_error == 0)
		zio->io_error = zio->io_child_error[c];
}

int
zio_bookmark_compare(const void *x1, const void *x2)
{
	const zio_t *z1 = x1;
	const zio_t *z2 = x2;

	if (z1->io_bookmark.zb_objset < z2->io_bookmark.zb_objset)
		return (-1);
	if (z1->io_bookmark.zb_objset > z2->io_bookmark.zb_objset)
		return (1);

	if (z1->io_bookmark.zb_object < z2->io_bookmark.zb_object)
		return (-1);
	if (z1->io_bookmark.zb_object > z2->io_bookmark.zb_object)
		return (1);

	if (z1->io_bookmark.zb_level < z2->io_bookmark.zb_level)
		return (-1);
	if (z1->io_bookmark.zb_level > z2->io_bookmark.zb_level)
		return (1);

	if (z1->io_bookmark.zb_blkid < z2->io_bookmark.zb_blkid)
		return (-1);
	if (z1->io_bookmark.zb_blkid > z2->io_bookmark.zb_blkid)
		return (1);

	if (z1 < z2)
		return (-1);
	if (z1 > z2)
		return (1);

	return (0);
}

/*
 * ==========================================================================
 * Create the various types of I/O (read, write, free, etc)
 * ==========================================================================
 */
static zio_t *
zio_create(zio_t *pio, spa_t *spa, uint64_t txg, const blkptr_t *bp,
    abd_t *data, uint64_t lsize, uint64_t psize, zio_done_func_t *done,
    void *private, zio_type_t type, zio_priority_t priority,
    enum zio_flag flags, vdev_t *vd, uint64_t offset,
    const zbookmark_phys_t *zb, enum zio_stage stage,
    enum zio_stage pipeline)
{
	zio_t *zio;

	IMPLY(type != ZIO_TYPE_TRIM, psize <= SPA_MAXBLOCKSIZE);
	ASSERT(P2PHASE(psize, SPA_MINBLOCKSIZE) == 0);
	ASSERT(P2PHASE(offset, SPA_MINBLOCKSIZE) == 0);

	ASSERT(!vd || spa_config_held(spa, SCL_STATE_ALL, RW_READER));
	ASSERT(!bp || !(flags & ZIO_FLAG_CONFIG_WRITER));
	ASSERT(vd || stage == ZIO_STAGE_OPEN);

	IMPLY(lsize != psize, (flags & ZIO_FLAG_RAW_COMPRESS) != 0);

	zio = kmem_cache_alloc(zio_cache, KM_SLEEP);
	bzero(zio, sizeof (zio_t));

	mutex_init(&zio->io_lock, NULL, MUTEX_NOLOCKDEP, NULL);
	cv_init(&zio->io_cv, NULL, CV_DEFAULT, NULL);

	list_create(&zio->io_parent_list, sizeof (zio_link_t),
	    offsetof(zio_link_t, zl_parent_node));
	list_create(&zio->io_child_list, sizeof (zio_link_t),
	    offsetof(zio_link_t, zl_child_node));
	metaslab_trace_init(&zio->io_alloc_list);

	if (vd != NULL)
		zio->io_child_type = ZIO_CHILD_VDEV;
	else if (flags & ZIO_FLAG_GANG_CHILD)
		zio->io_child_type = ZIO_CHILD_GANG;
	else if (flags & ZIO_FLAG_DDT_CHILD)
		zio->io_child_type = ZIO_CHILD_DDT;
	else
		zio->io_child_type = ZIO_CHILD_LOGICAL;

	if (bp != NULL) {
		zio->io_bp = (blkptr_t *)bp;
		zio->io_bp_copy = *bp;
		zio->io_bp_orig = *bp;
		if (type != ZIO_TYPE_WRITE ||
		    zio->io_child_type == ZIO_CHILD_DDT)
			zio->io_bp = &zio->io_bp_copy;	/* so caller can free */
		if (zio->io_child_type == ZIO_CHILD_LOGICAL)
			zio->io_logical = zio;
		if (zio->io_child_type > ZIO_CHILD_GANG && BP_IS_GANG(bp))
			pipeline |= ZIO_GANG_STAGES;
	}

	zio->io_spa = spa;
	zio->io_txg = txg;
	zio->io_done = done;
	zio->io_private = private;
	zio->io_type = type;
	zio->io_priority = priority;
	zio->io_vd = vd;
	zio->io_offset = offset;
	zio->io_orig_abd = zio->io_abd = data;
	zio->io_orig_size = zio->io_size = psize;
	zio->io_lsize = lsize;
	zio->io_orig_flags = zio->io_flags = flags;
	zio->io_orig_stage = zio->io_stage = stage;
	zio->io_orig_pipeline = zio->io_pipeline = pipeline;
	zio->io_pipeline_trace = ZIO_STAGE_OPEN;

	zio->io_state[ZIO_WAIT_READY] = (stage >= ZIO_STAGE_READY);
	zio->io_state[ZIO_WAIT_DONE] = (stage >= ZIO_STAGE_DONE);

	if (zb != NULL)
		zio->io_bookmark = *zb;

	if (pio != NULL) {
		if (zio->io_metaslab_class == NULL)
			zio->io_metaslab_class = pio->io_metaslab_class;
		if (zio->io_logical == NULL)
			zio->io_logical = pio->io_logical;
		if (zio->io_child_type == ZIO_CHILD_GANG)
			zio->io_gang_leader = pio->io_gang_leader;
		zio_add_child(pio, zio);
	}

	taskq_init_ent(&zio->io_tqent);

	return (zio);
}

static void
zio_destroy(zio_t *zio)
{
	metaslab_trace_fini(&zio->io_alloc_list);
	list_destroy(&zio->io_parent_list);
	list_destroy(&zio->io_child_list);
	mutex_destroy(&zio->io_lock);
	cv_destroy(&zio->io_cv);
	kmem_cache_free(zio_cache, zio);
}

zio_t *
zio_null(zio_t *pio, spa_t *spa, vdev_t *vd, zio_done_func_t *done,
    void *private, enum zio_flag flags)
{
	zio_t *zio;

	zio = zio_create(pio, spa, 0, NULL, NULL, 0, 0, done, private,
	    ZIO_TYPE_NULL, ZIO_PRIORITY_NOW, flags, vd, 0, NULL,
	    ZIO_STAGE_OPEN, ZIO_INTERLOCK_PIPELINE);

	return (zio);
}

zio_t *
zio_root(spa_t *spa, zio_done_func_t *done, void *private, enum zio_flag flags)
{
	return (zio_null(NULL, spa, NULL, done, private, flags));
}

void
zfs_blkptr_verify(spa_t *spa, const blkptr_t *bp)
{
	if (!DMU_OT_IS_VALID(BP_GET_TYPE(bp))) {
		zfs_panic_recover("blkptr at %p has invalid TYPE %llu",
		    bp, (longlong_t)BP_GET_TYPE(bp));
	}
	if (BP_GET_CHECKSUM(bp) >= ZIO_CHECKSUM_FUNCTIONS ||
	    BP_GET_CHECKSUM(bp) <= ZIO_CHECKSUM_ON) {
		zfs_panic_recover("blkptr at %p has invalid CHECKSUM %llu",
		    bp, (longlong_t)BP_GET_CHECKSUM(bp));
	}
	if (BP_GET_COMPRESS(bp) >= ZIO_COMPRESS_FUNCTIONS ||
	    BP_GET_COMPRESS(bp) <= ZIO_COMPRESS_ON) {
		zfs_panic_recover("blkptr at %p has invalid COMPRESS %llu",
		    bp, (longlong_t)BP_GET_COMPRESS(bp));
	}
	if (BP_GET_LSIZE(bp) > SPA_MAXBLOCKSIZE) {
		zfs_panic_recover("blkptr at %p has invalid LSIZE %llu",
		    bp, (longlong_t)BP_GET_LSIZE(bp));
	}
	if (BP_GET_PSIZE(bp) > SPA_MAXBLOCKSIZE) {
		zfs_panic_recover("blkptr at %p has invalid PSIZE %llu",
		    bp, (longlong_t)BP_GET_PSIZE(bp));
	}

	if (BP_IS_EMBEDDED(bp)) {
		if (BPE_GET_ETYPE(bp) > NUM_BP_EMBEDDED_TYPES) {
			zfs_panic_recover("blkptr at %p has invalid ETYPE %llu",
			    bp, (longlong_t)BPE_GET_ETYPE(bp));
		}
	}

	/*
	 * Do not verify individual DVAs if the config is not trusted. This
	 * will be done once the zio is executed in vdev_mirror_map_alloc.
	 */
	if (!spa->spa_trust_config)
		return;

	/*
	 * Pool-specific checks.
	 *
	 * Note: it would be nice to verify that the blk_birth and
	 * BP_PHYSICAL_BIRTH() are not too large.  However, spa_freeze()
	 * allows the birth time of log blocks (and dmu_sync()-ed blocks
	 * that are in the log) to be arbitrarily large.
	 */
	for (int i = 0; i < BP_GET_NDVAS(bp); i++) {
		uint64_t vdevid = DVA_GET_VDEV(&bp->blk_dva[i]);

		if (vdevid >= spa->spa_root_vdev->vdev_children) {
			zfs_panic_recover("blkptr at %p DVA %u has invalid "
			    "VDEV %llu",
			    bp, i, (longlong_t)vdevid);
			continue;
		}
		vdev_t *vd = spa->spa_root_vdev->vdev_child[vdevid];
		if (vd == NULL) {
			zfs_panic_recover("blkptr at %p DVA %u has invalid "
			    "VDEV %llu",
			    bp, i, (longlong_t)vdevid);
			continue;
		}
		if (vd->vdev_ops == &vdev_hole_ops) {
			zfs_panic_recover("blkptr at %p DVA %u has hole "
			    "VDEV %llu",
			    bp, i, (longlong_t)vdevid);
			continue;
		}
		if (vd->vdev_ops == &vdev_missing_ops) {
			/*
			 * "missing" vdevs are valid during import, but we
			 * don't have their detailed info (e.g. asize), so
			 * we can't perform any more checks on them.
			 */
			continue;
		}
		uint64_t offset = DVA_GET_OFFSET(&bp->blk_dva[i]);
		uint64_t asize = DVA_GET_ASIZE(&bp->blk_dva[i]);
		if (BP_IS_GANG(bp))
			asize = vdev_psize_to_asize(vd, SPA_GANGBLOCKSIZE);
		if (offset + asize > vd->vdev_asize) {
			zfs_panic_recover("blkptr at %p DVA %u has invalid "
			    "OFFSET %llu",
			    bp, i, (longlong_t)offset);
		}
	}
}

boolean_t
zfs_dva_valid(spa_t *spa, const dva_t *dva, const blkptr_t *bp)
{
	uint64_t vdevid = DVA_GET_VDEV(dva);

	if (vdevid >= spa->spa_root_vdev->vdev_children)
		return (B_FALSE);

	vdev_t *vd = spa->spa_root_vdev->vdev_child[vdevid];
	if (vd == NULL)
		return (B_FALSE);

	if (vd->vdev_ops == &vdev_hole_ops)
		return (B_FALSE);

	if (vd->vdev_ops == &vdev_missing_ops) {
		return (B_FALSE);
	}

	uint64_t offset = DVA_GET_OFFSET(dva);
	uint64_t asize = DVA_GET_ASIZE(dva);

	if (BP_IS_GANG(bp))
		asize = vdev_psize_to_asize(vd, SPA_GANGBLOCKSIZE);
	if (offset + asize > vd->vdev_asize)
		return (B_FALSE);

	return (B_TRUE);
}

zio_t *
zio_read(zio_t *pio, spa_t *spa, const blkptr_t *bp,
    abd_t *data, uint64_t size, zio_done_func_t *done, void *private,
    zio_priority_t priority, enum zio_flag flags, const zbookmark_phys_t *zb)
{
	zio_t *zio;

	zfs_blkptr_verify(spa, bp);

	zio = zio_create(pio, spa, BP_PHYSICAL_BIRTH(bp), bp,
	    data, size, size, done, private,
	    ZIO_TYPE_READ, priority, flags, NULL, 0, zb,
	    ZIO_STAGE_OPEN, (flags & ZIO_FLAG_DDT_CHILD) ?
	    ZIO_DDT_CHILD_READ_PIPELINE : ZIO_READ_PIPELINE);

	return (zio);
}

zio_t *
zio_write(zio_t *pio, spa_t *spa, uint64_t txg, blkptr_t *bp,
    abd_t *data, uint64_t lsize, uint64_t psize, const zio_prop_t *zp,
    zio_done_func_t *ready, zio_done_func_t *children_ready,
    zio_done_func_t *physdone, zio_done_func_t *done,
    void *private, zio_priority_t priority, enum zio_flag flags,
    const zbookmark_phys_t *zb)
{
	zio_t *zio;

	ASSERT(zp->zp_checksum >= ZIO_CHECKSUM_OFF &&
	    zp->zp_checksum < ZIO_CHECKSUM_FUNCTIONS &&
	    zp->zp_compress >= ZIO_COMPRESS_OFF &&
	    zp->zp_compress < ZIO_COMPRESS_FUNCTIONS &&
	    DMU_OT_IS_VALID(zp->zp_type) &&
	    zp->zp_level < 32 &&
	    zp->zp_copies > 0 &&
	    zp->zp_copies <= spa_max_replication(spa));

	zio = zio_create(pio, spa, txg, bp, data, lsize, psize, done, private,
	    ZIO_TYPE_WRITE, priority, flags, NULL, 0, zb,
	    ZIO_STAGE_OPEN, (flags & ZIO_FLAG_DDT_CHILD) ?
	    ZIO_DDT_CHILD_WRITE_PIPELINE : ZIO_WRITE_PIPELINE);

	zio->io_ready = ready;
	zio->io_children_ready = children_ready;
	zio->io_physdone = physdone;
	zio->io_prop = *zp;

	/*
	 * Data can be NULL if we are going to call zio_write_override() to
	 * provide the already-allocated BP.  But we may need the data to
	 * verify a dedup hit (if requested).  In this case, don't try to
	 * dedup (just take the already-allocated BP verbatim). Encrypted
	 * dedup blocks need data as well so we also disable dedup in this
	 * case.
	 */
	if (data == NULL &&
	    (zio->io_prop.zp_dedup_verify || zio->io_prop.zp_encrypt)) {
		zio->io_prop.zp_dedup = zio->io_prop.zp_dedup_verify = B_FALSE;
	}

	return (zio);
}

zio_t *
zio_rewrite(zio_t *pio, spa_t *spa, uint64_t txg, blkptr_t *bp, abd_t *data,
    uint64_t size, zio_done_func_t *done, void *private,
    zio_priority_t priority, enum zio_flag flags, zbookmark_phys_t *zb)
{
	zio_t *zio;

	zio = zio_create(pio, spa, txg, bp, data, size, size, done, private,
	    ZIO_TYPE_WRITE, priority, flags | ZIO_FLAG_IO_REWRITE, NULL, 0, zb,
	    ZIO_STAGE_OPEN, ZIO_REWRITE_PIPELINE);

	return (zio);
}

void
zio_write_override(zio_t *zio, blkptr_t *bp, int copies, boolean_t nopwrite)
{
	ASSERT(zio->io_type == ZIO_TYPE_WRITE);
	ASSERT(zio->io_child_type == ZIO_CHILD_LOGICAL);
	ASSERT(zio->io_stage == ZIO_STAGE_OPEN);
	ASSERT(zio->io_txg == spa_syncing_txg(zio->io_spa));

	/*
	 * We must reset the io_prop to match the values that existed
	 * when the bp was first written by dmu_sync() keeping in mind
	 * that nopwrite and dedup are mutually exclusive.
	 */
	zio->io_prop.zp_dedup = nopwrite ? B_FALSE : zio->io_prop.zp_dedup;
	zio->io_prop.zp_nopwrite = nopwrite;
	zio->io_prop.zp_copies = copies;
	zio->io_bp_override = bp;
}

void
zio_free(spa_t *spa, uint64_t txg, const blkptr_t *bp)
{

	zfs_blkptr_verify(spa, bp);

	/*
	 * The check for EMBEDDED is a performance optimization.  We
	 * process the free here (by ignoring it) rather than
	 * putting it on the list and then processing it in zio_free_sync().
	 */
	if (BP_IS_EMBEDDED(bp))
		return;
	metaslab_check_free(spa, bp);

	/*
	 * Frees that are for the currently-syncing txg, are not going to be
	 * deferred, and which will not need to do a read (i.e. not GANG or
	 * DEDUP), can be processed immediately.  Otherwise, put them on the
	 * in-memory list for later processing.
	 */
	if (BP_IS_GANG(bp) || BP_GET_DEDUP(bp) ||
	    txg != spa->spa_syncing_txg ||
	    spa_sync_pass(spa) >= zfs_sync_pass_deferred_free) {
		bplist_append(&spa->spa_free_bplist[txg & TXG_MASK], bp);
	} else {
		VERIFY0(zio_wait(zio_free_sync(NULL, spa, txg, bp, 0)));
	}
}

zio_t *
zio_free_sync(zio_t *pio, spa_t *spa, uint64_t txg, const blkptr_t *bp,
    enum zio_flag flags)
{
	zio_t *zio;
	enum zio_stage stage = ZIO_FREE_PIPELINE;

	ASSERT(!BP_IS_HOLE(bp));
	ASSERT(spa_syncing_txg(spa) == txg);
	ASSERT(spa_sync_pass(spa) < zfs_sync_pass_deferred_free);

	if (BP_IS_EMBEDDED(bp))
		return (zio_null(pio, spa, NULL, NULL, NULL, 0));

	metaslab_check_free(spa, bp);
	arc_freed(spa, bp);
	dsl_scan_freed(spa, bp);

	/*
	 * GANG and DEDUP blocks can induce a read (for the gang block header,
	 * or the DDT), so issue them asynchronously so that this thread is
	 * not tied up.
	 */
	if (BP_IS_GANG(bp) || BP_GET_DEDUP(bp))
		stage |= ZIO_STAGE_ISSUE_ASYNC;

	zio = zio_create(pio, spa, txg, bp, NULL, BP_GET_PSIZE(bp),
	    BP_GET_PSIZE(bp), NULL, NULL, ZIO_TYPE_FREE, ZIO_PRIORITY_NOW,
	    flags, NULL, 0, NULL, ZIO_STAGE_OPEN, stage);

	return (zio);
}

zio_t *
zio_claim(zio_t *pio, spa_t *spa, uint64_t txg, const blkptr_t *bp,
    zio_done_func_t *done, void *private, enum zio_flag flags)
{
	zio_t *zio;

	zfs_blkptr_verify(spa, bp);

	if (BP_IS_EMBEDDED(bp))
		return (zio_null(pio, spa, NULL, NULL, NULL, 0));

	/*
	 * A claim is an allocation of a specific block.  Claims are needed
	 * to support immediate writes in the intent log.  The issue is that
	 * immediate writes contain committed data, but in a txg that was
	 * *not* committed.  Upon opening the pool after an unclean shutdown,
	 * the intent log claims all blocks that contain immediate write data
	 * so that the SPA knows they're in use.
	 *
	 * All claims *must* be resolved in the first txg -- before the SPA
	 * starts allocating blocks -- so that nothing is allocated twice.
	 * If txg == 0 we just verify that the block is claimable.
	 */
	ASSERT3U(spa->spa_uberblock.ub_rootbp.blk_birth, <,
	    spa_min_claim_txg(spa));
	ASSERT(txg == spa_min_claim_txg(spa) || txg == 0);
	ASSERT(!BP_GET_DEDUP(bp) || !spa_writeable(spa));	/* zdb(1M) */

	zio = zio_create(pio, spa, txg, bp, NULL, BP_GET_PSIZE(bp),
	    BP_GET_PSIZE(bp), done, private, ZIO_TYPE_CLAIM, ZIO_PRIORITY_NOW,
	    flags, NULL, 0, NULL, ZIO_STAGE_OPEN, ZIO_CLAIM_PIPELINE);
	ASSERT0(zio->io_queued_timestamp);

	return (zio);
}

zio_t *
zio_ioctl(zio_t *pio, spa_t *spa, vdev_t *vd, int cmd,
    zio_done_func_t *done, void *private, enum zio_flag flags)
{
	zio_t *zio;
	int c;

	if (vd->vdev_children == 0) {
		zio = zio_create(pio, spa, 0, NULL, NULL, 0, 0, done, private,
		    ZIO_TYPE_IOCTL, ZIO_PRIORITY_NOW, flags, vd, 0, NULL,
		    ZIO_STAGE_OPEN, ZIO_IOCTL_PIPELINE);

		zio->io_cmd = cmd;
	} else {
		zio = zio_null(pio, spa, NULL, NULL, NULL, flags);

		for (c = 0; c < vd->vdev_children; c++)
			zio_nowait(zio_ioctl(zio, spa, vd->vdev_child[c], cmd,
			    done, private, flags));
	}

	return (zio);
}

zio_t *
zio_trim(zio_t *pio, vdev_t *vd, uint64_t offset, uint64_t size,
    zio_done_func_t *done, void *private, zio_priority_t priority,
    enum zio_flag flags, enum trim_flag trim_flags)
{
	zio_t *zio;

