1. Ruslan Osmanov
  2. pecl-eio

Source

pecl-eio / libeio / eio.c

   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
/*
 * libeio implementation
 *
 * Copyright (c) 2007,2008,2009,2010,2011,2012 Marc Alexander Lehmann <libeio@schmorp.de>
 * All rights reserved.
 *
 * Redistribution and use in source and binary forms, with or without modifica-
 * tion, are permitted provided that the following conditions are met:
 * 
 *   1.  Redistributions of source code must retain the above copyright notice,
 *       this list of conditions and the following disclaimer.
 * 
 *   2.  Redistributions in binary form must reproduce the above copyright
 *       notice, this list of conditions and the following disclaimer in the
 *       documentation and/or other materials provided with the distribution.
 * 
 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR IMPLIED
 * WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MER-
 * CHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.  IN NO
 * EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPE-
 * CIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO,
 * PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS;
 * OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY,
 * WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTH-
 * ERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED
 * OF THE POSSIBILITY OF SUCH DAMAGE.
 *
 * Alternatively, the contents of this file may be used under the terms of
 * the GNU General Public License ("GPL") version 2 or any later version,
 * in which case the provisions of the GPL are applicable instead of
 * the above. If you wish to allow the use of your version of this file
 * only under the terms of the GPL and not to allow others to use your
 * version of this file under the BSD license, indicate your decision
 * by deleting the provisions above and replace them with the notice
 * and other provisions required by the GPL. If you do not delete the
 * provisions above, a recipient may use your version of this file under
 * either the BSD or the GPL.
 */

#ifndef _WIN32
# include "config.h"
#endif

#include "eio.h"
#include "ecb.h"

#ifdef EIO_STACKSIZE
# define X_STACKSIZE EIO_STACKSIZE
#endif
#include "xthread.h"

#include <errno.h>
#include <stddef.h>
#include <stdlib.h>
#include <string.h>
#include <errno.h>
#include <sys/types.h>
#include <sys/stat.h>
#include <limits.h>
#include <fcntl.h>
#include <assert.h>

/* intptr_t comes from unistd.h, says POSIX/UNIX/tradition */
/* intptr_t only comes from stdint.h, says idiot openbsd coder */
#if HAVE_STDINT_H
# include <stdint.h>
#endif

#ifndef ECANCELED
# define ECANCELED EDOM
#endif
#ifndef ELOOP
# define ELOOP EDOM
#endif

#if !defined(ENOTSOCK) && defined(WSAENOTSOCK)
# define ENOTSOCK WSAENOTSOCK
#endif

static void eio_destroy (eio_req *req);

#ifndef EIO_FINISH
# define EIO_FINISH(req)  ((req)->finish) && !EIO_CANCELLED (req) ? (req)->finish (req) : 0
#endif

#ifndef EIO_DESTROY
# define EIO_DESTROY(req) do { if ((req)->destroy) (req)->destroy (req); } while (0)
#endif

#ifndef EIO_FEED
# define EIO_FEED(req)    do { if ((req)->feed   ) (req)->feed    (req); } while (0)
#endif

#ifndef EIO_FD_TO_WIN32_HANDLE
# define EIO_FD_TO_WIN32_HANDLE(fd) _get_osfhandle (fd)
#endif
#ifndef EIO_WIN32_HANDLE_TO_FD
# define EIO_WIN32_HANDLE_TO_FD(handle) _open_osfhandle (handle, 0)
#endif

#define EIO_ERRNO(errval,retval) ((errno = errval), retval)

#define EIO_ENOSYS() EIO_ERRNO (ENOSYS, -1)

#ifdef _WIN32

  #undef PAGESIZE
  #define PAGESIZE 4096 /* GetSystemInfo? */

  /* TODO: look at how perl does stat (non-sloppy), unlink (ro-files), utime, link */

  #ifdef EIO_STRUCT_STATI64
    /* look at perl's non-sloppy stat */
    #define stat(path,buf)       _stati64 (path,buf)
    #define fstat(fd,buf)        _fstati64 (fd,buf)
  #endif
  #define lstat(path,buf)      stat (path,buf)
  #define fsync(fd)            (FlushFileBuffers ((HANDLE)EIO_FD_TO_WIN32_HANDLE (fd)) ? 0 : EIO_ERRNO (EBADF, -1))
  #define mkdir(path,mode)     _mkdir (path)
  #define link(old,neu)        (CreateHardLink (neu, old, 0) ? 0 : EIO_ERRNO (ENOENT, -1))

  #define chmod(path,mode)     _chmod (path, mode)
  #define dup(fd)              _dup (fd)
  #define dup2(fd1,fd2)        _dup2 (fd1, fd2)

  #define fchmod(fd,mode)      EIO_ENOSYS ()
  #define chown(path,uid,gid)  EIO_ENOSYS ()
  #define fchown(fd,uid,gid)   EIO_ENOSYS ()
  #define truncate(path,offs)  EIO_ENOSYS () /* far-miss: SetEndOfFile */
  #define ftruncate(fd,offs)   EIO_ENOSYS () /* near-miss: SetEndOfFile */
  #define mknod(path,mode,dev) EIO_ENOSYS ()
  #define sync()               EIO_ENOSYS ()
  #define readlink(path,buf,s) EIO_ENOSYS ()
  #define statvfs(path,buf)    EIO_ENOSYS ()
  #define fstatvfs(fd,buf)     EIO_ENOSYS ()

  /* rename() uses MoveFile, which fails to overwrite */
  #define rename(old,neu)      eio__rename (old, neu)

  static int
  eio__rename (const char *old, const char *neu)
  {
    if (MoveFileEx (old, neu, MOVEFILE_REPLACE_EXISTING))
      return 0;

    /* should steal _dosmaperr */
    switch (GetLastError ())
      {
        case ERROR_FILE_NOT_FOUND:
        case ERROR_PATH_NOT_FOUND:
        case ERROR_INVALID_DRIVE:
        case ERROR_NO_MORE_FILES:
        case ERROR_BAD_NETPATH:
        case ERROR_BAD_NET_NAME:
        case ERROR_BAD_PATHNAME:
        case ERROR_FILENAME_EXCED_RANGE:
          errno = ENOENT;
          break;

        default:
          errno = EACCES;
          break;
      }

    return -1;
  }

  /* we could even stat and see if it exists */
  static int
  symlink (const char *old, const char *neu)
  {
    #if WINVER >= 0x0600
      if (CreateSymbolicLink (neu, old, 1))
        return 0;

      if (CreateSymbolicLink (neu, old, 0))
        return 0;
    #endif

    return EIO_ERRNO (ENOENT, -1);
  }

  /* POSIX API only */
  #define CreateHardLink(neu,old,flags) 0
  #define CreateSymbolicLink(neu,old,flags) 0

  struct statvfs
  {
    int dummy;
  };

  #define DT_DIR EIO_DT_DIR
  #define DT_REG EIO_DT_REG
  #define D_NAME(entp) entp.cFileName
  #define D_TYPE(entp) (entp.dwFileAttributes & FILE_ATTRIBUTE_DIRECTORY ? DT_DIR : DT_REG)

#else

  #include <sys/time.h>
  #include <sys/select.h>
  #include <sys/statvfs.h>
  #include <unistd.h>
  #include <signal.h>
  #include <dirent.h>

  #if _POSIX_MEMLOCK || _POSIX_MEMLOCK_RANGE || _POSIX_MAPPED_FILES
    #include <sys/mman.h>
  #endif