	ASSERT0(vd->vdev_children);
	ASSERT0(P2PHASE(offset, 1ULL << vd->vdev_ashift));
	ASSERT0(P2PHASE(size, 1ULL << vd->vdev_ashift));
	ASSERT3U(size, !=, 0);

	zio = zio_create(pio, vd->vdev_spa, 0, NULL, NULL, size, size, done,
	    private, ZIO_TYPE_TRIM, priority, flags | ZIO_FLAG_PHYSICAL,
	    vd, offset, NULL, ZIO_STAGE_OPEN, ZIO_TRIM_PIPELINE);
	zio->io_trim_flags = trim_flags;

	return (zio);
}

zio_t *
zio_read_phys(zio_t *pio, vdev_t *vd, uint64_t offset, uint64_t size,
    abd_t *data, int checksum, zio_done_func_t *done, void *private,
    zio_priority_t priority, enum zio_flag flags, boolean_t labels)
{
	zio_t *zio;

	ASSERT(vd->vdev_children == 0);
	ASSERT(!labels || offset + size <= VDEV_LABEL_START_SIZE ||
	    offset >= vd->vdev_psize - VDEV_LABEL_END_SIZE);
	ASSERT3U(offset + size, <=, vd->vdev_psize);

	zio = zio_create(pio, vd->vdev_spa, 0, NULL, data, size, size, done,
	    private, ZIO_TYPE_READ, priority, flags | ZIO_FLAG_PHYSICAL, vd,
	    offset, NULL, ZIO_STAGE_OPEN, ZIO_READ_PHYS_PIPELINE);

	zio->io_prop.zp_checksum = checksum;

	return (zio);
}

zio_t *
zio_write_phys(zio_t *pio, vdev_t *vd, uint64_t offset, uint64_t size,
    abd_t *data, int checksum, zio_done_func_t *done, void *private,
    zio_priority_t priority, enum zio_flag flags, boolean_t labels)
{
	zio_t *zio;

	ASSERT(vd->vdev_children == 0);
	ASSERT(!labels || offset + size <= VDEV_LABEL_START_SIZE ||
	    offset >= vd->vdev_psize - VDEV_LABEL_END_SIZE);
	ASSERT3U(offset + size, <=, vd->vdev_psize);

	zio = zio_create(pio, vd->vdev_spa, 0, NULL, data, size, size, done,
	    private, ZIO_TYPE_WRITE, priority, flags | ZIO_FLAG_PHYSICAL, vd,
	    offset, NULL, ZIO_STAGE_OPEN, ZIO_WRITE_PHYS_PIPELINE);

	zio->io_prop.zp_checksum = checksum;

	if (zio_checksum_table[checksum].ci_flags & ZCHECKSUM_FLAG_EMBEDDED) {
		/*
		 * zec checksums are necessarily destructive -- they modify
		 * the end of the write buffer to hold the verifier/checksum.
		 * Therefore, we must make a local copy in case the data is
		 * being written to multiple places in parallel.
		 */
		abd_t *wbuf = abd_alloc_sametype(data, size);
		abd_copy(wbuf, data, size);

		zio_push_transform(zio, wbuf, size, size, NULL);
	}

	return (zio);
}

/*
 * Create a child I/O to do some work for us.
 */
zio_t *
zio_vdev_child_io(zio_t *pio, blkptr_t *bp, vdev_t *vd, uint64_t offset,
    abd_t *data, uint64_t size, int type, zio_priority_t priority,
    enum zio_flag flags, zio_done_func_t *done, void *private)
{
	enum zio_stage pipeline = ZIO_VDEV_CHILD_PIPELINE;
	zio_t *zio;

	/*
	 * vdev child I/Os do not propagate their error to the parent.
	 * Therefore, for correct operation the caller *must* check for
	 * and handle the error in the child i/o's done callback.
	 * The only exceptions are i/os that we don't care about
	 * (OPTIONAL or REPAIR).
	 */
	ASSERT((flags & ZIO_FLAG_OPTIONAL) || (flags & ZIO_FLAG_IO_REPAIR) ||
	    done != NULL);

	if (type == ZIO_TYPE_READ && bp != NULL) {
		/*
		 * If we have the bp, then the child should perform the
		 * checksum and the parent need not.  This pushes error
		 * detection as close to the leaves as possible and
		 * eliminates redundant checksums in the interior nodes.
		 */
		pipeline |= ZIO_STAGE_CHECKSUM_VERIFY;
		pio->io_pipeline &= ~ZIO_STAGE_CHECKSUM_VERIFY;
	}

	if (vd->vdev_ops->vdev_op_leaf) {
		ASSERT0(vd->vdev_children);
		offset += VDEV_LABEL_START_SIZE;
	}

	flags |= ZIO_VDEV_CHILD_FLAGS(pio);

	/*
	 * If we've decided to do a repair, the write is not speculative --
	 * even if the original read was.
	 */
	if (flags & ZIO_FLAG_IO_REPAIR)
		flags &= ~ZIO_FLAG_SPECULATIVE;

	/*
	 * If we're creating a child I/O that is not associated with a
	 * top-level vdev, then the child zio is not an allocating I/O.
	 * If this is a retried I/O then we ignore it since we will
	 * have already processed the original allocating I/O.
	 */
	if (flags & ZIO_FLAG_IO_ALLOCATING &&
	    (vd != vd->vdev_top || (flags & ZIO_FLAG_IO_RETRY))) {
		ASSERT(pio->io_metaslab_class != NULL);
		ASSERT(pio->io_metaslab_class->mc_alloc_throttle_enabled);
		ASSERT(type == ZIO_TYPE_WRITE);
		ASSERT(priority == ZIO_PRIORITY_ASYNC_WRITE);
		ASSERT(!(flags & ZIO_FLAG_IO_REPAIR));
		ASSERT(!(pio->io_flags & ZIO_FLAG_IO_REWRITE) ||
		    pio->io_child_type == ZIO_CHILD_GANG);

		flags &= ~ZIO_FLAG_IO_ALLOCATING;
	}


	zio = zio_create(pio, pio->io_spa, pio->io_txg, bp, data, size, size,
	    done, private, type, priority, flags, vd, offset, &pio->io_bookmark,
	    ZIO_STAGE_VDEV_IO_START >> 1, pipeline);
	ASSERT3U(zio->io_child_type, ==, ZIO_CHILD_VDEV);

	zio->io_physdone = pio->io_physdone;
	if (vd->vdev_ops->vdev_op_leaf && zio->io_logical != NULL)
		zio->io_logical->io_phys_children++;

	return (zio);
}

zio_t *
zio_vdev_delegated_io(vdev_t *vd, uint64_t offset, abd_t *data, uint64_t size,
    zio_type_t type, zio_priority_t priority, enum zio_flag flags,
    zio_done_func_t *done, void *private)
{
	zio_t *zio;

	ASSERT(vd->vdev_ops->vdev_op_leaf);

	zio = zio_create(NULL, vd->vdev_spa, 0, NULL,
	    data, size, size, done, private, type, priority,
	    flags | ZIO_FLAG_CANFAIL | ZIO_FLAG_DONT_RETRY | ZIO_FLAG_DELEGATED,
	    vd, offset, NULL,
	    ZIO_STAGE_VDEV_IO_START >> 1, ZIO_VDEV_CHILD_PIPELINE);

	return (zio);
}

void
zio_flush(zio_t *zio, vdev_t *vd)
{
	zio_nowait(zio_ioctl(zio, zio->io_spa, vd, DKIOCFLUSHWRITECACHE,
	    NULL, NULL,
	    ZIO_FLAG_CANFAIL | ZIO_FLAG_DONT_PROPAGATE | ZIO_FLAG_DONT_RETRY));
}

void
zio_shrink(zio_t *zio, uint64_t size)
{
	ASSERT3P(zio->io_executor, ==, NULL);
	ASSERT3U(zio->io_orig_size, ==, zio->io_size);
	ASSERT3U(size, <=, zio->io_size);

	/*
	 * We don't shrink for raidz because of problems with the
	 * reconstruction when reading back less than the block size.
	 * Note, BP_IS_RAIDZ() assumes no compression.
	 */
	ASSERT(BP_GET_COMPRESS(zio->io_bp) == ZIO_COMPRESS_OFF);
	if (!BP_IS_RAIDZ(zio->io_bp)) {
		/* we are not doing a raw write */
		ASSERT3U(zio->io_size, ==, zio->io_lsize);
		zio->io_orig_size = zio->io_size = zio->io_lsize = size;
	}
}

/*
 * ==========================================================================
 * Prepare to read and write logical blocks
 * ==========================================================================
 */

static zio_t *
zio_read_bp_init(zio_t *zio)
{
	blkptr_t *bp = zio->io_bp;
	uint64_t psize =
	    BP_IS_EMBEDDED(bp) ? BPE_GET_PSIZE(bp) : BP_GET_PSIZE(bp);

	ASSERT3P(zio->io_bp, ==, &zio->io_bp_copy);

	if (BP_GET_COMPRESS(bp) != ZIO_COMPRESS_OFF &&
	    zio->io_child_type == ZIO_CHILD_LOGICAL &&
	    !(zio->io_flags & ZIO_FLAG_RAW_COMPRESS)) {
		zio_push_transform(zio, abd_alloc_sametype(zio->io_abd, psize),
		    psize, psize, zio_decompress);
	}

	if (((BP_IS_PROTECTED(bp) && !(zio->io_flags & ZIO_FLAG_RAW_ENCRYPT)) ||
	    BP_HAS_INDIRECT_MAC_CKSUM(bp)) &&
	    zio->io_child_type == ZIO_CHILD_LOGICAL) {
		zio_push_transform(zio, abd_alloc_sametype(zio->io_abd, psize),
		    psize, psize, zio_decrypt);
	}

	if (BP_IS_EMBEDDED(bp) && BPE_GET_ETYPE(bp) == BP_EMBEDDED_TYPE_DATA) {
		int psize = BPE_GET_PSIZE(bp);
		void *data = abd_borrow_buf(zio->io_abd, psize);

		zio->io_pipeline = ZIO_INTERLOCK_PIPELINE;
		decode_embedded_bp_compressed(bp, data);
		abd_return_buf_copy(zio->io_abd, data, psize);
	} else {
		ASSERT(!BP_IS_EMBEDDED(bp));
		ASSERT3P(zio->io_bp, ==, &zio->io_bp_copy);
	}

	if (!DMU_OT_IS_METADATA(BP_GET_TYPE(bp)) && BP_GET_LEVEL(bp) == 0)
		zio->io_flags |= ZIO_FLAG_DONT_CACHE;

	if (BP_GET_TYPE(bp) == DMU_OT_DDT_ZAP)
		zio->io_flags |= ZIO_FLAG_DONT_CACHE;

	if (BP_GET_DEDUP(bp) && zio->io_child_type == ZIO_CHILD_LOGICAL)
		zio->io_pipeline = ZIO_DDT_READ_PIPELINE;

	return (zio);
}

static zio_t *
zio_write_bp_init(zio_t *zio)
{
	if (!IO_IS_ALLOCATING(zio))
		return (zio);

	ASSERT(zio->io_child_type != ZIO_CHILD_DDT);

	if (zio->io_bp_override) {
		blkptr_t *bp = zio->io_bp;
		zio_prop_t *zp = &zio->io_prop;

		ASSERT(bp->blk_birth != zio->io_txg);
		ASSERT(BP_GET_DEDUP(zio->io_bp_override) == 0);

		*bp = *zio->io_bp_override;
		zio->io_pipeline = ZIO_INTERLOCK_PIPELINE;

		if (BP_IS_EMBEDDED(bp))
			return (zio);

		/*
		 * If we've been overridden and nopwrite is set then
		 * set the flag accordingly to indicate that a nopwrite
		 * has already occurred.
		 */
		if (!BP_IS_HOLE(bp) && zp->zp_nopwrite) {
			ASSERT(!zp->zp_dedup);
			ASSERT3U(BP_GET_CHECKSUM(bp), ==, zp->zp_checksum);
			zio->io_flags |= ZIO_FLAG_NOPWRITE;
			return (zio);
		}

		ASSERT(!zp->zp_nopwrite);

		if (BP_IS_HOLE(bp) || !zp->zp_dedup)
			return (zio);

		ASSERT((zio_checksum_table[zp->zp_checksum].ci_flags &
		    ZCHECKSUM_FLAG_DEDUP) || zp->zp_dedup_verify);

		if (BP_GET_CHECKSUM(bp) == zp->zp_checksum &&
		    !zp->zp_encrypt) {
			BP_SET_DEDUP(bp, 1);
			zio->io_pipeline |= ZIO_STAGE_DDT_WRITE;
			return (zio);
		}

		/*
		 * We were unable to handle this as an override bp, treat
		 * it as a regular write I/O.
		 */
		zio->io_bp_override = NULL;
		*bp = zio->io_bp_orig;
		zio->io_pipeline = zio->io_orig_pipeline;
	}

	return (zio);
}

static zio_t *
zio_write_compress(zio_t *zio)
{
	spa_t *spa = zio->io_spa;
	zio_prop_t *zp = &zio->io_prop;
	enum zio_compress compress = zp->zp_compress;
	blkptr_t *bp = zio->io_bp;
	uint64_t lsize = zio->io_lsize;
	uint64_t psize = zio->io_size;
	int pass = 1;

	/*
	 * If our children haven't all reached the ready stage,
	 * wait for them and then repeat this pipeline stage.
	 */
	if (zio_wait_for_children(zio, ZIO_CHILD_LOGICAL_BIT |
	    ZIO_CHILD_GANG_BIT, ZIO_WAIT_READY)) {
		return (NULL);
	}

	if (!IO_IS_ALLOCATING(zio))
		return (zio);

	if (zio->io_children_ready != NULL) {
		/*
		 * Now that all our children are ready, run the callback
		 * associated with this zio in case it wants to modify the
		 * data to be written.
		 */
		ASSERT3U(zp->zp_level, >, 0);
		zio->io_children_ready(zio);
	}

	ASSERT(zio->io_child_type != ZIO_CHILD_DDT);
	ASSERT(zio->io_bp_override == NULL);

	if (!BP_IS_HOLE(bp) && bp->blk_birth == zio->io_txg) {
		/*
		 * We're rewriting an existing block, which means we're
		 * working on behalf of spa_sync().  For spa_sync() to
		 * converge, it must eventually be the case that we don't
		 * have to allocate new blocks.  But compression changes
		 * the blocksize, which forces a reallocate, and makes
		 * convergence take longer.  Therefore, after the first
		 * few passes, stop compressing to ensure convergence.
		 */
		pass = spa_sync_pass(spa);

		ASSERT(zio->io_txg == spa_syncing_txg(spa));
		ASSERT(zio->io_child_type == ZIO_CHILD_LOGICAL);
		ASSERT(!BP_GET_DEDUP(bp));

		if (pass >= zfs_sync_pass_dont_compress)
			compress = ZIO_COMPRESS_OFF;

		/* Make sure someone doesn't change their mind on overwrites */
		ASSERT(BP_IS_EMBEDDED(bp) || MIN(zp->zp_copies + BP_IS_GANG(bp),
		    spa_max_replication(spa)) == BP_GET_NDVAS(bp));
	}

	/* If it's a compressed write that is not raw, compress the buffer. */
	if (compress != ZIO_COMPRESS_OFF &&
	    !(zio->io_flags & ZIO_FLAG_RAW_COMPRESS)) {
		void *cbuf = zio_buf_alloc(lsize);
		psize = zio_compress_data(compress, zio->io_abd, cbuf, lsize);
		if (psize == 0 || psize == lsize) {
			compress = ZIO_COMPRESS_OFF;
			zio_buf_free(cbuf, lsize);
		} else if (!zp->zp_dedup && !zp->zp_encrypt &&
		    psize <= BPE_PAYLOAD_SIZE &&
		    zp->zp_level == 0 && !DMU_OT_HAS_FILL(zp->zp_type) &&
		    spa_feature_is_enabled(spa, SPA_FEATURE_EMBEDDED_DATA)) {
			encode_embedded_bp_compressed(bp,
			    cbuf, compress, lsize, psize);
			BPE_SET_ETYPE(bp, BP_EMBEDDED_TYPE_DATA);
			BP_SET_TYPE(bp, zio->io_prop.zp_type);
			BP_SET_LEVEL(bp, zio->io_prop.zp_level);
			zio_buf_free(cbuf, lsize);
			bp->blk_birth = zio->io_txg;
			zio->io_pipeline = ZIO_INTERLOCK_PIPELINE;
			ASSERT(spa_feature_is_active(spa,
			    SPA_FEATURE_EMBEDDED_DATA));
			return (zio);
		} else {
			/*
			 * Round up compressed size up to the ashift
			 * of the smallest-ashift device, and zero the tail.
			 * This ensures that the compressed size of the BP
			 * (and thus compressratio property) are correct,
			 * in that we charge for the padding used to fill out
			 * the last sector.
			 */
			ASSERT3U(spa->spa_min_ashift, >=, SPA_MINBLOCKSHIFT);
			size_t rounded = (size_t)P2ROUNDUP(psize,
			    1ULL << spa->spa_min_ashift);
			if (rounded >= lsize) {
				compress = ZIO_COMPRESS_OFF;
				zio_buf_free(cbuf, lsize);
				psize = lsize;
			} else {
				abd_t *cdata = abd_get_from_buf(cbuf, lsize);
				abd_take_ownership_of_buf(cdata, B_TRUE);
				abd_zero_off(cdata, psize, rounded - psize);
				psize = rounded;
				zio_push_transform(zio, cdata,
				    psize, lsize, NULL);
			}
		}

		/*
		 * We were unable to handle this as an override bp, treat
		 * it as a regular write I/O.
		 */
		zio->io_bp_override = NULL;
		*bp = zio->io_bp_orig;
		zio->io_pipeline = zio->io_orig_pipeline;

	} else if ((zio->io_flags & ZIO_FLAG_RAW_ENCRYPT) != 0 &&
	    zp->zp_type == DMU_OT_DNODE) {
		/*
		 * The DMU actually relies on the zio layer's compression
		 * to free metadnode blocks that have had all contained
		 * dnodes freed. As a result, even when doing a raw
		 * receive, we must check whether the block can be compressed
		 * to a hole.
		 */
		psize = zio_compress_data(ZIO_COMPRESS_EMPTY,
		    zio->io_abd, NULL, lsize);
		if (psize == 0)
			compress = ZIO_COMPRESS_OFF;
	} else {
		ASSERT3U(psize, !=, 0);
	}

	/*
	 * The final pass of spa_sync() must be all rewrites, but the first
	 * few passes offer a trade-off: allocating blocks defers convergence,
	 * but newly allocated blocks are sequential, so they can be written
	 * to disk faster.  Therefore, we allow the first few passes of
	 * spa_sync() to allocate new blocks, but force rewrites after that.
	 * There should only be a handful of blocks after pass 1 in any case.
	 */
	if (!BP_IS_HOLE(bp) && bp->blk_birth == zio->io_txg &&
	    BP_GET_PSIZE(bp) == psize &&
	    pass >= zfs_sync_pass_rewrite) {
		VERIFY3U(psize, !=, 0);
		enum zio_stage gang_stages = zio->io_pipeline & ZIO_GANG_STAGES;

		zio->io_pipeline = ZIO_REWRITE_PIPELINE | gang_stages;
		zio->io_flags |= ZIO_FLAG_IO_REWRITE;
	} else {
		BP_ZERO(bp);
		zio->io_pipeline = ZIO_WRITE_PIPELINE;
	}

	if (psize == 0) {
		if (zio->io_bp_orig.blk_birth != 0 &&
		    spa_feature_is_active(spa, SPA_FEATURE_HOLE_BIRTH)) {
			BP_SET_LSIZE(bp, lsize);
			BP_SET_TYPE(bp, zp->zp_type);
			BP_SET_LEVEL(bp, zp->zp_level);
			BP_SET_BIRTH(bp, zio->io_txg, 0);
		}
		zio->io_pipeline = ZIO_INTERLOCK_PIPELINE;
	} else {
		ASSERT(zp->zp_checksum != ZIO_CHECKSUM_GANG_HEADER);
		BP_SET_LSIZE(bp, lsize);
		BP_SET_TYPE(bp, zp->zp_type);
		BP_SET_LEVEL(bp, zp->zp_level);
		BP_SET_PSIZE(bp, psize);
		BP_SET_COMPRESS(bp, compress);
		BP_SET_CHECKSUM(bp, zp->zp_checksum);
		BP_SET_DEDUP(bp, zp->zp_dedup);
		BP_SET_BYTEORDER(bp, ZFS_HOST_BYTEORDER);
		if (zp->zp_dedup) {
			ASSERT(zio->io_child_type == ZIO_CHILD_LOGICAL);
			ASSERT(!(zio->io_flags & ZIO_FLAG_IO_REWRITE));
			ASSERT(!zp->zp_encrypt ||
			    DMU_OT_IS_ENCRYPTED(zp->zp_type));
			zio->io_pipeline = ZIO_DDT_WRITE_PIPELINE;
		}
		if (zp->zp_nopwrite) {
			ASSERT(zio->io_child_type == ZIO_CHILD_LOGICAL);
			ASSERT(!(zio->io_flags & ZIO_FLAG_IO_REWRITE));
			zio->io_pipeline |= ZIO_STAGE_NOP_WRITE;
		}
	}
	return (zio);
}

static zio_t *
zio_free_bp_init(zio_t *zio)
{
	blkptr_t *bp = zio->io_bp;

	if (zio->io_child_type == ZIO_CHILD_LOGICAL) {
		if (BP_GET_DEDUP(bp))
			zio->io_pipeline = ZIO_DDT_FREE_PIPELINE;
	}

	ASSERT3P(zio->io_bp, ==, &zio->io_bp_copy);

	return (zio);
}

/*
 * ==========================================================================
 * Execute the I/O pipeline
 * ==========================================================================
 */

static void
zio_taskq_dispatch(zio_t *zio, zio_taskq_type_t q, boolean_t cutinline)
{
	spa_t *spa = zio->io_spa;
	zio_type_t t = zio->io_type;
	int flags = (cutinline ? TQ_FRONT : 0);

	/*
	 * If we're a config writer or a probe, the normal issue and
	 * interrupt threads may all be blocked waiting for the config lock.
	 * In this case, select the otherwise-unused taskq for ZIO_TYPE_NULL.
	 */
	if (zio->io_flags & (ZIO_FLAG_CONFIG_WRITER | ZIO_FLAG_PROBE))
		t = ZIO_TYPE_NULL;

	/*
	 * A similar issue exists for the L2ARC write thread until L2ARC 2.0.
	 */
	if (t == ZIO_TYPE_WRITE && zio->io_vd && zio->io_vd->vdev_aux)
		t = ZIO_TYPE_NULL;

	/*
	 * If this is a high priority I/O, then use the high priority taskq if
	 * available.
	 */
	if ((zio->io_priority == ZIO_PRIORITY_NOW ||
	    zio->io_priority == ZIO_PRIORITY_SYNC_WRITE) &&
	    spa->spa_zio_taskq[t][q + 1].stqs_count != 0)
		q++;

	ASSERT3U(q, <, ZIO_TASKQ_TYPES);

	/*
	 * NB: We are assuming that the zio can only be dispatched
	 * to a single taskq at a time.  It would be a grievous error
	 * to dispatch the zio to another taskq at the same time.
	 */
	ASSERT(taskq_empty_ent(&zio->io_tqent));
	spa_taskq_dispatch_ent(spa, t, q, (task_func_t *)zio_execute, zio,
	    flags, &zio->io_tqent);
}

static boolean_t
zio_taskq_member(zio_t *zio, zio_taskq_type_t q)
{
	kthread_t *executor = zio->io_executor;
	spa_t *spa = zio->io_spa;

	for (zio_type_t t = 0; t < ZIO_TYPES; t++) {
		spa_taskqs_t *tqs = &spa->spa_zio_taskq[t][q];
		uint_t i;
		for (i = 0; i < tqs->stqs_count; i++) {
			if (taskq_member(tqs->stqs_taskq[i], executor))
				return (B_TRUE);
		}
	}

	return (B_FALSE);
}

static zio_t *
zio_issue_async(zio_t *zio)
{
	zio_taskq_dispatch(zio, ZIO_TASKQ_ISSUE, B_FALSE);

	return (NULL);
}

void
zio_interrupt(zio_t *zio)
{
	zio_taskq_dispatch(zio, ZIO_TASKQ_INTERRUPT, B_FALSE);
}

void
zio_delay_interrupt(zio_t *zio)
{
	/*
	 * The timeout_generic() function isn't defined in userspace, so
	 * rather than trying to implement the function, the zio delay
	 * functionality has been disabled for userspace builds.
	 */

#ifdef _KERNEL
	/*
	 * If io_target_timestamp is zero, then no delay has been registered
	 * for this IO, thus jump to the end of this function and "skip" the
	 * delay; issuing it directly to the zio layer.
	 */
	if (zio->io_target_timestamp != 0) {
		hrtime_t now = gethrtime();

		if (now >= zio->io_target_timestamp) {
			/*
			 * This IO has already taken longer than the target
			 * delay to complete, so we don't want to delay it
			 * any longer; we "miss" the delay and issue it
			 * directly to the zio layer. This is likely due to
			 * the target latency being set to a value less than
			 * the underlying hardware can satisfy (e.g. delay
			 * set to 1ms, but the disks take 10ms to complete an
			 * IO request).
			 */

			DTRACE_PROBE2(zio__delay__miss, zio_t *, zio,
			    hrtime_t, now);

			zio_interrupt(zio);
		} else {
			taskqid_t tid;
			hrtime_t diff = zio->io_target_timestamp - now;
			clock_t expire_at_tick = ddi_get_lbolt() +
			    NSEC_TO_TICK(diff);

			DTRACE_PROBE3(zio__delay__hit, zio_t *, zio,
			    hrtime_t, now, hrtime_t, diff);

			if (NSEC_TO_TICK(diff) == 0) {
				/* Our delay is less than a jiffy - just spin */
				zfs_sleep_until(zio->io_target_timestamp);
				zio_interrupt(zio);
			} else {
				/*
				 * Use taskq_dispatch_delay() in the place of
				 * OpenZFS's timeout_generic().
				 */
				tid = taskq_dispatch_delay(system_taskq,
				    (task_func_t *)zio_interrupt,
				    zio, TQ_NOSLEEP, expire_at_tick);
				if (tid == TASKQID_INVALID) {
					/*
					 * Couldn't allocate a task.  Just
					 * finish the zio without a delay.
					 */
					zio_interrupt(zio);
				}
			}
		}
		return;
	}
#endif
	DTRACE_PROBE1(zio__delay__skip, zio_t *, zio);
	zio_interrupt(zio);
}

static void
zio_deadman_impl(zio_t *pio, int ziodepth)
{
	zio_t *cio, *cio_next;
	zio_link_t *zl = NULL;
	vdev_t *vd = pio->io_vd;

	if (zio_deadman_log_all || (vd != NULL && vd->vdev_ops->vdev_op_leaf)) {
		vdev_queue_t *vq = vd ? &vd->vdev_queue : NULL;
		zbookmark_phys_t *zb = &pio->io_bookmark;
		uint64_t delta = gethrtime() - pio->io_timestamp;
		uint64_t failmode = spa_get_deadman_failmode(pio->io_spa);

		zfs_dbgmsg("slow zio[%d]: zio=%px timestamp=%llu "
		    "delta=%llu queued=%llu io=%llu "
		    "path=%s last=%llu "
		    "type=%d priority=%d flags=0x%x "
		    "stage=0x%x pipeline=0x%x pipeline-trace=0x%x "
		    "objset=%llu object=%llu level=%llu blkid=%llu "
		    "offset=%llu size=%llu error=%d",
		    ziodepth, pio, pio->io_timestamp,
		    delta, pio->io_delta, pio->io_delay,
		    vd ? vd->vdev_path : "NULL", vq ? vq->vq_io_complete_ts : 0,
		    pio->io_type, pio->io_priority, pio->io_flags,
		    pio->io_stage, pio->io_pipeline, pio->io_pipeline_trace,
		    zb->zb_objset, zb->zb_object, zb->zb_level, zb->zb_blkid,
		    pio->io_offset, pio->io_size, pio->io_error);
		zfs_ereport_post(FM_EREPORT_ZFS_DEADMAN,
		    pio->io_spa, vd, zb, pio, 0, 0);

		if (failmode == ZIO_FAILURE_MODE_CONTINUE &&
		    taskq_empty_ent(&pio->io_tqent)) {
			zio_interrupt(pio);
		}
	}

	mutex_enter(&pio->io_lock);
	for (cio = zio_walk_children(pio, &zl); cio != NULL; cio = cio_next) {
		cio_next = zio_walk_children(pio, &zl);
		zio_deadman_impl(cio, ziodepth + 1);
	}
	mutex_exit(&pio->io_lock);
}

/*
 * Log the critical information describing this zio and all of its children
 * using the zfs_dbgmsg() interface then post deadman event for the ZED.
 */
void
zio_deadman(zio_t *pio, char *tag)
{
	spa_t *spa = pio->io_spa;
	char *name = spa_name(spa);

	if (!zfs_deadman_enabled || spa_suspended(spa))
		return;

	zio_deadman_impl(pio, 0);

	switch (spa_get_deadman_failmode(spa)) {
	case ZIO_FAILURE_MODE_WAIT:
		zfs_dbgmsg("%s waiting for hung I/O to pool '%s'", tag, name);
		break;

	case ZIO_FAILURE_MODE_CONTINUE:
		zfs_dbgmsg("%s restarting hung I/O for pool '%s'", tag, name);
		break;

	case ZIO_FAILURE_MODE_PANIC:
		fm_panic("%s determined I/O to pool '%s' is hung.", tag, name);
		break;
	}
}

/*
 * Execute the I/O pipeline until one of the following occurs:
 * (1) the I/O completes; (2) the pipeline stalls waiting for
 * dependent child I/Os; (3) the I/O issues, so we're waiting
 * for an I/O completion interrupt; (4) the I/O is delegated by
 * vdev-level caching or aggregation; (5) the I/O is deferred
 * due to vdev-level queueing; (6) the I/O is handed off to
 * another thread.  In all cases, the pipeline stops whenever
 * there's no CPU work; it never burns a thread in cv_wait_io().
 *
 * There's no locking on io_stage because there's no legitimate way
 * for multiple threads to be attempting to process the same I/O.
 */
static zio_pipe_stage_t *zio_pipeline[];

/*
 * zio_execute() is a wrapper around the static function
 * __zio_execute() so that we can force  __zio_execute() to be
 * inlined.  This reduces stack overhead which is important
 * because __zio_execute() is called recursively in several zio
 * code paths.  zio_execute() itself cannot be inlined because
 * it is externally visible.
 */
void
zio_execute(zio_t *zio)
{
	fstrans_cookie_t cookie;

	cookie = spl_fstrans_mark();
	__zio_execute(zio);
	spl_fstrans_unmark(cookie);
}

/*
 * Used to determine if in the current context the stack is sized large
 * enough to allow zio_execute() to be called recursively.  A minimum
 * stack size of 16K is required to avoid needing to re-dispatch the zio.
 */
boolean_t
zio_execute_stack_check(zio_t *zio)
{
#if !defined(HAVE_LARGE_STACKS)
	dsl_pool_t *dp = spa_get_dsl(zio->io_spa);

	/* Executing in txg_sync_thread() context. */
	if (dp && curthread == dp->dp_tx.tx_sync_thread)
		return (B_TRUE);

	/* Pool initialization outside of zio_taskq context. */
	if (dp && spa_is_initializing(dp->dp_spa) &&
	    !zio_taskq_member(zio, ZIO_TASKQ_ISSUE) &&
	    !zio_taskq_member(zio, ZIO_TASKQ_ISSUE_HIGH))
		return (B_TRUE);
#endif /* HAVE_LARGE_STACKS */

	return (B_FALSE);
}

__attribute__((always_inline))
static inline void
__zio_execute(zio_t *zio)
{
	ASSERT3U(zio->io_queued_timestamp, >, 0);

	while (zio->io_stage < ZIO_STAGE_DONE) {
		enum zio_stage pipeline = zio->io_pipeline;
		enum zio_stage stage = zio->io_stage;

		zio->io_executor = curthread;

		ASSERT(!MUTEX_HELD(&zio->io_lock));
		ASSERT(ISP2(stage));
		ASSERT(zio->io_stall == NULL);

		do {
			stage <<= 1;
		} while ((stage & pipeline) == 0);

		ASSERT(stage <= ZIO_STAGE_DONE);

		/*
		 * If we are in interrupt context and this pipeline stage
		 * will grab a config lock that is held across I/O,
		 * or may wait for an I/O that needs an interrupt thread
		 * to complete, issue async to avoid deadlock.
		 *
		 * For VDEV_IO_START, we cut in line so that the io will
		 * be sent to disk promptly.
		 */
		if ((stage & ZIO_BLOCKING_STAGES) && zio->io_vd == NULL &&
		    zio_taskq_member(zio, ZIO_TASKQ_INTERRUPT)) {
			boolean_t cut = (stage == ZIO_STAGE_VDEV_IO_START) ?
			    zio_requeue_io_start_cut_in_line : B_FALSE;
			zio_taskq_dispatch(zio, ZIO_TASKQ_ISSUE, cut);
			return;
		}

		/*
		 * If the current context doesn't have large enough stacks
		 * the zio must be issued asynchronously to prevent overflow.
		 */
		if (zio_execute_stack_check(zio)) {
			boolean_t cut = (stage == ZIO_STAGE_VDEV_IO_START) ?
			    zio_requeue_io_start_cut_in_line : B_FALSE;
			zio_taskq_dispatch(zio, ZIO_TASKQ_ISSUE, cut);
			return;
		}

		zio->io_stage = stage;
		zio->io_pipeline_trace |= zio->io_stage;

		/*
		 * The zio pipeline stage returns the next zio to execute
		 * (typically the same as this one), or NULL if we should
		 * stop.
		 */
		zio = zio_pipeline[highbit64(stage) - 1](zio);

		if (zio == NULL)
			return;
	}
}


/*
 * ==========================================================================
 * Initiate I/O, either sync or async
 * ==========================================================================
 */
int
zio_wait(zio_t *zio)
{
	long timeout = MSEC_TO_TICK(zfs_deadman_ziotime_ms);
	int error;

	ASSERT3S(zio->io_stage, ==, ZIO_STAGE_OPEN);
	ASSERT3P(zio->io_executor, ==, NULL);

	zio->io_waiter = curthread;
	ASSERT0(zio->io_queued_timestamp);
	zio->io_queued_timestamp = gethrtime();

	__zio_execute(zio);

	mutex_enter(&zio->io_lock);
	while (zio->io_executor != NULL) {
		error = cv_timedwait_io(&zio->io_cv, &zio->io_lock,
		    ddi_get_lbolt() + timeout);

		if (zfs_deadman_enabled && error == -1 &&
		    gethrtime() - zio->io_queued_timestamp >
		    spa_deadman_ziotime(zio->io_spa)) {
			mutex_exit(&zio->io_lock);
			timeout = MSEC_TO_TICK(zfs_deadman_checktime_ms);
			zio_deadman(zio, FTAG);
			mutex_enter(&zio->io_lock);
		}
	}
	mutex_exit(&zio->io_lock);

	error = zio->io_error;
	zio_destroy(zio);

	return (error);
}

void
zio_nowait(zio_t *zio)
{
	ASSERT3P(zio->io_executor, ==, NULL);

	if (zio->io_child_type == ZIO_CHILD_LOGICAL &&
	    zio_unique_parent(zio) == NULL) {
		zio_t *pio;

		/*
		 * This is a logical async I/O with no parent to wait for it.
		 * We add it to the spa_async_root_zio "Godfather" I/O which
		 * will ensure they complete prior to unloading the pool.
		 */
		spa_t *spa = zio->io_spa;
		kpreempt_disable();
		pio = spa->spa_async_zio_root[CPU_SEQID];
		kpreempt_enable();

		zio_add_child(pio, zio);
	}

	ASSERT0(zio->io_queued_timestamp);
	zio->io_queued_timestamp = gethrtime();
	__zio_execute(zio);
}

/*
 * ==========================================================================
 * Reexecute, cancel, or suspend/resume failed I/O
 * ==========================================================================
 */

static void
zio_reexecute(zio_t *pio)
{
	zio_t *cio, *cio_next;

	ASSERT(pio->io_child_type == ZIO_CHILD_LOGICAL);
	ASSERT(pio->io_orig_stage == ZIO_STAGE_OPEN);
	ASSERT(pio->io_gang_leader == NULL);
	ASSERT(pio->io_gang_tree == NULL);

	pio->io_flags = pio->io_orig_flags;
	pio->io_stage = pio->io_orig_stage;
	pio->io_pipeline = pio->io_orig_pipeline;
	pio->io_reexecute = 0;
	pio->io_flags |= ZIO_FLAG_REEXECUTED;
	pio->io_pipeline_trace = 0;
	pio->io_error = 0;
	for (int w = 0; w < ZIO_WAIT_TYPES; w++)
		pio->io_state[w] = 0;
	for (int c = 0; c < ZIO_CHILD_TYPES; c++)
		pio->io_child_error[c] = 0;

	if (IO_IS_ALLOCATING(pio))
		BP_ZERO(pio->io_bp);

	/*
	 * As we reexecute pio's children, new children could be created.
	 * New children go to the head of pio's io_child_list, however,
	 * so we will (correctly) not reexecute them.  The key is that
	 * the remainder of pio's io_child_list, from 'cio_next' onward,
	 * cannot be affected by any side effects of reexecuting 'cio'.
	 */
	zio_link_t *zl = NULL;
	mutex_enter(&pio->io_lock);
	for (cio = zio_walk_children(pio, &zl); cio != NULL; cio = cio_next) {
		cio_next = zio_walk_children(pio, &zl);
		for (int w = 0; w < ZIO_WAIT_TYPES; w++)
			pio->io_children[cio->io_child_type][w]++;
		mutex_exit(&pio->io_lock);
		zio_reexecute(cio);
		mutex_enter(&pio->io_lock);
	}
	mutex_exit(&pio->io_lock);

	/*
	 * Now that all children have been reexecuted, execute the parent.
	 * We don't reexecute "The Godfather" I/O here as it's the
	 * responsibility of the caller to wait on it.
	 */
	if (!(pio->io_flags & ZIO_FLAG_GODFATHER)) {
		pio->io_queued_timestamp = gethrtime();
		__zio_execute(pio);
	}
}

void
zio_suspend(spa_t *spa, zio_t *zio, zio_suspend_reason_t reason)
{
	if (spa_get_failmode(spa) == ZIO_FAILURE_MODE_PANIC)
		fm_panic("Pool '%s' has encountered an uncorrectable I/O "
		    "failure and the failure mode property for this pool "
		    "is set to panic.", spa_name(spa));

	cmn_err(CE_WARN, "Pool '%s' has encountered an uncorrectable I/O "
	    "failure and has been suspended.\n", spa_name(spa));

	zfs_ereport_post(FM_EREPORT_ZFS_IO_FAILURE, spa, NULL,
	    NULL, NULL, 0, 0);

	mutex_enter(&spa->spa_suspend_lock);

	if (spa->spa_suspend_zio_root == NULL)
		spa->spa_suspend_zio_root = zio_root(spa, NULL, NULL,
		    ZIO_FLAG_CANFAIL | ZIO_FLAG_SPECULATIVE |
		    ZIO_FLAG_GODFATHER);

	spa->spa_suspended = reason;

	if (zio != NULL) {
		ASSERT(!(zio->io_flags & ZIO_FLAG_GODFATHER));
		ASSERT(zio != spa->spa_suspend_zio_root);
		ASSERT(zio->io_child_type == ZIO_CHILD_LOGICAL);
		ASSERT(zio_unique_parent(zio) == NULL);
		ASSERT(zio->io_stage == ZIO_STAGE_DONE);
		zio_add_child(spa->spa_suspend_zio_root, zio);
	}

	mutex_exit(&spa->spa_suspend_lock);
}

int
zio_resume(spa_t *spa)
{
	zio_t *pio;