  #define D_NAME(entp) entp->d_name

  /* POSIX_SOURCE is useless on bsd's, and XOPEN_SOURCE is unreliable there, too */
  #if __FreeBSD__ || __NetBSD__ || __OpenBSD__
    #define _DIRENT_HAVE_D_TYPE /* sigh */
    #define D_INO(de) (de)->d_fileno
    #define D_NAMLEN(de) (de)->d_namlen
  #elif __linux || defined d_ino || _XOPEN_SOURCE >= 600
    #define D_INO(de) (de)->d_ino
  #endif

  #ifdef _D_EXACT_NAMLEN
    #undef D_NAMLEN
    #define D_NAMLEN(de) _D_EXACT_NAMLEN (de)
  #endif

  #ifdef _DIRENT_HAVE_D_TYPE
    #define D_TYPE(de) (de)->d_type
  #endif

  #ifndef EIO_STRUCT_DIRENT
    #define EIO_STRUCT_DIRENT struct dirent
  #endif

#endif

#if HAVE_UTIMES
# include <utime.h>
#endif

#if HAVE_SYS_SYSCALL_H
# include <sys/syscall.h>
#endif

#if HAVE_SYS_PRCTL_H
# include <sys/prctl.h>
#endif

#if HAVE_SENDFILE
# if __linux
#  include <sys/sendfile.h>
# elif __FreeBSD__ || defined __APPLE__
#  include <sys/socket.h>
#  include <sys/uio.h>
# elif __hpux
#  include <sys/socket.h>
# elif __solaris
#  include <sys/sendfile.h>
# else
#  error sendfile support requested but not available
# endif
#endif

#ifndef D_TYPE
# define D_TYPE(de) 0
#endif
#ifndef D_INO
# define D_INO(de) 0
#endif
#ifndef D_NAMLEN
# define D_NAMLEN(entp) strlen (D_NAME (entp))
#endif

/* used for struct dirent, AIX doesn't provide it */
#ifndef NAME_MAX
# define NAME_MAX 4096
#endif

/* used for readlink etc. */
#ifndef PATH_MAX
# define PATH_MAX 4096
#endif

/* buffer size for various temporary buffers */
#define EIO_BUFSIZE 65536

#define dBUF	 				\
  char *eio_buf = malloc (EIO_BUFSIZE);		\
  errno = ENOMEM;				\
  if (!eio_buf)					\
    return -1

#define FUBd					\
  free (eio_buf)

#define EIO_TICKS ((1000000 + 1023) >> 10)

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

struct tmpbuf
{
  void *ptr;
  int len;
};

static void *
tmpbuf_get (struct tmpbuf *buf, int len)
{
  if (buf->len < len)
    {
      free (buf->ptr);
      buf->ptr = malloc (buf->len = len);
    }

  return buf->ptr;
}

struct tmpbuf;

#if _POSIX_VERSION >= 200809L
  #define HAVE_AT 1
  #define WD2FD(wd) ((wd) ? (wd)->fd : AT_FDCWD)
  #ifndef O_SEARCH
    #define O_SEARCH O_RDONLY
  #endif
#else
  #define HAVE_AT 0
  static const char *wd_expand (struct tmpbuf *tmpbuf, eio_wd wd, const char *path);
#endif

struct eio_pwd
{
#if HAVE_AT
  int fd;
#endif
  int len;
  char str[1]; /* actually, a 0-terminated canonical path */
};

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

#define ETP_PRI_MIN EIO_PRI_MIN
#define ETP_PRI_MAX EIO_PRI_MAX

struct etp_worker;

#define ETP_REQ eio_req
#define ETP_DESTROY(req) eio_destroy (req)
static int eio_finish (eio_req *req);
#define ETP_FINISH(req)  eio_finish (req)
static void eio_execute (struct etp_worker *self, eio_req *req);
#define ETP_EXECUTE(wrk,req) eio_execute (wrk,req)

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

#define ETP_NUM_PRI (ETP_PRI_MAX - ETP_PRI_MIN + 1)