	/*
	 * Reexecute all previously suspended i/o.
	 */
	mutex_enter(&spa->spa_suspend_lock);
	spa->spa_suspended = ZIO_SUSPEND_NONE;
	cv_broadcast(&spa->spa_suspend_cv);
	pio = spa->spa_suspend_zio_root;
	spa->spa_suspend_zio_root = NULL;
	mutex_exit(&spa->spa_suspend_lock);

	if (pio == NULL)
		return (0);

	zio_reexecute(pio);
	return (zio_wait(pio));
}

void
zio_resume_wait(spa_t *spa)
{
	mutex_enter(&spa->spa_suspend_lock);
	while (spa_suspended(spa))
		cv_wait(&spa->spa_suspend_cv, &spa->spa_suspend_lock);
	mutex_exit(&spa->spa_suspend_lock);
}

/*
 * ==========================================================================
 * Gang blocks.
 *
 * A gang block is a collection of small blocks that looks to the DMU
 * like one large block.  When zio_dva_allocate() cannot find a block
 * of the requested size, due to either severe fragmentation or the pool
 * being nearly full, it calls zio_write_gang_block() to construct the
 * block from smaller fragments.
 *
 * A gang block consists of a gang header (zio_gbh_phys_t) and up to
 * three (SPA_GBH_NBLKPTRS) gang members.  The gang header is just like
 * an indirect block: it's an array of block pointers.  It consumes
 * only one sector and hence is allocatable regardless of fragmentation.
 * The gang header's bps point to its gang members, which hold the data.
 *
 * Gang blocks are self-checksumming, using the bp's <vdev, offset, txg>
 * as the verifier to ensure uniqueness of the SHA256 checksum.
 * Critically, the gang block bp's blk_cksum is the checksum of the data,
 * not the gang header.  This ensures that data block signatures (needed for
 * deduplication) are independent of how the block is physically stored.
 *
 * Gang blocks can be nested: a gang member may itself be a gang block.
 * Thus every gang block is a tree in which root and all interior nodes are
 * gang headers, and the leaves are normal blocks that contain user data.
 * The root of the gang tree is called the gang leader.
 *
 * To perform any operation (read, rewrite, free, claim) on a gang block,
 * zio_gang_assemble() first assembles the gang tree (minus data leaves)
 * in the io_gang_tree field of the original logical i/o by recursively
 * reading the gang leader and all gang headers below it.  This yields
 * an in-core tree containing the contents of every gang header and the
 * bps for every constituent of the gang block.
 *
 * With the gang tree now assembled, zio_gang_issue() just walks the gang tree
 * and invokes a callback on each bp.  To free a gang block, zio_gang_issue()
 * calls zio_free_gang() -- a trivial wrapper around zio_free() -- for each bp.
 * zio_claim_gang() provides a similarly trivial wrapper for zio_claim().
 * zio_read_gang() is a wrapper around zio_read() that omits reading gang
 * headers, since we already have those in io_gang_tree.  zio_rewrite_gang()
 * performs a zio_rewrite() of the data or, for gang headers, a zio_rewrite()
 * of the gang header plus zio_checksum_compute() of the data to update the
 * gang header's blk_cksum as described above.
 *
 * The two-phase assemble/issue model solves the problem of partial failure --
 * what if you'd freed part of a gang block but then couldn't read the
 * gang header for another part?  Assembling the entire gang tree first
 * ensures that all the necessary gang header I/O has succeeded before
 * starting the actual work of free, claim, or write.  Once the gang tree
 * is assembled, free and claim are in-memory operations that cannot fail.
 *
 * In the event that a gang write fails, zio_dva_unallocate() walks the
 * gang tree to immediately free (i.e. insert back into the space map)
 * everything we've allocated.  This ensures that we don't get ENOSPC
 * errors during repeated suspend/resume cycles due to a flaky device.
 *
 * Gang rewrites only happen during sync-to-convergence.  If we can't assemble
 * the gang tree, we won't modify the block, so we can safely defer the free
 * (knowing that the block is still intact).  If we *can* assemble the gang
 * tree, then even if some of the rewrites fail, zio_dva_unallocate() will free
 * each constituent bp and we can allocate a new block on the next sync pass.
 *
 * In all cases, the gang tree allows complete recovery from partial failure.
 * ==========================================================================
 */

static void
zio_gang_issue_func_done(zio_t *zio)
{
	abd_put(zio->io_abd);
}

static zio_t *
zio_read_gang(zio_t *pio, blkptr_t *bp, zio_gang_node_t *gn, abd_t *data,
    uint64_t offset)
{
	if (gn != NULL)
		return (pio);

	return (zio_read(pio, pio->io_spa, bp, abd_get_offset(data, offset),
	    BP_GET_PSIZE(bp), zio_gang_issue_func_done,
	    NULL, pio->io_priority, ZIO_GANG_CHILD_FLAGS(pio),
	    &pio->io_bookmark));
}

static zio_t *
zio_rewrite_gang(zio_t *pio, blkptr_t *bp, zio_gang_node_t *gn, abd_t *data,
    uint64_t offset)
{
	zio_t *zio;

	if (gn != NULL) {
		abd_t *gbh_abd =
		    abd_get_from_buf(gn->gn_gbh, SPA_GANGBLOCKSIZE);
		zio = zio_rewrite(pio, pio->io_spa, pio->io_txg, bp,
		    gbh_abd, SPA_GANGBLOCKSIZE, zio_gang_issue_func_done, NULL,
		    pio->io_priority, ZIO_GANG_CHILD_FLAGS(pio),
		    &pio->io_bookmark);
		/*
		 * As we rewrite each gang header, the pipeline will compute
		 * a new gang block header checksum for it; but no one will
		 * compute a new data checksum, so we do that here.  The one
		 * exception is the gang leader: the pipeline already computed
		 * its data checksum because that stage precedes gang assembly.
		 * (Presently, nothing actually uses interior data checksums;
		 * this is just good hygiene.)
		 */
		if (gn != pio->io_gang_leader->io_gang_tree) {
			abd_t *buf = abd_get_offset(data, offset);

			zio_checksum_compute(zio, BP_GET_CHECKSUM(bp),
			    buf, BP_GET_PSIZE(bp));

			abd_put(buf);
		}
		/*
		 * If we are here to damage data for testing purposes,
		 * leave the GBH alone so that we can detect the damage.
		 */
		if (pio->io_gang_leader->io_flags & ZIO_FLAG_INDUCE_DAMAGE)
			zio->io_pipeline &= ~ZIO_VDEV_IO_STAGES;
	} else {
		zio = zio_rewrite(pio, pio->io_spa, pio->io_txg, bp,
		    abd_get_offset(data, offset), BP_GET_PSIZE(bp),
		    zio_gang_issue_func_done, NULL, pio->io_priority,
		    ZIO_GANG_CHILD_FLAGS(pio), &pio->io_bookmark);
	}

	return (zio);
}

/* ARGSUSED */
static zio_t *
zio_free_gang(zio_t *pio, blkptr_t *bp, zio_gang_node_t *gn, abd_t *data,
    uint64_t offset)
{
	return (zio_free_sync(pio, pio->io_spa, pio->io_txg, bp,
	    ZIO_GANG_CHILD_FLAGS(pio)));
}

/* ARGSUSED */
static zio_t *
zio_claim_gang(zio_t *pio, blkptr_t *bp, zio_gang_node_t *gn, abd_t *data,
    uint64_t offset)
{
	return (zio_claim(pio, pio->io_spa, pio->io_txg, bp,
	    NULL, NULL, ZIO_GANG_CHILD_FLAGS(pio)));
}

static zio_gang_issue_func_t *zio_gang_issue_func[ZIO_TYPES] = {
	NULL,
	zio_read_gang,
	zio_rewrite_gang,
	zio_free_gang,
	zio_claim_gang,
	NULL
};

static void zio_gang_tree_assemble_done(zio_t *zio);

static zio_gang_node_t *
zio_gang_node_alloc(zio_gang_node_t **gnpp)
{
	zio_gang_node_t *gn;

	ASSERT(*gnpp == NULL);

	gn = kmem_zalloc(sizeof (*gn), KM_SLEEP);
	gn->gn_gbh = zio_buf_alloc(SPA_GANGBLOCKSIZE);
	*gnpp = gn;

	return (gn);
}

static void
zio_gang_node_free(zio_gang_node_t **gnpp)
{
	zio_gang_node_t *gn = *gnpp;

	for (int g = 0; g < SPA_GBH_NBLKPTRS; g++)
		ASSERT(gn->gn_child[g] == NULL);

	zio_buf_free(gn->gn_gbh, SPA_GANGBLOCKSIZE);
	kmem_free(gn, sizeof (*gn));
	*gnpp = NULL;
}

static void
zio_gang_tree_free(zio_gang_node_t **gnpp)
{
	zio_gang_node_t *gn = *gnpp;

	if (gn == NULL)
		return;

	for (int g = 0; g < SPA_GBH_NBLKPTRS; g++)
		zio_gang_tree_free(&gn->gn_child[g]);

	zio_gang_node_free(gnpp);
}

static void
zio_gang_tree_assemble(zio_t *gio, blkptr_t *bp, zio_gang_node_t **gnpp)
{
	zio_gang_node_t *gn = zio_gang_node_alloc(gnpp);
	abd_t *gbh_abd = abd_get_from_buf(gn->gn_gbh, SPA_GANGBLOCKSIZE);

	ASSERT(gio->io_gang_leader == gio);
	ASSERT(BP_IS_GANG(bp));

	zio_nowait(zio_read(gio, gio->io_spa, bp, gbh_abd, SPA_GANGBLOCKSIZE,
	    zio_gang_tree_assemble_done, gn, gio->io_priority,
	    ZIO_GANG_CHILD_FLAGS(gio), &gio->io_bookmark));
}

static void
zio_gang_tree_assemble_done(zio_t *zio)
{
	zio_t *gio = zio->io_gang_leader;
	zio_gang_node_t *gn = zio->io_private;
	blkptr_t *bp = zio->io_bp;

	ASSERT(gio == zio_unique_parent(zio));
	ASSERT(zio->io_child_count == 0);

	if (zio->io_error)
		return;

	/* this ABD was created from a linear buf in zio_gang_tree_assemble */
	if (BP_SHOULD_BYTESWAP(bp))
		byteswap_uint64_array(abd_to_buf(zio->io_abd), zio->io_size);

	ASSERT3P(abd_to_buf(zio->io_abd), ==, gn->gn_gbh);
	ASSERT(zio->io_size == SPA_GANGBLOCKSIZE);
	ASSERT(gn->gn_gbh->zg_tail.zec_magic == ZEC_MAGIC);

	abd_put(zio->io_abd);

	for (int g = 0; g < SPA_GBH_NBLKPTRS; g++) {
		blkptr_t *gbp = &gn->gn_gbh->zg_blkptr[g];
		if (!BP_IS_GANG(gbp))
			continue;
		zio_gang_tree_assemble(gio, gbp, &gn->gn_child[g]);
	}
}

static void
zio_gang_tree_issue(zio_t *pio, zio_gang_node_t *gn, blkptr_t *bp, abd_t *data,
    uint64_t offset)
{
	zio_t *gio = pio->io_gang_leader;
	zio_t *zio;

	ASSERT(BP_IS_GANG(bp) == !!gn);
	ASSERT(BP_GET_CHECKSUM(bp) == BP_GET_CHECKSUM(gio->io_bp));
	ASSERT(BP_GET_LSIZE(bp) == BP_GET_PSIZE(bp) || gn == gio->io_gang_tree);

	/*
	 * If you're a gang header, your data is in gn->gn_gbh.
	 * If you're a gang member, your data is in 'data' and gn == NULL.
	 */
	zio = zio_gang_issue_func[gio->io_type](pio, bp, gn, data, offset);

	if (gn != NULL) {
		ASSERT(gn->gn_gbh->zg_tail.zec_magic == ZEC_MAGIC);

		for (int g = 0; g < SPA_GBH_NBLKPTRS; g++) {
			blkptr_t *gbp = &gn->gn_gbh->zg_blkptr[g];
			if (BP_IS_HOLE(gbp))
				continue;
			zio_gang_tree_issue(zio, gn->gn_child[g], gbp, data,
			    offset);
			offset += BP_GET_PSIZE(gbp);
		}
	}

	if (gn == gio->io_gang_tree)
		ASSERT3U(gio->io_size, ==, offset);

	if (zio != pio)
		zio_nowait(zio);
}

static zio_t *
zio_gang_assemble(zio_t *zio)
{
	blkptr_t *bp = zio->io_bp;

	ASSERT(BP_IS_GANG(bp) && zio->io_gang_leader == NULL);
	ASSERT(zio->io_child_type > ZIO_CHILD_GANG);

	zio->io_gang_leader = zio;

	zio_gang_tree_assemble(zio, bp, &zio->io_gang_tree);

	return (zio);
}

static zio_t *
zio_gang_issue(zio_t *zio)
{
	blkptr_t *bp = zio->io_bp;

	if (zio_wait_for_children(zio, ZIO_CHILD_GANG_BIT, ZIO_WAIT_DONE)) {
		return (NULL);
	}

	ASSERT(BP_IS_GANG(bp) && zio->io_gang_leader == zio);
	ASSERT(zio->io_child_type > ZIO_CHILD_GANG);

	if (zio->io_child_error[ZIO_CHILD_GANG] == 0)
		zio_gang_tree_issue(zio, zio->io_gang_tree, bp, zio->io_abd,
		    0);
	else
		zio_gang_tree_free(&zio->io_gang_tree);

	zio->io_pipeline = ZIO_INTERLOCK_PIPELINE;

	return (zio);
}

static void
zio_write_gang_member_ready(zio_t *zio)
{
	zio_t *pio = zio_unique_parent(zio);
	dva_t *cdva = zio->io_bp->blk_dva;
	dva_t *pdva = pio->io_bp->blk_dva;
	uint64_t asize;
	ASSERTV(zio_t *gio = zio->io_gang_leader);

	if (BP_IS_HOLE(zio->io_bp))
		return;

	ASSERT(BP_IS_HOLE(&zio->io_bp_orig));

	ASSERT(zio->io_child_type == ZIO_CHILD_GANG);
	ASSERT3U(zio->io_prop.zp_copies, ==, gio->io_prop.zp_copies);
	ASSERT3U(zio->io_prop.zp_copies, <=, BP_GET_NDVAS(zio->io_bp));
	ASSERT3U(pio->io_prop.zp_copies, <=, BP_GET_NDVAS(pio->io_bp));
	ASSERT3U(BP_GET_NDVAS(zio->io_bp), <=, BP_GET_NDVAS(pio->io_bp));

	mutex_enter(&pio->io_lock);
	for (int d = 0; d < BP_GET_NDVAS(zio->io_bp); d++) {
		ASSERT(DVA_GET_GANG(&pdva[d]));
		asize = DVA_GET_ASIZE(&pdva[d]);
		asize += DVA_GET_ASIZE(&cdva[d]);
		DVA_SET_ASIZE(&pdva[d], asize);
	}
	mutex_exit(&pio->io_lock);
}

static void
zio_write_gang_done(zio_t *zio)
{
	/*
	 * The io_abd field will be NULL for a zio with no data.  The io_flags
	 * will initially have the ZIO_FLAG_NODATA bit flag set, but we can't
	 * check for it here as it is cleared in zio_ready.
	 */
	if (zio->io_abd != NULL)
		abd_put(zio->io_abd);
}

static zio_t *
zio_write_gang_block(zio_t *pio)
{
	spa_t *spa = pio->io_spa;
	metaslab_class_t *mc = spa_normal_class(spa);
	blkptr_t *bp = pio->io_bp;
	zio_t *gio = pio->io_gang_leader;
	zio_t *zio;
	zio_gang_node_t *gn, **gnpp;
	zio_gbh_phys_t *gbh;
	abd_t *gbh_abd;
	uint64_t txg = pio->io_txg;
	uint64_t resid = pio->io_size;
	uint64_t lsize;
	int copies = gio->io_prop.zp_copies;
	int gbh_copies;
	zio_prop_t zp;
	int error;
	boolean_t has_data = !(pio->io_flags & ZIO_FLAG_NODATA);

	/*
	 * encrypted blocks need DVA[2] free so encrypted gang headers can't
	 * have a third copy.
	 */
	gbh_copies = MIN(copies + 1, spa_max_replication(spa));
	if (gio->io_prop.zp_encrypt && gbh_copies >= SPA_DVAS_PER_BP)
		gbh_copies = SPA_DVAS_PER_BP - 1;

	int flags = METASLAB_HINTBP_FAVOR | METASLAB_GANG_HEADER;
	if (pio->io_flags & ZIO_FLAG_IO_ALLOCATING) {
		ASSERT(pio->io_priority == ZIO_PRIORITY_ASYNC_WRITE);
		ASSERT(has_data);

		flags |= METASLAB_ASYNC_ALLOC;
		VERIFY(zfs_refcount_held(&mc->mc_alloc_slots[pio->io_allocator],
		    pio));

		/*
		 * The logical zio has already placed a reservation for
		 * 'copies' allocation slots but gang blocks may require
		 * additional copies. These additional copies
		 * (i.e. gbh_copies - copies) are guaranteed to succeed
		 * since metaslab_class_throttle_reserve() always allows
		 * additional reservations for gang blocks.
		 */
		VERIFY(metaslab_class_throttle_reserve(mc, gbh_copies - copies,
		    pio->io_allocator, pio, flags));
	}

	error = metaslab_alloc(spa, mc, SPA_GANGBLOCKSIZE,
	    bp, gbh_copies, txg, pio == gio ? NULL : gio->io_bp, flags,
	    &pio->io_alloc_list, pio, pio->io_allocator);
	if (error) {
		if (pio->io_flags & ZIO_FLAG_IO_ALLOCATING) {
			ASSERT(pio->io_priority == ZIO_PRIORITY_ASYNC_WRITE);
			ASSERT(has_data);

			/*
			 * If we failed to allocate the gang block header then
			 * we remove any additional allocation reservations that
			 * we placed here. The original reservation will
			 * be removed when the logical I/O goes to the ready
			 * stage.
			 */
			metaslab_class_throttle_unreserve(mc,
			    gbh_copies - copies, pio->io_allocator, pio);
		}

		pio->io_error = error;
		return (pio);
	}

	if (pio == gio) {
		gnpp = &gio->io_gang_tree;
	} else {
		gnpp = pio->io_private;
		ASSERT(pio->io_ready == zio_write_gang_member_ready);
	}

	gn = zio_gang_node_alloc(gnpp);
	gbh = gn->gn_gbh;
	bzero(gbh, SPA_GANGBLOCKSIZE);
	gbh_abd = abd_get_from_buf(gbh, SPA_GANGBLOCKSIZE);

	/*
	 * Create the gang header.
	 */
	zio = zio_rewrite(pio, spa, txg, bp, gbh_abd, SPA_GANGBLOCKSIZE,
	    zio_write_gang_done, NULL, pio->io_priority,
	    ZIO_GANG_CHILD_FLAGS(pio), &pio->io_bookmark);

	/*
	 * Create and nowait the gang children.
	 */
	for (int g = 0; resid != 0; resid -= lsize, g++) {
		lsize = P2ROUNDUP(resid / (SPA_GBH_NBLKPTRS - g),
		    SPA_MINBLOCKSIZE);
		ASSERT(lsize >= SPA_MINBLOCKSIZE && lsize <= resid);

		zp.zp_checksum = gio->io_prop.zp_checksum;
		zp.zp_compress = ZIO_COMPRESS_OFF;
		zp.zp_type = DMU_OT_NONE;
		zp.zp_level = 0;
		zp.zp_copies = gio->io_prop.zp_copies;
		zp.zp_dedup = B_FALSE;
		zp.zp_dedup_verify = B_FALSE;
		zp.zp_nopwrite = B_FALSE;
		zp.zp_encrypt = gio->io_prop.zp_encrypt;
		zp.zp_byteorder = gio->io_prop.zp_byteorder;
		bzero(zp.zp_salt, ZIO_DATA_SALT_LEN);
		bzero(zp.zp_iv, ZIO_DATA_IV_LEN);
		bzero(zp.zp_mac, ZIO_DATA_MAC_LEN);

		zio_t *cio = zio_write(zio, spa, txg, &gbh->zg_blkptr[g],
		    has_data ? abd_get_offset(pio->io_abd, pio->io_size -
		    resid) : NULL, lsize, lsize, &zp,
		    zio_write_gang_member_ready, NULL, NULL,
		    zio_write_gang_done, &gn->gn_child[g], pio->io_priority,
		    ZIO_GANG_CHILD_FLAGS(pio), &pio->io_bookmark);

		if (pio->io_flags & ZIO_FLAG_IO_ALLOCATING) {
			ASSERT(pio->io_priority == ZIO_PRIORITY_ASYNC_WRITE);
			ASSERT(has_data);

			/*
			 * Gang children won't throttle but we should
			 * account for their work, so reserve an allocation
			 * slot for them here.
			 */
			VERIFY(metaslab_class_throttle_reserve(mc,
			    zp.zp_copies, cio->io_allocator, cio, flags));
		}
		zio_nowait(cio);
	}

	/*
	 * Set pio's pipeline to just wait for zio to finish.
	 */
	pio->io_pipeline = ZIO_INTERLOCK_PIPELINE;

	/*
	 * We didn't allocate this bp, so make sure it doesn't get unmarked.
	 */
	pio->io_flags &= ~ZIO_FLAG_FASTWRITE;

	zio_nowait(zio);

	return (pio);
}

/*
 * The zio_nop_write stage in the pipeline determines if allocating a
 * new bp is necessary.  The nopwrite feature can handle writes in
 * either syncing or open context (i.e. zil writes) and as a result is
 * mutually exclusive with dedup.
 *
 * By leveraging a cryptographically secure checksum, such as SHA256, we
 * can compare the checksums of the new data and the old to determine if
 * allocating a new block is required.  Note that our requirements for
 * cryptographic strength are fairly weak: there can't be any accidental
 * hash collisions, but we don't need to be secure against intentional
 * (malicious) collisions.  To trigger a nopwrite, you have to be able
 * to write the file to begin with, and triggering an incorrect (hash
 * collision) nopwrite is no worse than simply writing to the file.
 * That said, there are no known attacks against the checksum algorithms
 * used for nopwrite, assuming that the salt and the checksums
 * themselves remain secret.
 */
static zio_t *
zio_nop_write(zio_t *zio)
{
	blkptr_t *bp = zio->io_bp;
	blkptr_t *bp_orig = &zio->io_bp_orig;
	zio_prop_t *zp = &zio->io_prop;

	ASSERT(BP_GET_LEVEL(bp) == 0);
	ASSERT(!(zio->io_flags & ZIO_FLAG_IO_REWRITE));
	ASSERT(zp->zp_nopwrite);
	ASSERT(!zp->zp_dedup);
	ASSERT(zio->io_bp_override == NULL);
	ASSERT(IO_IS_ALLOCATING(zio));

	/*
	 * Check to see if the original bp and the new bp have matching
	 * characteristics (i.e. same checksum, compression algorithms, etc).
	 * If they don't then just continue with the pipeline which will
	 * allocate a new bp.
	 */
	if (BP_IS_HOLE(bp_orig) ||
	    !(zio_checksum_table[BP_GET_CHECKSUM(bp)].ci_flags &
	    ZCHECKSUM_FLAG_NOPWRITE) ||
	    BP_IS_ENCRYPTED(bp) || BP_IS_ENCRYPTED(bp_orig) ||
	    BP_GET_CHECKSUM(bp) != BP_GET_CHECKSUM(bp_orig) ||
	    BP_GET_COMPRESS(bp) != BP_GET_COMPRESS(bp_orig) ||
	    BP_GET_DEDUP(bp) != BP_GET_DEDUP(bp_orig) ||
	    zp->zp_copies != BP_GET_NDVAS(bp_orig))
		return (zio);

	/*
	 * If the checksums match then reset the pipeline so that we
	 * avoid allocating a new bp and issuing any I/O.
	 */
	if (ZIO_CHECKSUM_EQUAL(bp->blk_cksum, bp_orig->blk_cksum)) {
		ASSERT(zio_checksum_table[zp->zp_checksum].ci_flags &
		    ZCHECKSUM_FLAG_NOPWRITE);
		ASSERT3U(BP_GET_PSIZE(bp), ==, BP_GET_PSIZE(bp_orig));
		ASSERT3U(BP_GET_LSIZE(bp), ==, BP_GET_LSIZE(bp_orig));
		ASSERT(zp->zp_compress != ZIO_COMPRESS_OFF);
		ASSERT(bcmp(&bp->blk_prop, &bp_orig->blk_prop,
		    sizeof (uint64_t)) == 0);

		*bp = *bp_orig;
		zio->io_pipeline = ZIO_INTERLOCK_PIPELINE;
		zio->io_flags |= ZIO_FLAG_NOPWRITE;
	}

	return (zio);
}

/*
 * ==========================================================================
 * Dedup
 * ==========================================================================
 */
static void
zio_ddt_child_read_done(zio_t *zio)
{
	blkptr_t *bp = zio->io_bp;
	ddt_entry_t *dde = zio->io_private;
	ddt_phys_t *ddp;
	zio_t *pio = zio_unique_parent(zio);

	mutex_enter(&pio->io_lock);
	ddp = ddt_phys_select(dde, bp);
	if (zio->io_error == 0)
		ddt_phys_clear(ddp);	/* this ddp doesn't need repair */

	if (zio->io_error == 0 && dde->dde_repair_abd == NULL)
		dde->dde_repair_abd = zio->io_abd;
	else
		abd_free(zio->io_abd);
	mutex_exit(&pio->io_lock);
}

static zio_t *
zio_ddt_read_start(zio_t *zio)
{
	blkptr_t *bp = zio->io_bp;

	ASSERT(BP_GET_DEDUP(bp));
	ASSERT(BP_GET_PSIZE(bp) == zio->io_size);
	ASSERT(zio->io_child_type == ZIO_CHILD_LOGICAL);

	if (zio->io_child_error[ZIO_CHILD_DDT]) {
		ddt_t *ddt = ddt_select(zio->io_spa, bp);
		ddt_entry_t *dde = ddt_repair_start(ddt, bp);
		ddt_phys_t *ddp = dde->dde_phys;
		ddt_phys_t *ddp_self = ddt_phys_select(dde, bp);
		blkptr_t blk;

		ASSERT(zio->io_vsd == NULL);
		zio->io_vsd = dde;

		if (ddp_self == NULL)
			return (zio);

		for (int p = 0; p < DDT_PHYS_TYPES; p++, ddp++) {
			if (ddp->ddp_phys_birth == 0 || ddp == ddp_self)
				continue;
			ddt_bp_create(ddt->ddt_checksum, &dde->dde_key, ddp,
			    &blk);
			zio_nowait(zio_read(zio, zio->io_spa, &blk,
			    abd_alloc_for_io(zio->io_size, B_TRUE),
			    zio->io_size, zio_ddt_child_read_done, dde,
			    zio->io_priority, ZIO_DDT_CHILD_FLAGS(zio) |
			    ZIO_FLAG_DONT_PROPAGATE, &zio->io_bookmark));
		}
		return (zio);
	}

	zio_nowait(zio_read(zio, zio->io_spa, bp,
	    zio->io_abd, zio->io_size, NULL, NULL, zio->io_priority,
	    ZIO_DDT_CHILD_FLAGS(zio), &zio->io_bookmark));

	return (zio);
}

static zio_t *
zio_ddt_read_done(zio_t *zio)
{
	blkptr_t *bp = zio->io_bp;

	if (zio_wait_for_children(zio, ZIO_CHILD_DDT_BIT, ZIO_WAIT_DONE)) {
		return (NULL);
	}

	ASSERT(BP_GET_DEDUP(bp));
	ASSERT(BP_GET_PSIZE(bp) == zio->io_size);
	ASSERT(zio->io_child_type == ZIO_CHILD_LOGICAL);

	if (zio->io_child_error[ZIO_CHILD_DDT]) {
		ddt_t *ddt = ddt_select(zio->io_spa, bp);
		ddt_entry_t *dde = zio->io_vsd;
		if (ddt == NULL) {
			ASSERT(spa_load_state(zio->io_spa) != SPA_LOAD_NONE);
			return (zio);
		}
		if (dde == NULL) {
			zio->io_stage = ZIO_STAGE_DDT_READ_START >> 1;
			zio_taskq_dispatch(zio, ZIO_TASKQ_ISSUE, B_FALSE);
			return (NULL);
		}
		if (dde->dde_repair_abd != NULL) {
			abd_copy(zio->io_abd, dde->dde_repair_abd,
			    zio->io_size);
			zio->io_child_error[ZIO_CHILD_DDT] = 0;
		}
		ddt_repair_done(ddt, dde);
		zio->io_vsd = NULL;
	}

	ASSERT(zio->io_vsd == NULL);

	return (zio);
}

static boolean_t
zio_ddt_collision(zio_t *zio, ddt_t *ddt, ddt_entry_t *dde)
{
	spa_t *spa = zio->io_spa;
	boolean_t do_raw = !!(zio->io_flags & ZIO_FLAG_RAW);

	ASSERT(!(zio->io_bp_override && do_raw));

	/*
	 * Note: we compare the original data, not the transformed data,
	 * because when zio->io_bp is an override bp, we will not have
	 * pushed the I/O transforms.  That's an important optimization
	 * because otherwise we'd compress/encrypt all dmu_sync() data twice.
	 * However, we should never get a raw, override zio so in these
	 * cases we can compare the io_abd directly. This is useful because
	 * it allows us to do dedup verification even if we don't have access
	 * to the original data (for instance, if the encryption keys aren't
	 * loaded).
	 */

	for (int p = DDT_PHYS_SINGLE; p <= DDT_PHYS_TRIPLE; p++) {
		zio_t *lio = dde->dde_lead_zio[p];

		if (lio != NULL && do_raw) {
			return (lio->io_size != zio->io_size ||
			    abd_cmp(zio->io_abd, lio->io_abd) != 0);
		} else if (lio != NULL) {
			return (lio->io_orig_size != zio->io_orig_size ||
			    abd_cmp(zio->io_orig_abd, lio->io_orig_abd) != 0);
		}
	}

	for (int p = DDT_PHYS_SINGLE; p <= DDT_PHYS_TRIPLE; p++) {
		ddt_phys_t *ddp = &dde->dde_phys[p];

		if (ddp->ddp_phys_birth != 0 && do_raw) {
			blkptr_t blk = *zio->io_bp;
			uint64_t psize;
			abd_t *tmpabd;
			int error;

			ddt_bp_fill(ddp, &blk, ddp->ddp_phys_birth);
			psize = BP_GET_PSIZE(&blk);

			if (psize != zio->io_size)
				return (B_TRUE);

			ddt_exit(ddt);

			tmpabd = abd_alloc_for_io(psize, B_TRUE);

			error = zio_wait(zio_read(NULL, spa, &blk, tmpabd,
			    psize, NULL, NULL, ZIO_PRIORITY_SYNC_READ,
			    ZIO_FLAG_CANFAIL | ZIO_FLAG_SPECULATIVE |
			    ZIO_FLAG_RAW, &zio->io_bookmark));

			if (error == 0) {
				if (abd_cmp(tmpabd, zio->io_abd) != 0)
					error = SET_ERROR(ENOENT);
			}

			abd_free(tmpabd);
			ddt_enter(ddt);
			return (error != 0);
		} else if (ddp->ddp_phys_birth != 0) {
			arc_buf_t *abuf = NULL;
			arc_flags_t aflags = ARC_FLAG_WAIT;
			blkptr_t blk = *zio->io_bp;
			int error;

			ddt_bp_fill(ddp, &blk, ddp->ddp_phys_birth);

			if (BP_GET_LSIZE(&blk) != zio->io_orig_size)
				return (B_TRUE);

			ddt_exit(ddt);

			error = arc_read(NULL, spa, &blk,
			    arc_getbuf_func, &abuf, ZIO_PRIORITY_SYNC_READ,
			    ZIO_FLAG_CANFAIL | ZIO_FLAG_SPECULATIVE,
			    &aflags, &zio->io_bookmark);

			if (error == 0) {
				if (abd_cmp_buf(zio->io_orig_abd, abuf->b_data,
				    zio->io_orig_size) != 0)
					error = SET_ERROR(ENOENT);
				arc_buf_destroy(abuf, &abuf);
			}

			ddt_enter(ddt);
			return (error != 0);
		}
	}

	return (B_FALSE);
}

static void
zio_ddt_child_write_ready(zio_t *zio)
{
	int p = zio->io_prop.zp_copies;
	ddt_t *ddt = ddt_select(zio->io_spa, zio->io_bp);
	ddt_entry_t *dde = zio->io_private;
	ddt_phys_t *ddp = &dde->dde_phys[p];
	zio_t *pio;

	if (zio->io_error)
		return;

	ddt_enter(ddt);

	ASSERT(dde->dde_lead_zio[p] == zio);

	ddt_phys_fill(ddp, zio->io_bp);

	zio_link_t *zl = NULL;
	while ((pio = zio_walk_parents(zio, &zl)) != NULL)
		ddt_bp_fill(ddp, pio->io_bp, zio->io_txg);

	ddt_exit(ddt);
}

static void
zio_ddt_child_write_done(zio_t *zio)
{
	int p = zio->io_prop.zp_copies;
	ddt_t *ddt = ddt_select(zio->io_spa, zio->io_bp);
	ddt_entry_t *dde = zio->io_private;
	ddt_phys_t *ddp = &dde->dde_phys[p];

	ddt_enter(ddt);

	ASSERT(ddp->ddp_refcnt == 0);
	ASSERT(dde->dde_lead_zio[p] == zio);
	dde->dde_lead_zio[p] = NULL;

	if (zio->io_error == 0) {
		zio_link_t *zl = NULL;
		while (zio_walk_parents(zio, &zl) != NULL)
			ddt_phys_addref(ddp);
	} else {
		ddt_phys_clear(ddp);
	}

	ddt_exit(ddt);
}

static zio_t *
zio_ddt_write(zio_t *zio)
{
	spa_t *spa = zio->io_spa;
	blkptr_t *bp = zio->io_bp;
	uint64_t txg = zio->io_txg;
	zio_prop_t *zp = &zio->io_prop;
	int p = zp->zp_copies;
	zio_t *cio = NULL;
	ddt_t *ddt = ddt_select(spa, bp);
	ddt_entry_t *dde;
	ddt_phys_t *ddp;

	ASSERT(BP_GET_DEDUP(bp));
	ASSERT(BP_GET_CHECKSUM(bp) == zp->zp_checksum);
	ASSERT(BP_IS_HOLE(bp) || zio->io_bp_override);
	ASSERT(!(zio->io_bp_override && (zio->io_flags & ZIO_FLAG_RAW)));

	ddt_enter(ddt);
	dde = ddt_lookup(ddt, bp, B_TRUE);
	ddp = &dde->dde_phys[p];

	if (zp->zp_dedup_verify && zio_ddt_collision(zio, ddt, dde)) {
		/*
		 * If we're using a weak checksum, upgrade to a strong checksum
		 * and try again.  If we're already using a strong checksum,
		 * we can't resolve it, so just convert to an ordinary write.
		 * (And automatically e-mail a paper to Nature?)
		 */
		if (!(zio_checksum_table[zp->zp_checksum].ci_flags &
		    ZCHECKSUM_FLAG_DEDUP)) {
			zp->zp_checksum = spa_dedup_checksum(spa);
			zio_pop_transforms(zio);
			zio->io_stage = ZIO_STAGE_OPEN;
			BP_ZERO(bp);
		} else {
			zp->zp_dedup = B_FALSE;
		}
		zio->io_pipeline = ZIO_WRITE_PIPELINE;
		ddt_exit(ddt);
		return (zio);
	}

	if (ddp->ddp_phys_birth != 0 || dde->dde_lead_zio[p] != NULL) {
		if (ddp->ddp_phys_birth != 0)
			ddt_bp_fill(ddp, bp, txg);
		if (dde->dde_lead_zio[p] != NULL)
			zio_add_child(zio, dde->dde_lead_zio[p]);
		else
			ddt_phys_addref(ddp);
	} else if (zio->io_bp_override) {
		ASSERT(bp->blk_birth == txg);
		ASSERT(BP_EQUAL(bp, zio->io_bp_override));
		ddt_phys_fill(ddp, bp);
		ddt_phys_addref(ddp);
	} else {
		cio = zio_write(zio, spa, txg, bp, zio->io_orig_abd,
		    zio->io_orig_size, zio->io_orig_size, zp,
		    zio_ddt_child_write_ready, NULL, NULL,
		    zio_ddt_child_write_done, dde, zio->io_priority,
		    ZIO_DDT_CHILD_FLAGS(zio), &zio->io_bookmark);

		zio_push_transform(cio, zio->io_abd, zio->io_size, 0, NULL);
		dde->dde_lead_zio[p] = cio;
	}

	ddt_exit(ddt);

	if (cio)
		zio_nowait(cio);

	return (zio);
}

ddt_entry_t *freedde; /* for debugging */

static zio_t *
zio_ddt_free(zio_t *zio)
{
	spa_t *spa = zio->io_spa;
	blkptr_t *bp = zio->io_bp;
	ddt_t *ddt = ddt_select(spa, bp);
	ddt_entry_t *dde;
	ddt_phys_t *ddp;

	ASSERT(BP_GET_DEDUP(bp));
	ASSERT(zio->io_child_type == ZIO_CHILD_LOGICAL);

	ddt_enter(ddt);
	freedde = dde = ddt_lookup(ddt, bp, B_TRUE);
	if (dde) {
		ddp = ddt_phys_select(dde, bp);
		if (ddp)
			ddt_phys_decref(ddp);
	}
	ddt_exit(ddt);

	return (zio);
}

/*
 * ==========================================================================
 * Allocate and free blocks
 * ==========================================================================
 */

static zio_t *
zio_io_to_allocate(spa_t *spa, int allocator)
{
	zio_t *zio;

	ASSERT(MUTEX_HELD(&spa->spa_alloc_locks[allocator]));

	zio = avl_first(&spa->spa_alloc_trees[allocator]);
	if (zio == NULL)
		return (NULL);

	ASSERT(IO_IS_ALLOCATING(zio));

	/*
	 * Try to place a reservation for this zio. If we're unable to
	 * reserve then we throttle.
	 */
	ASSERT3U(zio->io_allocator, ==, allocator);
	if (!metaslab_class_throttle_reserve(zio->io_metaslab_class,
	    zio->io_prop.zp_copies, zio->io_allocator, zio, 0)) {
		return (NULL);
	}

	avl_remove(&spa->spa_alloc_trees[allocator], zio);
	ASSERT3U(zio->io_stage, <, ZIO_STAGE_DVA_ALLOCATE);

	return (zio);
}

static zio_t *
zio_dva_throttle(zio_t *zio)
{
	spa_t *spa = zio->io_spa;
	zio_t *nio;
	metaslab_class_t *mc;

	/* locate an appropriate allocation class */
	mc = spa_preferred_class(spa, zio->io_size, zio->io_prop.zp_type,
	    zio->io_prop.zp_level, zio->io_prop.zp_zpl_smallblk);

	if (zio->io_priority == ZIO_PRIORITY_SYNC_WRITE ||
	    !mc->mc_alloc_throttle_enabled ||
	    zio->io_child_type == ZIO_CHILD_GANG ||
	    zio->io_flags & ZIO_FLAG_NODATA) {
		return (zio);
	}

	ASSERT(zio->io_child_type > ZIO_CHILD_GANG);

	ASSERT3U(zio->io_queued_timestamp, >, 0);
	ASSERT(zio->io_stage == ZIO_STAGE_DVA_THROTTLE);

	zbookmark_phys_t *bm = &zio->io_bookmark;
	/*
	 * We want to try to use as many allocators as possible to help improve
	 * performance, but we also want logically adjacent IOs to be physically
	 * adjacent to improve sequential read performance. We chunk each object
	 * into 2^20 block regions, and then hash based on the objset, object,
	 * level, and region to accomplish both of these goals.
	 */
	zio->io_allocator = cityhash4(bm->zb_objset, bm->zb_object,
	    bm->zb_level, bm->zb_blkid >> 20) % spa->spa_alloc_count;
	mutex_enter(&spa->spa_alloc_locks[zio->io_allocator]);
	ASSERT(zio->io_type == ZIO_TYPE_WRITE);
	zio->io_metaslab_class = mc;
	avl_add(&spa->spa_alloc_trees[zio->io_allocator], zio);
	nio = zio_io_to_allocate(spa, zio->io_allocator);
	mutex_exit(&spa->spa_alloc_locks[zio->io_allocator]);
	return (nio);
}

static void
zio_allocate_dispatch(spa_t *spa, int allocator)
{
	zio_t *zio;

	mutex_enter(&spa->spa_alloc_locks[allocator]);
	zio = zio_io_to_allocate(spa, allocator);
	mutex_exit(&spa->spa_alloc_locks[allocator]);
	if (zio == NULL)
		return;