/* calculate time difference in ~1/EIO_TICKS of a second */
ecb_inline int
tvdiff (struct timeval *tv1, struct timeval *tv2)
{
  return  (tv2->tv_sec  - tv1->tv_sec ) * EIO_TICKS
       + ((tv2->tv_usec - tv1->tv_usec) >> 10);
}

static unsigned int started, idle, wanted = 4;

static void (*want_poll_cb) (void);
static void (*done_poll_cb) (void);
 
static unsigned int max_poll_time;     /* reslock */
static unsigned int max_poll_reqs;     /* reslock */

static unsigned int nreqs;    /* reqlock */
static unsigned int nready;   /* reqlock */
static unsigned int npending; /* reqlock */
static unsigned int max_idle = 4;      /* maximum number of threads that can idle indefinitely */
static unsigned int idle_timeout = 10; /* number of seconds after which an idle threads exit */

static xmutex_t wrklock;
static xmutex_t reslock;
static xmutex_t reqlock;
static xcond_t  reqwait;

typedef struct etp_worker
{
  struct tmpbuf tmpbuf;

  /* locked by wrklock */
  struct etp_worker *prev, *next;

  xthread_t tid;

#ifdef ETP_WORKER_COMMON
  ETP_WORKER_COMMON
#endif
} etp_worker;

static etp_worker wrk_first; /* NOT etp */

#define ETP_WORKER_LOCK(wrk)   X_LOCK   (wrklock)
#define ETP_WORKER_UNLOCK(wrk) X_UNLOCK (wrklock)

/* worker threads management */

static void
etp_worker_clear (etp_worker *wrk)
{
}

static void ecb_cold
etp_worker_free (etp_worker *wrk)
{
  free (wrk->tmpbuf.ptr);

  wrk->next->prev = wrk->prev;
  wrk->prev->next = wrk->next;

  free (wrk);
}

static unsigned int
etp_nreqs (void)
{
  int retval;
  if (WORDACCESS_UNSAFE) X_LOCK   (reqlock);
  retval = nreqs;
  if (WORDACCESS_UNSAFE) X_UNLOCK (reqlock);
  return retval;
}

static unsigned int
etp_nready (void)
{
  unsigned int retval;

  if (WORDACCESS_UNSAFE) X_LOCK   (reqlock);
  retval = nready;
  if (WORDACCESS_UNSAFE) X_UNLOCK (reqlock);

  return retval;
}

static unsigned int
etp_npending (void)
{
  unsigned int retval;

  if (WORDACCESS_UNSAFE) X_LOCK   (reqlock);
  retval = npending;
  if (WORDACCESS_UNSAFE) X_UNLOCK (reqlock);

  return retval;
}

static unsigned int
etp_nthreads (void)
{
  unsigned int retval;

  if (WORDACCESS_UNSAFE) X_LOCK   (reqlock);
  retval = started;
  if (WORDACCESS_UNSAFE) X_UNLOCK (reqlock);

  return retval;
}

/*
 * a somewhat faster data structure might be nice, but
 * with 8 priorities this actually needs <20 insns
 * per shift, the most expensive operation.
 */
typedef struct {
  ETP_REQ *qs[ETP_NUM_PRI], *qe[ETP_NUM_PRI]; /* qstart, qend */
  int size;
} etp_reqq;

static etp_reqq req_queue;
static etp_reqq res_queue;

static void ecb_noinline ecb_cold
reqq_init (etp_reqq *q)
{
  int pri;

  for (pri = 0; pri < ETP_NUM_PRI; ++pri)
    q->qs[pri] = q->qe[pri] = 0;

  q->size = 0;
}

static int ecb_noinline
reqq_push (etp_reqq *q, ETP_REQ *req)
{
  int pri = req->pri;
  req->next = 0;

  if (q->qe[pri])
    {
      q->qe[pri]->next = req;
      q->qe[pri] = req;
    }
  else
    q->qe[pri] = q->qs[pri] = req;

  return q->size++;
}

static ETP_REQ * ecb_noinline
reqq_shift (etp_reqq *q)
{
  int pri;

  if (!q->size)
    return 0;

  --q->size;

  for (pri = ETP_NUM_PRI; pri--; )
    {
      eio_req *req = q->qs[pri];

      if (req)
        {
          if (!(q->qs[pri] = (eio_req *)req->next))
            q->qe[pri] = 0;

          return req;
        }
    }

  abort ();
}

static int ecb_cold
etp_init (void (*want_poll)(void), void (*done_poll)(void))
{
  X_MUTEX_CREATE (wrklock);
  X_MUTEX_CREATE (reslock);
  X_MUTEX_CREATE (reqlock);
  X_COND_CREATE  (reqwait);

  reqq_init (&req_queue);
  reqq_init (&res_queue);

  wrk_first.next =
  wrk_first.prev = &wrk_first;

  started  = 0;
  idle     = 0;
  nreqs    = 0;
  nready   = 0;
  npending = 0;

  want_poll_cb = want_poll;
  done_poll_cb = done_poll;

  return 0;
}

X_THREAD_PROC (etp_proc);

static void ecb_cold
etp_start_thread (void)
{
  etp_worker *wrk = calloc (1, sizeof (etp_worker));

  /*TODO*/
  assert (("unable to allocate worker thread data", wrk));

  X_LOCK (wrklock);

  if (thread_create (&wrk->tid, etp_proc, (void *)wrk))
    {
      wrk->prev = &wrk_first;
      wrk->next = wrk_first.next;
      wrk_first.next->prev = wrk;
      wrk_first.next = wrk;
      ++started;
    }
  else
    free (wrk);

  X_UNLOCK (wrklock);
}

static void
etp_maybe_start_thread (void)
{
  if (ecb_expect_true (etp_nthreads () >= wanted))
    return;
  
  /* todo: maybe use idle here, but might be less exact */
  if (ecb_expect_true (0 <= (int)etp_nthreads () + (int)etp_npending () - (int)etp_nreqs ()))
    return;

  etp_start_thread ();
}

static void ecb_cold
etp_end_thread (void)
{
  eio_req *req = calloc (1, sizeof (eio_req)); /* will be freed by worker */

  req->type = -1;
  req->pri  = ETP_PRI_MAX - ETP_PRI_MIN;

  X_LOCK (reqlock);
  reqq_push (&req_queue, req);
  X_COND_SIGNAL (reqwait);
  X_UNLOCK (reqlock);

  X_LOCK (wrklock);
  --started;
  X_UNLOCK (wrklock);
}

static int
etp_poll (void)
{
  unsigned int maxreqs;
  unsigned int maxtime;
  struct timeval tv_start, tv_now;

  X_LOCK (reslock);
  maxreqs = max_poll_reqs;
  maxtime = max_poll_time;
  X_UNLOCK (reslock);

  if (maxtime)
    gettimeofday (&tv_start, 0);

  for (;;)
    {
      ETP_REQ *req;

      etp_maybe_start_thread ();

      X_LOCK (reslock);
      req = reqq_shift (&res_queue);

      if (req)
        {
          --npending;

          if (!res_queue.size && done_poll_cb)
            done_poll_cb ();
        }

      X_UNLOCK (reslock);

      if (!req)
        return 0;