	ASSERT3U(zio->io_stage, ==, ZIO_STAGE_DVA_THROTTLE);
	ASSERT0(zio->io_error);
	zio_taskq_dispatch(zio, ZIO_TASKQ_ISSUE, B_TRUE);
}

static zio_t *
zio_dva_allocate(zio_t *zio)
{
	spa_t *spa = zio->io_spa;
	metaslab_class_t *mc;
	blkptr_t *bp = zio->io_bp;
	int error;
	int flags = 0;

	if (zio->io_gang_leader == NULL) {
		ASSERT(zio->io_child_type > ZIO_CHILD_GANG);
		zio->io_gang_leader = zio;
	}

	ASSERT(BP_IS_HOLE(bp));
	ASSERT0(BP_GET_NDVAS(bp));
	ASSERT3U(zio->io_prop.zp_copies, >, 0);
	ASSERT3U(zio->io_prop.zp_copies, <=, spa_max_replication(spa));
	ASSERT3U(zio->io_size, ==, BP_GET_PSIZE(bp));

	flags |= (zio->io_flags & ZIO_FLAG_FASTWRITE) ? METASLAB_FASTWRITE : 0;
	if (zio->io_flags & ZIO_FLAG_NODATA)
		flags |= METASLAB_DONT_THROTTLE;
	if (zio->io_flags & ZIO_FLAG_GANG_CHILD)
		flags |= METASLAB_GANG_CHILD;
	if (zio->io_priority == ZIO_PRIORITY_ASYNC_WRITE)
		flags |= METASLAB_ASYNC_ALLOC;

	/*
	 * if not already chosen, locate an appropriate allocation class
	 */
	mc = zio->io_metaslab_class;
	if (mc == NULL) {
		mc = spa_preferred_class(spa, zio->io_size,
		    zio->io_prop.zp_type, zio->io_prop.zp_level,
		    zio->io_prop.zp_zpl_smallblk);
		zio->io_metaslab_class = mc;
	}

	error = metaslab_alloc(spa, mc, zio->io_size, bp,
	    zio->io_prop.zp_copies, zio->io_txg, NULL, flags,
	    &zio->io_alloc_list, zio, zio->io_allocator);

	/*
	 * Fallback to normal class when an alloc class is full
	 */
	if (error == ENOSPC && mc != spa_normal_class(spa)) {
		/*
		 * If throttling, transfer reservation over to normal class.
		 * The io_allocator slot can remain the same even though we
		 * are switching classes.
		 */
		if (mc->mc_alloc_throttle_enabled &&
		    (zio->io_flags & ZIO_FLAG_IO_ALLOCATING)) {
			metaslab_class_throttle_unreserve(mc,
			    zio->io_prop.zp_copies, zio->io_allocator, zio);
			zio->io_flags &= ~ZIO_FLAG_IO_ALLOCATING;

			mc = spa_normal_class(spa);
			VERIFY(metaslab_class_throttle_reserve(mc,
			    zio->io_prop.zp_copies, zio->io_allocator, zio,
			    flags | METASLAB_MUST_RESERVE));
		} else {
			mc = spa_normal_class(spa);
		}
		zio->io_metaslab_class = mc;

		error = metaslab_alloc(spa, mc, zio->io_size, bp,
		    zio->io_prop.zp_copies, zio->io_txg, NULL, flags,
		    &zio->io_alloc_list, zio, zio->io_allocator);
	}

	if (error != 0) {
		zfs_dbgmsg("%s: metaslab allocation failure: zio %px, "
		    "size %llu, error %d", spa_name(spa), zio, zio->io_size,
		    error);
		if (error == ENOSPC && zio->io_size > SPA_MINBLOCKSIZE)
			return (zio_write_gang_block(zio));
		zio->io_error = error;
	}

	return (zio);
}

static zio_t *
zio_dva_free(zio_t *zio)
{
	metaslab_free(zio->io_spa, zio->io_bp, zio->io_txg, B_FALSE);

	return (zio);
}

static zio_t *
zio_dva_claim(zio_t *zio)
{
	int error;

	error = metaslab_claim(zio->io_spa, zio->io_bp, zio->io_txg);
	if (error)
		zio->io_error = error;

	return (zio);
}

/*
 * Undo an allocation.  This is used by zio_done() when an I/O fails
 * and we want to give back the block we just allocated.
 * This handles both normal blocks and gang blocks.
 */
static void
zio_dva_unallocate(zio_t *zio, zio_gang_node_t *gn, blkptr_t *bp)
{
	ASSERT(bp->blk_birth == zio->io_txg || BP_IS_HOLE(bp));
	ASSERT(zio->io_bp_override == NULL);

	if (!BP_IS_HOLE(bp))
		metaslab_free(zio->io_spa, bp, bp->blk_birth, B_TRUE);

	if (gn != NULL) {
		for (int g = 0; g < SPA_GBH_NBLKPTRS; g++) {
			zio_dva_unallocate(zio, gn->gn_child[g],
			    &gn->gn_gbh->zg_blkptr[g]);
		}
	}
}

/*
 * Try to allocate an intent log block.  Return 0 on success, errno on failure.
 */
int
zio_alloc_zil(spa_t *spa, objset_t *os, uint64_t txg, blkptr_t *new_bp,
    uint64_t size, boolean_t *slog)
{
	int error = 1;
	zio_alloc_list_t io_alloc_list;

	ASSERT(txg > spa_syncing_txg(spa));

	metaslab_trace_init(&io_alloc_list);

	/*
	 * Block pointer fields are useful to metaslabs for stats and debugging.
	 * Fill in the obvious ones before calling into metaslab_alloc().
	 */
	BP_SET_TYPE(new_bp, DMU_OT_INTENT_LOG);
	BP_SET_PSIZE(new_bp, size);
	BP_SET_LEVEL(new_bp, 0);

	/*
	 * When allocating a zil block, we don't have information about
	 * the final destination of the block except the objset it's part
	 * of, so we just hash the objset ID to pick the allocator to get
	 * some parallelism.
	 */
	error = metaslab_alloc(spa, spa_log_class(spa), size, new_bp, 1,
	    txg, NULL, METASLAB_FASTWRITE, &io_alloc_list, NULL,
	    cityhash4(0, 0, 0, os->os_dsl_dataset->ds_object) %
	    spa->spa_alloc_count);
	if (error == 0) {
		*slog = TRUE;
	} else {
		error = metaslab_alloc(spa, spa_normal_class(spa), size,
		    new_bp, 1, txg, NULL, METASLAB_FASTWRITE,
		    &io_alloc_list, NULL, cityhash4(0, 0, 0,
		    os->os_dsl_dataset->ds_object) % spa->spa_alloc_count);
		if (error == 0)
			*slog = FALSE;
	}
	metaslab_trace_fini(&io_alloc_list);

	if (error == 0) {
		BP_SET_LSIZE(new_bp, size);
		BP_SET_PSIZE(new_bp, size);
		BP_SET_COMPRESS(new_bp, ZIO_COMPRESS_OFF);
		BP_SET_CHECKSUM(new_bp,
		    spa_version(spa) >= SPA_VERSION_SLIM_ZIL
		    ? ZIO_CHECKSUM_ZILOG2 : ZIO_CHECKSUM_ZILOG);
		BP_SET_TYPE(new_bp, DMU_OT_INTENT_LOG);
		BP_SET_LEVEL(new_bp, 0);
		BP_SET_DEDUP(new_bp, 0);
		BP_SET_BYTEORDER(new_bp, ZFS_HOST_BYTEORDER);

		/*
		 * encrypted blocks will require an IV and salt. We generate
		 * these now since we will not be rewriting the bp at
		 * rewrite time.
		 */
		if (os->os_encrypted) {
			uint8_t iv[ZIO_DATA_IV_LEN];
			uint8_t salt[ZIO_DATA_SALT_LEN];

			BP_SET_CRYPT(new_bp, B_TRUE);
			VERIFY0(spa_crypt_get_salt(spa,
			    dmu_objset_id(os), salt));
			VERIFY0(zio_crypt_generate_iv(iv));

			zio_crypt_encode_params_bp(new_bp, salt, iv);
		}
	} else {
		zfs_dbgmsg("%s: zil block allocation failure: "
		    "size %llu, error %d", spa_name(spa), size, error);
	}

	return (error);
}

/*
 * ==========================================================================
 * Read and write to physical devices
 * ==========================================================================
 */

/*
 * Issue an I/O to the underlying vdev. Typically the issue pipeline
 * stops after this stage and will resume upon I/O completion.
 * However, there are instances where the vdev layer may need to
 * continue the pipeline when an I/O was not issued. Since the I/O
 * that was sent to the vdev layer might be different than the one
 * currently active in the pipeline (see vdev_queue_io()), we explicitly
 * force the underlying vdev layers to call either zio_execute() or
 * zio_interrupt() to ensure that the pipeline continues with the correct I/O.
 */
static zio_t *
zio_vdev_io_start(zio_t *zio)
{
	vdev_t *vd = zio->io_vd;
	uint64_t align;
	spa_t *spa = zio->io_spa;

	zio->io_delay = 0;

	ASSERT(zio->io_error == 0);
	ASSERT(zio->io_child_error[ZIO_CHILD_VDEV] == 0);

	if (vd == NULL) {
		if (!(zio->io_flags & ZIO_FLAG_CONFIG_WRITER))
			spa_config_enter(spa, SCL_ZIO, zio, RW_READER);

		/*
		 * The mirror_ops handle multiple DVAs in a single BP.
		 */
		vdev_mirror_ops.vdev_op_io_start(zio);
		return (NULL);
	}

	ASSERT3P(zio->io_logical, !=, zio);
	if (zio->io_type == ZIO_TYPE_WRITE) {
		ASSERT(spa->spa_trust_config);

		/*
		 * Note: the code can handle other kinds of writes,
		 * but we don't expect them.
		 */
		if (zio->io_vd->vdev_removing) {
			ASSERT(zio->io_flags &
			    (ZIO_FLAG_PHYSICAL | ZIO_FLAG_SELF_HEAL |
			    ZIO_FLAG_RESILVER | ZIO_FLAG_INDUCE_DAMAGE));
		}
	}

	align = 1ULL << vd->vdev_top->vdev_ashift;

	if (!(zio->io_flags & ZIO_FLAG_PHYSICAL) &&
	    P2PHASE(zio->io_size, align) != 0) {
		/* Transform logical writes to be a full physical block size. */
		uint64_t asize = P2ROUNDUP(zio->io_size, align);
		abd_t *abuf = abd_alloc_sametype(zio->io_abd, asize);
		ASSERT(vd == vd->vdev_top);
		if (zio->io_type == ZIO_TYPE_WRITE) {
			abd_copy(abuf, zio->io_abd, zio->io_size);
			abd_zero_off(abuf, zio->io_size, asize - zio->io_size);
		}
		zio_push_transform(zio, abuf, asize, asize, zio_subblock);
	}

	/*
	 * If this is not a physical io, make sure that it is properly aligned
	 * before proceeding.
	 */
	if (!(zio->io_flags & ZIO_FLAG_PHYSICAL)) {
		ASSERT0(P2PHASE(zio->io_offset, align));
		ASSERT0(P2PHASE(zio->io_size, align));
	} else {
		/*
		 * For physical writes, we allow 512b aligned writes and assume
		 * the device will perform a read-modify-write as necessary.
		 */
		ASSERT0(P2PHASE(zio->io_offset, SPA_MINBLOCKSIZE));
		ASSERT0(P2PHASE(zio->io_size, SPA_MINBLOCKSIZE));
	}

	VERIFY(zio->io_type != ZIO_TYPE_WRITE || spa_writeable(spa));

	/*
	 * If this is a repair I/O, and there's no self-healing involved --
	 * that is, we're just resilvering what we expect to resilver --
	 * then don't do the I/O unless zio's txg is actually in vd's DTL.
	 * This prevents spurious resilvering.
	 *
	 * There are a few ways that we can end up creating these spurious
	 * resilver i/os:
	 *
	 * 1. A resilver i/o will be issued if any DVA in the BP has a
	 * dirty DTL.  The mirror code will issue resilver writes to
	 * each DVA, including the one(s) that are not on vdevs with dirty
	 * DTLs.
	 *
	 * 2. With nested replication, which happens when we have a
	 * "replacing" or "spare" vdev that's a child of a mirror or raidz.
	 * For example, given mirror(replacing(A+B), C), it's likely that
	 * only A is out of date (it's the new device). In this case, we'll
	 * read from C, then use the data to resilver A+B -- but we don't
	 * actually want to resilver B, just A. The top-level mirror has no
	 * way to know this, so instead we just discard unnecessary repairs
	 * as we work our way down the vdev tree.
	 *
	 * 3. ZTEST also creates mirrors of mirrors, mirrors of raidz, etc.
	 * The same logic applies to any form of nested replication: ditto
	 * + mirror, RAID-Z + replacing, etc.
	 *
	 * However, indirect vdevs point off to other vdevs which may have
	 * DTL's, so we never bypass them.  The child i/os on concrete vdevs
	 * will be properly bypassed instead.
	 */
	if ((zio->io_flags & ZIO_FLAG_IO_REPAIR) &&
	    !(zio->io_flags & ZIO_FLAG_SELF_HEAL) &&
	    zio->io_txg != 0 &&	/* not a delegated i/o */
	    vd->vdev_ops != &vdev_indirect_ops &&
	    !vdev_dtl_contains(vd, DTL_PARTIAL, zio->io_txg, 1)) {
		ASSERT(zio->io_type == ZIO_TYPE_WRITE);
		zio_vdev_io_bypass(zio);
		return (zio);
	}

	if (vd->vdev_ops->vdev_op_leaf && (zio->io_type == ZIO_TYPE_READ ||
	    zio->io_type == ZIO_TYPE_WRITE || zio->io_type == ZIO_TYPE_TRIM)) {

		if (zio->io_type == ZIO_TYPE_READ && vdev_cache_read(zio))
			return (zio);

		if ((zio = vdev_queue_io(zio)) == NULL)
			return (NULL);

		if (!vdev_accessible(vd, zio)) {
			zio->io_error = SET_ERROR(ENXIO);
			zio_interrupt(zio);
			return (NULL);
		}
		zio->io_delay = gethrtime();
	}

	vd->vdev_ops->vdev_op_io_start(zio);
	return (NULL);
}

static zio_t *
zio_vdev_io_done(zio_t *zio)
{
	vdev_t *vd = zio->io_vd;
	vdev_ops_t *ops = vd ? vd->vdev_ops : &vdev_mirror_ops;
	boolean_t unexpected_error = B_FALSE;

	if (zio_wait_for_children(zio, ZIO_CHILD_VDEV_BIT, ZIO_WAIT_DONE)) {
		return (NULL);
	}

	ASSERT(zio->io_type == ZIO_TYPE_READ ||
	    zio->io_type == ZIO_TYPE_WRITE || zio->io_type == ZIO_TYPE_TRIM);

	if (zio->io_delay)
		zio->io_delay = gethrtime() - zio->io_delay;

	if (vd != NULL && vd->vdev_ops->vdev_op_leaf) {

		vdev_queue_io_done(zio);

		if (zio->io_type == ZIO_TYPE_WRITE)
			vdev_cache_write(zio);

		if (zio_injection_enabled && zio->io_error == 0)
			zio->io_error = zio_handle_device_injections(vd, zio,
			    EIO, EILSEQ);

		if (zio_injection_enabled && zio->io_error == 0)
			zio->io_error = zio_handle_label_injection(zio, EIO);

		if (zio->io_error && zio->io_type != ZIO_TYPE_TRIM) {
			if (!vdev_accessible(vd, zio)) {
				zio->io_error = SET_ERROR(ENXIO);
			} else {
				unexpected_error = B_TRUE;
			}
		}
	}

	ops->vdev_op_io_done(zio);

	if (unexpected_error)
		VERIFY(vdev_probe(vd, zio) == NULL);

	return (zio);
}

/*
 * This function is used to change the priority of an existing zio that is
 * currently in-flight. This is used by the arc to upgrade priority in the
 * event that a demand read is made for a block that is currently queued
 * as a scrub or async read IO. Otherwise, the high priority read request
 * would end up having to wait for the lower priority IO.
 */
void
zio_change_priority(zio_t *pio, zio_priority_t priority)
{
	zio_t *cio, *cio_next;
	zio_link_t *zl = NULL;

	ASSERT3U(priority, <, ZIO_PRIORITY_NUM_QUEUEABLE);

	if (pio->io_vd != NULL && pio->io_vd->vdev_ops->vdev_op_leaf) {
		vdev_queue_change_io_priority(pio, priority);
	} else {
		pio->io_priority = priority;
	}

	mutex_enter(&pio->io_lock);
	for (cio = zio_walk_children(pio, &zl); cio != NULL; cio = cio_next) {
		cio_next = zio_walk_children(pio, &zl);
		zio_change_priority(cio, priority);
	}
	mutex_exit(&pio->io_lock);
}

/*
 * For non-raidz ZIOs, we can just copy aside the bad data read from the
 * disk, and use that to finish the checksum ereport later.
 */
static void
zio_vsd_default_cksum_finish(zio_cksum_report_t *zcr,
    const abd_t *good_buf)
{
	/* no processing needed */
	zfs_ereport_finish_checksum(zcr, good_buf, zcr->zcr_cbdata, B_FALSE);
}

/*ARGSUSED*/
void
zio_vsd_default_cksum_report(zio_t *zio, zio_cksum_report_t *zcr, void *ignored)
{
	void *abd = abd_alloc_sametype(zio->io_abd, zio->io_size);

	abd_copy(abd, zio->io_abd, zio->io_size);

	zcr->zcr_cbinfo = zio->io_size;
	zcr->zcr_cbdata = abd;
	zcr->zcr_finish = zio_vsd_default_cksum_finish;
	zcr->zcr_free = zio_abd_free;
}

static zio_t *
zio_vdev_io_assess(zio_t *zio)
{
	vdev_t *vd = zio->io_vd;

	if (zio_wait_for_children(zio, ZIO_CHILD_VDEV_BIT, ZIO_WAIT_DONE)) {
		return (NULL);
	}

	if (vd == NULL && !(zio->io_flags & ZIO_FLAG_CONFIG_WRITER))
		spa_config_exit(zio->io_spa, SCL_ZIO, zio);

	if (zio->io_vsd != NULL) {
		zio->io_vsd_ops->vsd_free(zio);
		zio->io_vsd = NULL;
	}

	if (zio_injection_enabled && zio->io_error == 0)
		zio->io_error = zio_handle_fault_injection(zio, EIO);

	/*
	 * If the I/O failed, determine whether we should attempt to retry it.
	 *
	 * On retry, we cut in line in the issue queue, since we don't want
	 * compression/checksumming/etc. work to prevent our (cheap) IO reissue.
	 */
	if (zio->io_error && vd == NULL &&
	    !(zio->io_flags & (ZIO_FLAG_DONT_RETRY | ZIO_FLAG_IO_RETRY))) {
		ASSERT(!(zio->io_flags & ZIO_FLAG_DONT_QUEUE));	/* not a leaf */
		ASSERT(!(zio->io_flags & ZIO_FLAG_IO_BYPASS));	/* not a leaf */
		zio->io_error = 0;
		zio->io_flags |= ZIO_FLAG_IO_RETRY |
		    ZIO_FLAG_DONT_CACHE | ZIO_FLAG_DONT_AGGREGATE;
		zio->io_stage = ZIO_STAGE_VDEV_IO_START >> 1;
		zio_taskq_dispatch(zio, ZIO_TASKQ_ISSUE,
		    zio_requeue_io_start_cut_in_line);
		return (NULL);
	}

	/*
	 * If we got an error on a leaf device, convert it to ENXIO
	 * if the device is not accessible at all.
	 */
	if (zio->io_error && vd != NULL && vd->vdev_ops->vdev_op_leaf &&
	    !vdev_accessible(vd, zio))
		zio->io_error = SET_ERROR(ENXIO);

	/*
	 * If we can't write to an interior vdev (mirror or RAID-Z),
	 * set vdev_cant_write so that we stop trying to allocate from it.
	 */
	if (zio->io_error == ENXIO && zio->io_type == ZIO_TYPE_WRITE &&
	    vd != NULL && !vd->vdev_ops->vdev_op_leaf) {
		vd->vdev_cant_write = B_TRUE;
	}