      X_LOCK (reqlock);
      --nreqs;
      X_UNLOCK (reqlock);

      if (ecb_expect_false (req->type == EIO_GROUP && req->size))
        {
          req->int1 = 1; /* mark request as delayed */
          continue;
        }
      else
        {
          int res = ETP_FINISH (req);
          if (ecb_expect_false (res))
            return res;
        }

      if (ecb_expect_false (maxreqs && !--maxreqs))
        break;

      if (maxtime)
        {
          gettimeofday (&tv_now, 0);

          if (tvdiff (&tv_start, &tv_now) >= maxtime)
            break;
        }
    }

  errno = EAGAIN;
  return -1;
}

static void
etp_cancel (ETP_REQ *req)
{
  req->cancelled = 1;

  eio_grp_cancel (req);
}

static void
etp_submit (ETP_REQ *req)
{
  req->pri -= ETP_PRI_MIN;

  if (ecb_expect_false (req->pri < ETP_PRI_MIN - ETP_PRI_MIN)) req->pri = ETP_PRI_MIN - ETP_PRI_MIN;
  if (ecb_expect_false (req->pri > ETP_PRI_MAX - ETP_PRI_MIN)) req->pri = ETP_PRI_MAX - ETP_PRI_MIN;

  if (ecb_expect_false (req->type == EIO_GROUP))
    {
      /* I hope this is worth it :/ */
      X_LOCK (reqlock);
      ++nreqs;
      X_UNLOCK (reqlock);

      X_LOCK (reslock);

      ++npending;

      if (!reqq_push (&res_queue, req) && want_poll_cb)
        want_poll_cb ();

      X_UNLOCK (reslock);
    }
  else
    {
      X_LOCK (reqlock);
      ++nreqs;
      ++nready;
      reqq_push (&req_queue, req);
      X_COND_SIGNAL (reqwait);
      X_UNLOCK (reqlock);

      etp_maybe_start_thread ();
    }
}

static void ecb_cold
etp_set_max_poll_time (double nseconds)
{
  if (WORDACCESS_UNSAFE) X_LOCK   (reslock);
  max_poll_time = nseconds * EIO_TICKS;
  if (WORDACCESS_UNSAFE) X_UNLOCK (reslock);
}

static void ecb_cold
etp_set_max_poll_reqs (unsigned int maxreqs)
{
  if (WORDACCESS_UNSAFE) X_LOCK   (reslock);
  max_poll_reqs = maxreqs;
  if (WORDACCESS_UNSAFE) X_UNLOCK (reslock);
}

static void ecb_cold
etp_set_max_idle (unsigned int nthreads)
{
  if (WORDACCESS_UNSAFE) X_LOCK   (reqlock);
  max_idle = nthreads;
  if (WORDACCESS_UNSAFE) X_UNLOCK (reqlock);
}

static void ecb_cold
etp_set_idle_timeout (unsigned int seconds)
{
  if (WORDACCESS_UNSAFE) X_LOCK   (reqlock);
  idle_timeout = seconds;
  if (WORDACCESS_UNSAFE) X_UNLOCK (reqlock);
}

static void ecb_cold
etp_set_min_parallel (unsigned int nthreads)
{
  if (wanted < nthreads)
    wanted = nthreads;
}

static void ecb_cold
etp_set_max_parallel (unsigned int nthreads)
{
  if (wanted > nthreads)
    wanted = nthreads;

  while (started > wanted)
    etp_end_thread ();
}

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

static void
grp_try_feed (eio_req *grp)
{
  while (grp->size < grp->int2 && !EIO_CANCELLED (grp))
    {
      grp->flags &= ~EIO_FLAG_GROUPADD;

      EIO_FEED (grp);

      /* stop if no progress has been made */
      if (!(grp->flags & EIO_FLAG_GROUPADD))
        {
          grp->feed = 0;
          break;
        }
    }
}

static int
grp_dec (eio_req *grp)
{
  --grp->size;

  /* call feeder, if applicable */
  grp_try_feed (grp);

  /* finish, if done */
  if (!grp->size && grp->int1)
    return eio_finish (grp);
  else
    return 0;
}

static void
eio_destroy (eio_req *req)
{
  if ((req)->flags & EIO_FLAG_PTR1_FREE) free (req->ptr1);
  if ((req)->flags & EIO_FLAG_PTR2_FREE) free (req->ptr2);