	/*
	 * If a cache flush returns ENOTSUP or ENOTTY, we know that no future
	 * attempts will ever succeed. In this case we set a persistent
	 * boolean flag so that we don't bother with it in the future.
	 */
	if ((zio->io_error == ENOTSUP || zio->io_error == ENOTTY) &&
	    zio->io_type == ZIO_TYPE_IOCTL &&
	    zio->io_cmd == DKIOCFLUSHWRITECACHE && vd != NULL)
		vd->vdev_nowritecache = B_TRUE;

	if (zio->io_error)
		zio->io_pipeline = ZIO_INTERLOCK_PIPELINE;

	if (vd != NULL && vd->vdev_ops->vdev_op_leaf &&
	    zio->io_physdone != NULL) {
		ASSERT(!(zio->io_flags & ZIO_FLAG_DELEGATED));
		ASSERT(zio->io_child_type == ZIO_CHILD_VDEV);
		zio->io_physdone(zio->io_logical);
	}

	return (zio);
}

void
zio_vdev_io_reissue(zio_t *zio)
{
	ASSERT(zio->io_stage == ZIO_STAGE_VDEV_IO_START);
	ASSERT(zio->io_error == 0);

	zio->io_stage >>= 1;
}

void
zio_vdev_io_redone(zio_t *zio)
{
	ASSERT(zio->io_stage == ZIO_STAGE_VDEV_IO_DONE);

	zio->io_stage >>= 1;
}

void
zio_vdev_io_bypass(zio_t *zio)
{
	ASSERT(zio->io_stage == ZIO_STAGE_VDEV_IO_START);
	ASSERT(zio->io_error == 0);

	zio->io_flags |= ZIO_FLAG_IO_BYPASS;
	zio->io_stage = ZIO_STAGE_VDEV_IO_ASSESS >> 1;
}

/*
 * ==========================================================================
 * Encrypt and store encryption parameters
 * ==========================================================================
 */


/*
 * This function is used for ZIO_STAGE_ENCRYPT. It is responsible for
 * managing the storage of encryption parameters and passing them to the
 * lower-level encryption functions.
 */
static zio_t *
zio_encrypt(zio_t *zio)
{
	zio_prop_t *zp = &zio->io_prop;
	spa_t *spa = zio->io_spa;
	blkptr_t *bp = zio->io_bp;
	uint64_t psize = BP_GET_PSIZE(bp);
	uint64_t dsobj = zio->io_bookmark.zb_objset;
	dmu_object_type_t ot = BP_GET_TYPE(bp);
	void *enc_buf = NULL;
	abd_t *eabd = NULL;
	uint8_t salt[ZIO_DATA_SALT_LEN];
	uint8_t iv[ZIO_DATA_IV_LEN];
	uint8_t mac[ZIO_DATA_MAC_LEN];
	boolean_t no_crypt = B_FALSE;

	/* the root zio already encrypted the data */
	if (zio->io_child_type == ZIO_CHILD_GANG)
		return (zio);

	/* only ZIL blocks are re-encrypted on rewrite */
	if (!IO_IS_ALLOCATING(zio) && ot != DMU_OT_INTENT_LOG)
		return (zio);

	if (!(zp->zp_encrypt || BP_IS_ENCRYPTED(bp))) {
		BP_SET_CRYPT(bp, B_FALSE);
		return (zio);
	}

	/* if we are doing raw encryption set the provided encryption params */
	if (zio->io_flags & ZIO_FLAG_RAW_ENCRYPT) {
		ASSERT0(BP_GET_LEVEL(bp));
		BP_SET_CRYPT(bp, B_TRUE);
		BP_SET_BYTEORDER(bp, zp->zp_byteorder);
		if (ot != DMU_OT_OBJSET)
			zio_crypt_encode_mac_bp(bp, zp->zp_mac);

		/* dnode blocks must be written out in the provided byteorder */
		if (zp->zp_byteorder != ZFS_HOST_BYTEORDER &&
		    ot == DMU_OT_DNODE) {
			void *bswap_buf = zio_buf_alloc(psize);
			abd_t *babd = abd_get_from_buf(bswap_buf, psize);

			ASSERT3U(BP_GET_COMPRESS(bp), ==, ZIO_COMPRESS_OFF);
			abd_copy_to_buf(bswap_buf, zio->io_abd, psize);
			dmu_ot_byteswap[DMU_OT_BYTESWAP(ot)].ob_func(bswap_buf,
			    psize);

			abd_take_ownership_of_buf(babd, B_TRUE);
			zio_push_transform(zio, babd, psize, psize, NULL);
		}

		if (DMU_OT_IS_ENCRYPTED(ot))
			zio_crypt_encode_params_bp(bp, zp->zp_salt, zp->zp_iv);
		return (zio);
	}

	/* indirect blocks only maintain a cksum of the lower level MACs */
	if (BP_GET_LEVEL(bp) > 0) {
		BP_SET_CRYPT(bp, B_TRUE);
		VERIFY0(zio_crypt_do_indirect_mac_checksum_abd(B_TRUE,
		    zio->io_orig_abd, BP_GET_LSIZE(bp), BP_SHOULD_BYTESWAP(bp),
		    mac));
		zio_crypt_encode_mac_bp(bp, mac);
		return (zio);
	}

	/*
	 * Objset blocks are a special case since they have 2 256-bit MACs
	 * embedded within them.
	 */
	if (ot == DMU_OT_OBJSET) {
		ASSERT0(DMU_OT_IS_ENCRYPTED(ot));
		ASSERT3U(BP_GET_COMPRESS(bp), ==, ZIO_COMPRESS_OFF);
		BP_SET_CRYPT(bp, B_TRUE);
		VERIFY0(spa_do_crypt_objset_mac_abd(B_TRUE, spa, dsobj,
		    zio->io_abd, psize, BP_SHOULD_BYTESWAP(bp)));
		return (zio);
	}

	/* unencrypted object types are only authenticated with a MAC */
	if (!DMU_OT_IS_ENCRYPTED(ot)) {
		BP_SET_CRYPT(bp, B_TRUE);
		VERIFY0(spa_do_crypt_mac_abd(B_TRUE, spa, dsobj,
		    zio->io_abd, psize, mac));
		zio_crypt_encode_mac_bp(bp, mac);
		return (zio);
	}

	/*
	 * Later passes of sync-to-convergence may decide to rewrite data
	 * in place to avoid more disk reallocations. This presents a problem
	 * for encryption because this constitutes rewriting the new data with
	 * the same encryption key and IV. However, this only applies to blocks
	 * in the MOS (particularly the spacemaps) and we do not encrypt the
	 * MOS. We assert that the zio is allocating or an intent log write
	 * to enforce this.
	 */
	ASSERT(IO_IS_ALLOCATING(zio) || ot == DMU_OT_INTENT_LOG);
	ASSERT(BP_GET_LEVEL(bp) == 0 || ot == DMU_OT_INTENT_LOG);
	ASSERT(spa_feature_is_active(spa, SPA_FEATURE_ENCRYPTION));
	ASSERT3U(psize, !=, 0);

	enc_buf = zio_buf_alloc(psize);
	eabd = abd_get_from_buf(enc_buf, psize);
	abd_take_ownership_of_buf(eabd, B_TRUE);

	/*
	 * For an explanation of what encryption parameters are stored
	 * where, see the block comment in zio_crypt.c.
	 */
	if (ot == DMU_OT_INTENT_LOG) {
		zio_crypt_decode_params_bp(bp, salt, iv);
	} else {
		BP_SET_CRYPT(bp, B_TRUE);
	}

	/* Perform the encryption. This should not fail */
	VERIFY0(spa_do_crypt_abd(B_TRUE, spa, &zio->io_bookmark,
	    BP_GET_TYPE(bp), BP_GET_DEDUP(bp), BP_SHOULD_BYTESWAP(bp),
	    salt, iv, mac, psize, zio->io_abd, eabd, &no_crypt));

	/* encode encryption metadata into the bp */
	if (ot == DMU_OT_INTENT_LOG) {
		/*
		 * ZIL blocks store the MAC in the embedded checksum, so the
		 * transform must always be applied.
		 */
		zio_crypt_encode_mac_zil(enc_buf, mac);
		zio_push_transform(zio, eabd, psize, psize, NULL);
	} else {
		BP_SET_CRYPT(bp, B_TRUE);
		zio_crypt_encode_params_bp(bp, salt, iv);
		zio_crypt_encode_mac_bp(bp, mac);

		if (no_crypt) {
			ASSERT3U(ot, ==, DMU_OT_DNODE);
			abd_free(eabd);
		} else {
			zio_push_transform(zio, eabd, psize, psize, NULL);
		}
	}

	return (zio);
}

/*
 * ==========================================================================
 * Generate and verify checksums
 * ==========================================================================
 */
static zio_t *
zio_checksum_generate(zio_t *zio)
{
	blkptr_t *bp = zio->io_bp;
	enum zio_checksum checksum;

	if (bp == NULL) {
		/*
		 * This is zio_write_phys().
		 * We're either generating a label checksum, or none at all.
		 */
		checksum = zio->io_prop.zp_checksum;

		if (checksum == ZIO_CHECKSUM_OFF)
			return (zio);

		ASSERT(checksum == ZIO_CHECKSUM_LABEL);
	} else {
		if (BP_IS_GANG(bp) && zio->io_child_type == ZIO_CHILD_GANG) {
			ASSERT(!IO_IS_ALLOCATING(zio));
			checksum = ZIO_CHECKSUM_GANG_HEADER;
		} else {
			checksum = BP_GET_CHECKSUM(bp);
		}
	}

	zio_checksum_compute(zio, checksum, zio->io_abd, zio->io_size);

	return (zio);
}

static zio_t *
zio_checksum_verify(zio_t *zio)
{
	zio_bad_cksum_t info;
	blkptr_t *bp = zio->io_bp;
	int error;

	ASSERT(zio->io_vd != NULL);

	if (bp == NULL) {
		/*
		 * This is zio_read_phys().
		 * We're either verifying a label checksum, or nothing at all.
		 */
		if (zio->io_prop.zp_checksum == ZIO_CHECKSUM_OFF)
			return (zio);

		ASSERT(zio->io_prop.zp_checksum == ZIO_CHECKSUM_LABEL);
	}

	if ((error = zio_checksum_error(zio, &info)) != 0) {
		zio->io_error = error;
		if (error == ECKSUM &&
		    !(zio->io_flags & ZIO_FLAG_SPECULATIVE)) {
			mutex_enter(&zio->io_vd->vdev_stat_lock);
			zio->io_vd->vdev_stat.vs_checksum_errors++;
			mutex_exit(&zio->io_vd->vdev_stat_lock);

			zfs_ereport_start_checksum(zio->io_spa,
			    zio->io_vd, &zio->io_bookmark, zio,
			    zio->io_offset, zio->io_size, NULL, &info);
		}
	}

	return (zio);
}

/*
 * Called by RAID-Z to ensure we don't compute the checksum twice.
 */
void
zio_checksum_verified(zio_t *zio)
{
	zio->io_pipeline &= ~ZIO_STAGE_CHECKSUM_VERIFY;
}

/*
 * ==========================================================================
 * Error rank.  Error are ranked in the order 0, ENXIO, ECKSUM, EIO, other.
 * An error of 0 indicates success.  ENXIO indicates whole-device failure,
 * which may be transient (e.g. unplugged) or permanent.  ECKSUM and EIO
 * indicate errors that are specific to one I/O, and most likely permanent.
 * Any other error is presumed to be worse because we weren't expecting it.
 * ==========================================================================
 */
int
zio_worst_error(int e1, int e2)
{
	static int zio_error_rank[] = { 0, ENXIO, ECKSUM, EIO };
	int r1, r2;

	for (r1 = 0; r1 < sizeof (zio_error_rank) / sizeof (int); r1++)
		if (e1 == zio_error_rank[r1])
			break;

	for (r2 = 0; r2 < sizeof (zio_error_rank) / sizeof (int); r2++)
		if (e2 == zio_error_rank[r2])
			break;

	return (r1 > r2 ? e1 : e2);
}

/*
 * ==========================================================================
 * I/O completion
 * ==========================================================================
 */
static zio_t *
zio_ready(zio_t *zio)
{
	blkptr_t *bp = zio->io_bp;
	zio_t *pio, *pio_next;
	zio_link_t *zl = NULL;

	if (zio_wait_for_children(zio, ZIO_CHILD_GANG_BIT | ZIO_CHILD_DDT_BIT,
	    ZIO_WAIT_READY)) {
		return (NULL);
	}

	if (zio->io_ready) {
		ASSERT(IO_IS_ALLOCATING(zio));
		ASSERT(bp->blk_birth == zio->io_txg || BP_IS_HOLE(bp) ||
		    (zio->io_flags & ZIO_FLAG_NOPWRITE));
		ASSERT(zio->io_children[ZIO_CHILD_GANG][ZIO_WAIT_READY] == 0);

		zio->io_ready(zio);
	}

	if (bp != NULL && bp != &zio->io_bp_copy)
		zio->io_bp_copy = *bp;

	if (zio->io_error != 0) {
		zio->io_pipeline = ZIO_INTERLOCK_PIPELINE;

		if (zio->io_flags & ZIO_FLAG_IO_ALLOCATING) {
			ASSERT(IO_IS_ALLOCATING(zio));
			ASSERT(zio->io_priority == ZIO_PRIORITY_ASYNC_WRITE);
			ASSERT(zio->io_metaslab_class != NULL);

			/*
			 * We were unable to allocate anything, unreserve and
			 * issue the next I/O to allocate.
			 */
			metaslab_class_throttle_unreserve(
			    zio->io_metaslab_class, zio->io_prop.zp_copies,
			    zio->io_allocator, zio);
			zio_allocate_dispatch(zio->io_spa, zio->io_allocator);
		}
	}

	mutex_enter(&zio->io_lock);
	zio->io_state[ZIO_WAIT_READY] = 1;
	pio = zio_walk_parents(zio, &zl);
	mutex_exit(&zio->io_lock);

	/*
	 * As we notify zio's parents, new parents could be added.
	 * New parents go to the head of zio's io_parent_list, however,
	 * so we will (correctly) not notify them.  The remainder of zio's
	 * io_parent_list, from 'pio_next' onward, cannot change because
	 * all parents must wait for us to be done before they can be done.
	 */
	for (; pio != NULL; pio = pio_next) {
		pio_next = zio_walk_parents(zio, &zl);
		zio_notify_parent(pio, zio, ZIO_WAIT_READY, NULL);
	}

	if (zio->io_flags & ZIO_FLAG_NODATA) {
		if (BP_IS_GANG(bp)) {
			zio->io_flags &= ~ZIO_FLAG_NODATA;
		} else {
			ASSERT((uintptr_t)zio->io_abd < SPA_MAXBLOCKSIZE);
			zio->io_pipeline &= ~ZIO_VDEV_IO_STAGES;
		}
	}

	if (zio_injection_enabled &&
	    zio->io_spa->spa_syncing_txg == zio->io_txg)
		zio_handle_ignored_writes(zio);

	return (zio);
}

/*
 * Update the allocation throttle accounting.
 */
static void
zio_dva_throttle_done(zio_t *zio)
{
	ASSERTV(zio_t *lio = zio->io_logical);
	zio_t *pio = zio_unique_parent(zio);
	vdev_t *vd = zio->io_vd;
	int flags = METASLAB_ASYNC_ALLOC;

	ASSERT3P(zio->io_bp, !=, NULL);
	ASSERT3U(zio->io_type, ==, ZIO_TYPE_WRITE);
	ASSERT3U(zio->io_priority, ==, ZIO_PRIORITY_ASYNC_WRITE);
	ASSERT3U(zio->io_child_type, ==, ZIO_CHILD_VDEV);
	ASSERT(vd != NULL);
	ASSERT3P(vd, ==, vd->vdev_top);
	ASSERT(zio_injection_enabled || !(zio->io_flags & ZIO_FLAG_IO_RETRY));
	ASSERT(!(zio->io_flags & ZIO_FLAG_IO_REPAIR));
	ASSERT(zio->io_flags & ZIO_FLAG_IO_ALLOCATING);
	ASSERT(!(lio->io_flags & ZIO_FLAG_IO_REWRITE));
	ASSERT(!(lio->io_orig_flags & ZIO_FLAG_NODATA));

	/*
	 * Parents of gang children can have two flavors -- ones that
	 * allocated the gang header (will have ZIO_FLAG_IO_REWRITE set)
	 * and ones that allocated the constituent blocks. The allocation
	 * throttle needs to know the allocating parent zio so we must find
	 * it here.
	 */
	if (pio->io_child_type == ZIO_CHILD_GANG) {
		/*
		 * If our parent is a rewrite gang child then our grandparent
		 * would have been the one that performed the allocation.
		 */
		if (pio->io_flags & ZIO_FLAG_IO_REWRITE)
			pio = zio_unique_parent(pio);
		flags |= METASLAB_GANG_CHILD;
	}

	ASSERT(IO_IS_ALLOCATING(pio));
	ASSERT3P(zio, !=, zio->io_logical);
	ASSERT(zio->io_logical != NULL);
	ASSERT(!(zio->io_flags & ZIO_FLAG_IO_REPAIR));
	ASSERT0(zio->io_flags & ZIO_FLAG_NOPWRITE);
	ASSERT(zio->io_metaslab_class != NULL);

	mutex_enter(&pio->io_lock);
	metaslab_group_alloc_decrement(zio->io_spa, vd->vdev_id, pio, flags,
	    pio->io_allocator, B_TRUE);
	mutex_exit(&pio->io_lock);

	metaslab_class_throttle_unreserve(zio->io_metaslab_class, 1,
	    pio->io_allocator, pio);

	/*
	 * Call into the pipeline to see if there is more work that
	 * needs to be done. If there is work to be done it will be
	 * dispatched to another taskq thread.
	 */
	zio_allocate_dispatch(zio->io_spa, pio->io_allocator);
}

static zio_t *
zio_done(zio_t *zio)
{
	/*
	 * Always attempt to keep stack usage minimal here since
	 * we can be called recursively up to 19 levels deep.
	 */
	const uint64_t psize = zio->io_size;
	zio_t *pio, *pio_next;
	zio_link_t *zl = NULL;

	/*
	 * If our children haven't all completed,
	 * wait for them and then repeat this pipeline stage.
	 */
	if (zio_wait_for_children(zio, ZIO_CHILD_ALL_BITS, ZIO_WAIT_DONE)) {
		return (NULL);
	}

	/*
	 * If the allocation throttle is enabled, then update the accounting.
	 * We only track child I/Os that are part of an allocating async
	 * write. We must do this since the allocation is performed
	 * by the logical I/O but the actual write is done by child I/Os.
	 */
	if (zio->io_flags & ZIO_FLAG_IO_ALLOCATING &&
	    zio->io_child_type == ZIO_CHILD_VDEV) {
		ASSERT(zio->io_metaslab_class != NULL);
		ASSERT(zio->io_metaslab_class->mc_alloc_throttle_enabled);
		zio_dva_throttle_done(zio);
	}

	/*
	 * If the allocation throttle is enabled, verify that
	 * we have decremented the refcounts for every I/O that was throttled.
	 */
	if (zio->io_flags & ZIO_FLAG_IO_ALLOCATING) {
		ASSERT(zio->io_type == ZIO_TYPE_WRITE);
		ASSERT(zio->io_priority == ZIO_PRIORITY_ASYNC_WRITE);
		ASSERT(zio->io_bp != NULL);

		metaslab_group_alloc_verify(zio->io_spa, zio->io_bp, zio,
		    zio->io_allocator);
		VERIFY(zfs_refcount_not_held(
		    &zio->io_metaslab_class->mc_alloc_slots[zio->io_allocator],
		    zio));
	}


	for (int c = 0; c < ZIO_CHILD_TYPES; c++)
		for (int w = 0; w < ZIO_WAIT_TYPES; w++)
			ASSERT(zio->io_children[c][w] == 0);

	if (zio->io_bp != NULL && !BP_IS_EMBEDDED(zio->io_bp)) {
		ASSERT(zio->io_bp->blk_pad[0] == 0);
		ASSERT(zio->io_bp->blk_pad[1] == 0);
		ASSERT(bcmp(zio->io_bp, &zio->io_bp_copy,
		    sizeof (blkptr_t)) == 0 ||
		    (zio->io_bp == zio_unique_parent(zio)->io_bp));
		if (zio->io_type == ZIO_TYPE_WRITE && !BP_IS_HOLE(zio->io_bp) &&
		    zio->io_bp_override == NULL &&
		    !(zio->io_flags & ZIO_FLAG_IO_REPAIR)) {
			ASSERT3U(zio->io_prop.zp_copies, <=,
			    BP_GET_NDVAS(zio->io_bp));
			ASSERT(BP_COUNT_GANG(zio->io_bp) == 0 ||
			    (BP_COUNT_GANG(zio->io_bp) ==
			    BP_GET_NDVAS(zio->io_bp)));
		}
		if (zio->io_flags & ZIO_FLAG_NOPWRITE)
			VERIFY(BP_EQUAL(zio->io_bp, &zio->io_bp_orig));
	}