  EIO_DESTROY (req);
}

static int
eio_finish (eio_req *req)
{
  int res = EIO_FINISH (req);

  if (req->grp)
    {
      int res2;
      eio_req *grp = req->grp;

      /* unlink request */
      if (req->grp_next) req->grp_next->grp_prev = req->grp_prev;
      if (req->grp_prev) req->grp_prev->grp_next = req->grp_next;

      if (grp->grp_first == req)
        grp->grp_first = req->grp_next;

      res2 = grp_dec (grp);

      if (!res)
        res = res2;
    }

  eio_destroy (req);

  return res;
}

void
eio_grp_cancel (eio_req *grp)
{
  for (grp = grp->grp_first; grp; grp = grp->grp_next)
    eio_cancel (grp);
}

void
eio_cancel (eio_req *req)
{
  etp_cancel (req);
}

void
eio_submit (eio_req *req)
{
  etp_submit (req);
}

unsigned int
eio_nreqs (void)
{
  return etp_nreqs ();
}

unsigned int
eio_nready (void)
{
  return etp_nready ();
}

unsigned int
eio_npending (void)
{
  return etp_npending ();
}

unsigned int ecb_cold
eio_nthreads (void)
{
  return etp_nthreads ();
}

void ecb_cold
eio_set_max_poll_time (double nseconds)
{
  etp_set_max_poll_time (nseconds);
}

void ecb_cold
eio_set_max_poll_reqs (unsigned int maxreqs)
{
  etp_set_max_poll_reqs (maxreqs);
}

void ecb_cold
eio_set_max_idle (unsigned int nthreads)
{
  etp_set_max_idle (nthreads);
}

void ecb_cold
eio_set_idle_timeout (unsigned int seconds)
{
  etp_set_idle_timeout (seconds);
}

void ecb_cold
eio_set_min_parallel (unsigned int nthreads)
{
  etp_set_min_parallel (nthreads);
}

void ecb_cold
eio_set_max_parallel (unsigned int nthreads)
{
  etp_set_max_parallel (nthreads);
}

int eio_poll (void)
{
  return etp_poll ();
}

/*****************************************************************************/
/* work around various missing functions */

#ifndef HAVE_UTIMES

# undef utimes
# define utimes(path,times)  eio__utimes (path, times)

static int
eio__utimes (const char *filename, const struct timeval times[2])
{
  if (times)
    {
      struct utimbuf buf;

      buf.actime  = times[0].tv_sec;
      buf.modtime = times[1].tv_sec;

      return utime (filename, &buf);
    }
  else
    return utime (filename, 0);
}

#endif

#ifndef HAVE_FUTIMES

# undef futimes
# define futimes(fd,times) eio__futimes (fd, times)

static int
eio__futimes (int fd, const struct timeval tv[2])
{
  errno = ENOSYS;
  return -1;
}

#endif

#if !HAVE_FDATASYNC
# undef fdatasync
# define fdatasync(fd) fsync (fd)
#endif

static int
eio__syncfs (int fd)
{
  int res;

#if HAVE_SYS_SYNCFS
  res = (int)syscall (__NR_syncfs, (int)(fd));
#else
  res = EIO_ENOSYS ();
#endif

  if (res < 0 && errno == ENOSYS && fd >= 0)
    sync ();

  return res;
}

/* sync_file_range always needs emulation */
static int
eio__sync_file_range (int fd, off_t offset, size_t nbytes, unsigned int flags)
{
#if HAVE_SYNC_FILE_RANGE
  int res;