	/*
	 * If there were child vdev/gang/ddt errors, they apply to us now.
	 */
	zio_inherit_child_errors(zio, ZIO_CHILD_VDEV);
	zio_inherit_child_errors(zio, ZIO_CHILD_GANG);
	zio_inherit_child_errors(zio, ZIO_CHILD_DDT);

	/*
	 * If the I/O on the transformed data was successful, generate any
	 * checksum reports now while we still have the transformed data.
	 */
	if (zio->io_error == 0) {
		while (zio->io_cksum_report != NULL) {
			zio_cksum_report_t *zcr = zio->io_cksum_report;
			uint64_t align = zcr->zcr_align;
			uint64_t asize = P2ROUNDUP(psize, align);
			abd_t *adata = zio->io_abd;

			if (asize != psize) {
				adata = abd_alloc(asize, B_TRUE);
				abd_copy(adata, zio->io_abd, psize);
				abd_zero_off(adata, psize, asize - psize);
			}

			zio->io_cksum_report = zcr->zcr_next;
			zcr->zcr_next = NULL;
			zcr->zcr_finish(zcr, adata);
			zfs_ereport_free_checksum(zcr);

			if (asize != psize)
				abd_free(adata);
		}
	}

	zio_pop_transforms(zio);	/* note: may set zio->io_error */

	vdev_stat_update(zio, psize);

	/*
	 * If this I/O is attached to a particular vdev is slow, exceeding
	 * 30 seconds to complete, post an error described the I/O delay.
	 * We ignore these errors if the device is currently unavailable.
	 */
	if (zio->io_delay >= MSEC2NSEC(zio_slow_io_ms)) {
		if (zio->io_vd != NULL && !vdev_is_dead(zio->io_vd)) {
			/*
			 * We want to only increment our slow IO counters if
			 * the IO is valid (i.e. not if the drive is removed).
			 *
			 * zfs_ereport_post() will also do these checks, but
			 * it can also ratelimit and have other failures, so we
			 * need to increment the slow_io counters independent
			 * of it.
			 */
			if (zfs_ereport_is_valid(FM_EREPORT_ZFS_DELAY,
			    zio->io_spa, zio->io_vd, zio)) {
				mutex_enter(&zio->io_vd->vdev_stat_lock);
				zio->io_vd->vdev_stat.vs_slow_ios++;
				mutex_exit(&zio->io_vd->vdev_stat_lock);

				zfs_ereport_post(FM_EREPORT_ZFS_DELAY,
				    zio->io_spa, zio->io_vd, &zio->io_bookmark,
				    zio, 0, 0);
			}
		}
	}

	if (zio->io_error) {
		/*
		 * If this I/O is attached to a particular vdev,
		 * generate an error message describing the I/O failure
		 * at the block level.  We ignore these errors if the
		 * device is currently unavailable.
		 */
		if (zio->io_error != ECKSUM && zio->io_vd != NULL &&
		    !vdev_is_dead(zio->io_vd)) {
			mutex_enter(&zio->io_vd->vdev_stat_lock);
			if (zio->io_type == ZIO_TYPE_READ) {
				zio->io_vd->vdev_stat.vs_read_errors++;
			} else if (zio->io_type == ZIO_TYPE_WRITE) {
				zio->io_vd->vdev_stat.vs_write_errors++;
			}
			mutex_exit(&zio->io_vd->vdev_stat_lock);

			zfs_ereport_post(FM_EREPORT_ZFS_IO, zio->io_spa,
			    zio->io_vd, &zio->io_bookmark, zio, 0, 0);
		}

		if ((zio->io_error == EIO || !(zio->io_flags &
		    (ZIO_FLAG_SPECULATIVE | ZIO_FLAG_DONT_PROPAGATE))) &&
		    zio == zio->io_logical) {
			/*
			 * For logical I/O requests, tell the SPA to log the
			 * error and generate a logical data ereport.
			 */
			spa_log_error(zio->io_spa, &zio->io_bookmark);
			zfs_ereport_post(FM_EREPORT_ZFS_DATA, zio->io_spa,
			    NULL, &zio->io_bookmark, zio, 0, 0);
		}
	}

	if (zio->io_error && zio == zio->io_logical) {
		/*
		 * Determine whether zio should be reexecuted.  This will
		 * propagate all the way to the root via zio_notify_parent().
		 */
		ASSERT(zio->io_vd == NULL && zio->io_bp != NULL);
		ASSERT(zio->io_child_type == ZIO_CHILD_LOGICAL);

		if (IO_IS_ALLOCATING(zio) &&
		    !(zio->io_flags & ZIO_FLAG_CANFAIL)) {
			if (zio->io_error != ENOSPC)
				zio->io_reexecute |= ZIO_REEXECUTE_NOW;
			else
				zio->io_reexecute |= ZIO_REEXECUTE_SUSPEND;
		}

		if ((zio->io_type == ZIO_TYPE_READ ||
		    zio->io_type == ZIO_TYPE_FREE) &&
		    !(zio->io_flags & ZIO_FLAG_SCAN_THREAD) &&
		    zio->io_error == ENXIO &&
		    spa_load_state(zio->io_spa) == SPA_LOAD_NONE &&
		    spa_get_failmode(zio->io_spa) != ZIO_FAILURE_MODE_CONTINUE)
			zio->io_reexecute |= ZIO_REEXECUTE_SUSPEND;

		if (!(zio->io_flags & ZIO_FLAG_CANFAIL) && !zio->io_reexecute)
			zio->io_reexecute |= ZIO_REEXECUTE_SUSPEND;

		/*
		 * Here is a possibly good place to attempt to do
		 * either combinatorial reconstruction or error correction
		 * based on checksums.  It also might be a good place
		 * to send out preliminary ereports before we suspend
		 * processing.
		 */
	}

	/*
	 * If there were logical child errors, they apply to us now.
	 * We defer this until now to avoid conflating logical child
	 * errors with errors that happened to the zio itself when
	 * updating vdev stats and reporting FMA events above.
	 */
	zio_inherit_child_errors(zio, ZIO_CHILD_LOGICAL);

	if ((zio->io_error || zio->io_reexecute) &&
	    IO_IS_ALLOCATING(zio) && zio->io_gang_leader == zio &&
	    !(zio->io_flags & (ZIO_FLAG_IO_REWRITE | ZIO_FLAG_NOPWRITE)))
		zio_dva_unallocate(zio, zio->io_gang_tree, zio->io_bp);

	zio_gang_tree_free(&zio->io_gang_tree);

	/*
	 * Godfather I/Os should never suspend.
	 */
	if ((zio->io_flags & ZIO_FLAG_GODFATHER) &&
	    (zio->io_reexecute & ZIO_REEXECUTE_SUSPEND))
		zio->io_reexecute &= ~ZIO_REEXECUTE_SUSPEND;

	if (zio->io_reexecute) {
		/*
		 * This is a logical I/O that wants to reexecute.
		 *
		 * Reexecute is top-down.  When an i/o fails, if it's not
		 * the root, it simply notifies its parent and sticks around.
		 * The parent, seeing that it still has children in zio_done(),
		 * does the same.  This percolates all the way up to the root.
		 * The root i/o will reexecute or suspend the entire tree.
		 *
		 * This approach ensures that zio_reexecute() honors
		 * all the original i/o dependency relationships, e.g.
		 * parents not executing until children are ready.
		 */
		ASSERT(zio->io_child_type == ZIO_CHILD_LOGICAL);

		zio->io_gang_leader = NULL;

		mutex_enter(&zio->io_lock);
		zio->io_state[ZIO_WAIT_DONE] = 1;
		mutex_exit(&zio->io_lock);

		/*
		 * "The Godfather" I/O monitors its children but is
		 * not a true parent to them. It will track them through
		 * the pipeline but severs its ties whenever they get into
		 * trouble (e.g. suspended). This allows "The Godfather"
		 * I/O to return status without blocking.
		 */
		zl = NULL;
		for (pio = zio_walk_parents(zio, &zl); pio != NULL;
		    pio = pio_next) {
			zio_link_t *remove_zl = zl;
			pio_next = zio_walk_parents(zio, &zl);

			if ((pio->io_flags & ZIO_FLAG_GODFATHER) &&
			    (zio->io_reexecute & ZIO_REEXECUTE_SUSPEND)) {
				zio_remove_child(pio, zio, remove_zl);
				/*
				 * This is a rare code path, so we don't
				 * bother with "next_to_execute".
				 */
				zio_notify_parent(pio, zio, ZIO_WAIT_DONE,
				    NULL);
			}
		}

		if ((pio = zio_unique_parent(zio)) != NULL) {
			/*
			 * We're not a root i/o, so there's nothing to do
			 * but notify our parent.  Don't propagate errors
			 * upward since we haven't permanently failed yet.
			 */
			ASSERT(!(zio->io_flags & ZIO_FLAG_GODFATHER));
			zio->io_flags |= ZIO_FLAG_DONT_PROPAGATE;
			/*
			 * This is a rare code path, so we don't bother with
			 * "next_to_execute".
			 */
			zio_notify_parent(pio, zio, ZIO_WAIT_DONE, NULL);
		} else if (zio->io_reexecute & ZIO_REEXECUTE_SUSPEND) {
			/*
			 * We'd fail again if we reexecuted now, so suspend
			 * until conditions improve (e.g. device comes online).
			 */
			zio_suspend(zio->io_spa, zio, ZIO_SUSPEND_IOERR);
		} else {
			/*
			 * Reexecution is potentially a huge amount of work.
			 * Hand it off to the otherwise-unused claim taskq.
			 */
			ASSERT(taskq_empty_ent(&zio->io_tqent));
			spa_taskq_dispatch_ent(zio->io_spa,
			    ZIO_TYPE_CLAIM, ZIO_TASKQ_ISSUE,
			    (task_func_t *)zio_reexecute, zio, 0,
			    &zio->io_tqent);
		}
		return (NULL);
	}

	ASSERT(zio->io_child_count == 0);
	ASSERT(zio->io_reexecute == 0);
	ASSERT(zio->io_error == 0 || (zio->io_flags & ZIO_FLAG_CANFAIL));

	/*
	 * Report any checksum errors, since the I/O is complete.
	 */
	while (zio->io_cksum_report != NULL) {
		zio_cksum_report_t *zcr = zio->io_cksum_report;
		zio->io_cksum_report = zcr->zcr_next;
		zcr->zcr_next = NULL;
		zcr->zcr_finish(zcr, NULL);
		zfs_ereport_free_checksum(zcr);
	}

	if (zio->io_flags & ZIO_FLAG_FASTWRITE && zio->io_bp &&
	    !BP_IS_HOLE(zio->io_bp) && !BP_IS_EMBEDDED(zio->io_bp) &&
	    !(zio->io_flags & ZIO_FLAG_NOPWRITE)) {
		metaslab_fastwrite_unmark(zio->io_spa, zio->io_bp);
	}

	/*
	 * It is the responsibility of the done callback to ensure that this
	 * particular zio is no longer discoverable for adoption, and as
	 * such, cannot acquire any new parents.
	 */
	if (zio->io_done)
		zio->io_done(zio);

	mutex_enter(&zio->io_lock);
	zio->io_state[ZIO_WAIT_DONE] = 1;
	mutex_exit(&zio->io_lock);

	/*
	 * We are done executing this zio.  We may want to execute a parent
	 * next.  See the comment in zio_notify_parent().
	 */
	zio_t *next_to_execute = NULL;
	zl = NULL;
	for (pio = zio_walk_parents(zio, &zl); pio != NULL; pio = pio_next) {
		zio_link_t *remove_zl = zl;
		pio_next = zio_walk_parents(zio, &zl);
		zio_remove_child(pio, zio, remove_zl);
		zio_notify_parent(pio, zio, ZIO_WAIT_DONE, &next_to_execute);
	}

	if (zio->io_waiter != NULL) {
		mutex_enter(&zio->io_lock);
		zio->io_executor = NULL;
		cv_broadcast(&zio->io_cv);
		mutex_exit(&zio->io_lock);
	} else {
		zio_destroy(zio);
	}

	return (next_to_execute);
}

/*
 * ==========================================================================
 * I/O pipeline definition
 * ==========================================================================
 */
static zio_pipe_stage_t *zio_pipeline[] = {
	NULL,
	zio_read_bp_init,
	zio_write_bp_init,
	zio_free_bp_init,
	zio_issue_async,
	zio_write_compress,
	zio_encrypt,
	zio_checksum_generate,
	zio_nop_write,
	zio_ddt_read_start,
	zio_ddt_read_done,
	zio_ddt_write,
	zio_ddt_free,
	zio_gang_assemble,
	zio_gang_issue,
	zio_dva_throttle,
	zio_dva_allocate,
	zio_dva_free,
	zio_dva_claim,
	zio_ready,
	zio_vdev_io_start,
	zio_vdev_io_done,
	zio_vdev_io_assess,
	zio_checksum_verify,
	zio_done
};




/*
 * Compare two zbookmark_phys_t's to see which we would reach first in a
 * pre-order traversal of the object tree.
 *
 * This is simple in every case aside from the meta-dnode object. For all other
 * objects, we traverse them in order (object 1 before object 2, and so on).
 * However, all of these objects are traversed while traversing object 0, since
 * the data it points to is the list of objects.  Thus, we need to convert to a
 * canonical representation so we can compare meta-dnode bookmarks to
 * non-meta-dnode bookmarks.
 *
 * We do this by calculating "equivalents" for each field of the zbookmark.
 * zbookmarks outside of the meta-dnode use their own object and level, and
 * calculate the level 0 equivalent (the first L0 blkid that is contained in the
 * blocks this bookmark refers to) by multiplying their blkid by their span
 * (the number of L0 blocks contained within one block at their level).
 * zbookmarks inside the meta-dnode calculate their object equivalent
 * (which is L0equiv * dnodes per data block), use 0 for their L0equiv, and use
 * level + 1<<31 (any value larger than a level could ever be) for their level.
 * This causes them to always compare before a bookmark in their object
 * equivalent, compare appropriately to bookmarks in other objects, and to
 * compare appropriately to other bookmarks in the meta-dnode.
 */
int
zbookmark_compare(uint16_t dbss1, uint8_t ibs1, uint16_t dbss2, uint8_t ibs2,
    const zbookmark_phys_t *zb1, const zbookmark_phys_t *zb2)
{
	/*
	 * These variables represent the "equivalent" values for the zbookmark,
	 * after converting zbookmarks inside the meta dnode to their
	 * normal-object equivalents.
	 */
	uint64_t zb1obj, zb2obj;
	uint64_t zb1L0, zb2L0;
	uint64_t zb1level, zb2level;

	if (zb1->zb_object == zb2->zb_object &&
	    zb1->zb_level == zb2->zb_level &&
	    zb1->zb_blkid == zb2->zb_blkid)
		return (0);

	IMPLY(zb1->zb_level > 0, ibs1 >= SPA_MINBLOCKSHIFT);
	IMPLY(zb2->zb_level > 0, ibs2 >= SPA_MINBLOCKSHIFT);

	/*
	 * BP_SPANB calculates the span in blocks.
	 */
	zb1L0 = (zb1->zb_blkid) * BP_SPANB(ibs1, zb1->zb_level);
	zb2L0 = (zb2->zb_blkid) * BP_SPANB(ibs2, zb2->zb_level);

	if (zb1->zb_object == DMU_META_DNODE_OBJECT) {
		zb1obj = zb1L0 * (dbss1 << (SPA_MINBLOCKSHIFT - DNODE_SHIFT));
		zb1L0 = 0;
		zb1level = zb1->zb_level + COMPARE_META_LEVEL;
	} else {
		zb1obj = zb1->zb_object;
		zb1level = zb1->zb_level;
	}

	if (zb2->zb_object == DMU_META_DNODE_OBJECT) {
		zb2obj = zb2L0 * (dbss2 << (SPA_MINBLOCKSHIFT - DNODE_SHIFT));
		zb2L0 = 0;
		zb2level = zb2->zb_level + COMPARE_META_LEVEL;
	} else {
		zb2obj = zb2->zb_object;
		zb2level = zb2->zb_level;
	}

	/* Now that we have a canonical representation, do the comparison. */
	if (zb1obj != zb2obj)
		return (zb1obj < zb2obj ? -1 : 1);
	else if (zb1L0 != zb2L0)
		return (zb1L0 < zb2L0 ? -1 : 1);
	else if (zb1level != zb2level)
		return (zb1level > zb2level ? -1 : 1);
	/*
	 * This can (theoretically) happen if the bookmarks have the same object
	 * and level, but different blkids, if the block sizes are not the same.
	 * There is presently no way to change the indirect block sizes
	 */
	return (0);
}

/*
 *  This function checks the following: given that last_block is the place that
 *  our traversal stopped last time, does that guarantee that we've visited
 *  every node under subtree_root?  Therefore, we can't just use the raw output
 *  of zbookmark_compare.  We have to pass in a modified version of
 *  subtree_root; by incrementing the block id, and then checking whether
 *  last_block is before or equal to that, we can tell whether or not having
 *  visited last_block implies that all of subtree_root's children have been
 *  visited.
 */
boolean_t
zbookmark_subtree_completed(const dnode_phys_t *dnp,
    const zbookmark_phys_t *subtree_root, const zbookmark_phys_t *last_block)
{
	zbookmark_phys_t mod_zb = *subtree_root;
	mod_zb.zb_blkid++;
	ASSERT(last_block->zb_level == 0);

	/* The objset_phys_t isn't before anything. */
	if (dnp == NULL)
		return (B_FALSE);

	/*
	 * We pass in 1ULL << (DNODE_BLOCK_SHIFT - SPA_MINBLOCKSHIFT) for the
	 * data block size in sectors, because that variable is only used if
	 * the bookmark refers to a block in the meta-dnode.  Since we don't
	 * know without examining it what object it refers to, and there's no
	 * harm in passing in this value in other cases, we always pass it in.
	 *
	 * We pass in 0 for the indirect block size shift because zb2 must be
	 * level 0.  The indirect block size is only used to calculate the span
	 * of the bookmark, but since the bookmark must be level 0, the span is
	 * always 1, so the math works out.
	 *
	 * If you make changes to how the zbookmark_compare code works, be sure
	 * to make sure that this code still works afterwards.
	 */
	return (zbookmark_compare(dnp->dn_datablkszsec, dnp->dn_indblkshift,
	    1ULL << (DNODE_BLOCK_SHIFT - SPA_MINBLOCKSHIFT), 0, &mod_zb,
	    last_block) <= 0);
}

#if defined(_KERNEL)
EXPORT_SYMBOL(zio_type_name);
EXPORT_SYMBOL(zio_buf_alloc);
EXPORT_SYMBOL(zio_data_buf_alloc);
EXPORT_SYMBOL(zio_buf_free);
EXPORT_SYMBOL(zio_data_buf_free);

module_param(zio_slow_io_ms, int, 0644);
MODULE_PARM_DESC(zio_slow_io_ms,
	"Max I/O completion time (milliseconds) before marking it as slow");

module_param(zio_requeue_io_start_cut_in_line, int, 0644);
MODULE_PARM_DESC(zio_requeue_io_start_cut_in_line, "Prioritize requeued I/O");

module_param(zfs_sync_pass_deferred_free, int, 0644);
MODULE_PARM_DESC(zfs_sync_pass_deferred_free,
	"Defer frees starting in this pass");

module_param(zfs_sync_pass_dont_compress, int, 0644);
MODULE_PARM_DESC(zfs_sync_pass_dont_compress,
	"Don't compress starting in this pass");

module_param(zfs_sync_pass_rewrite, int, 0644);
MODULE_PARM_DESC(zfs_sync_pass_rewrite,
	"Rewrite new bps starting in this pass");

module_param(zio_dva_throttle_enabled, int, 0644);
MODULE_PARM_DESC(zio_dva_throttle_enabled,
	"Throttle block allocations in the ZIO pipeline");

module_param(zio_deadman_log_all, int, 0644);
MODULE_PARM_DESC(zio_deadman_log_all,
	"Log all slow ZIOs, not just those with vdevs");
#endif