  if (EIO_SYNC_FILE_RANGE_WAIT_BEFORE   != SYNC_FILE_RANGE_WAIT_BEFORE
      || EIO_SYNC_FILE_RANGE_WRITE      != SYNC_FILE_RANGE_WRITE
      || EIO_SYNC_FILE_RANGE_WAIT_AFTER != SYNC_FILE_RANGE_WAIT_AFTER)
    {
      flags = 0
         | (flags & EIO_SYNC_FILE_RANGE_WAIT_BEFORE ? SYNC_FILE_RANGE_WAIT_BEFORE : 0)
         | (flags & EIO_SYNC_FILE_RANGE_WRITE       ? SYNC_FILE_RANGE_WRITE       : 0)
         | (flags & EIO_SYNC_FILE_RANGE_WAIT_AFTER  ? SYNC_FILE_RANGE_WAIT_AFTER  : 0);
    }

  res = sync_file_range (fd, offset, nbytes, flags);

  if (!res || errno != ENOSYS)
    return res;
#endif

  /* even though we could play tricks with the flags, it's better to always
   * call fdatasync, as that matches the expectation of its users best */
  return fdatasync (fd);
}

static int
eio__fallocate (int fd, int mode, off_t offset, size_t len)
{
#if HAVE_LINUX_FALLOCATE
  return fallocate (fd, mode, offset, len);
#else
  return EIO_ENOSYS ();
#endif
}

#if !HAVE_READAHEAD
# undef readahead
# define readahead(fd,offset,count) eio__readahead (fd, offset, count, self)

static eio_ssize_t
eio__readahead (int fd, off_t offset, size_t count, etp_worker *self)
{
  size_t todo = count;
  dBUF;

  while (todo > 0)
    {
      size_t len = todo < EIO_BUFSIZE ? todo : EIO_BUFSIZE;

      pread (fd, eio_buf, len, offset);
      offset += len;
      todo   -= len;
    }

  FUBd;

  /* linux's readahead basically only fails for EBADF or EINVAL (not mmappable) */
  /* but not for e.g. EIO or eof, so we also never fail */
  return 0;
}

#endif

/* sendfile always needs emulation */
static eio_ssize_t
eio__sendfile (int ofd, int ifd, off_t offset, size_t count)
{
  eio_ssize_t written = 0;
  eio_ssize_t res;

  if (!count)
    return 0;

  for (;;)
    {
#ifdef __APPLE__
# undef HAVE_SENDFILE /* broken, as everything on os x */
#endif
#if HAVE_SENDFILE
# if __linux
      off_t soffset = offset;
      res = sendfile (ofd, ifd, &soffset, count);

# elif __FreeBSD__
      /*
       * Of course, the freebsd sendfile is a dire hack with no thoughts
       * wasted on making it similar to other I/O functions.
       */
      off_t sbytes;
      res = sendfile (ifd, ofd, offset, count, 0, &sbytes, 0);

      #if 0 /* according to the manpage, this is correct, but broken behaviour */
      /* freebsd' sendfile will return 0 on success */
      /* freebsd 8 documents it as only setting *sbytes on EINTR and EAGAIN, but */
      /* not on e.g. EIO or EPIPE - sounds broken */
      if ((res < 0 && (errno == EAGAIN || errno == EINTR) && sbytes) || res == 0)
        res = sbytes;
      #endif

      /* according to source inspection, this is correct, and useful behaviour */
      if (sbytes)
        res = sbytes;

# elif defined __APPLE__
      off_t sbytes = count;
      res = sendfile (ifd, ofd, offset, &sbytes, 0, 0);

      /* according to the manpage, sbytes is always valid */
      if (sbytes)
        res = sbytes;

# elif __hpux
      res = sendfile (ofd, ifd, offset, count, 0, 0);

# elif __solaris
      struct sendfilevec vec;
      size_t sbytes;

      vec.sfv_fd   = ifd;
      vec.sfv_flag = 0;
      vec.sfv_off  = offset;
      vec.sfv_len  = count;

      res = sendfilev (ofd, &vec, 1, &sbytes);

      if (res < 0 && sbytes)
        res = sbytes;

# endif

#elif defined (_WIN32) && 0
      /* does not work, just for documentation of what would need to be done */
      /* actually, cannot be done like this, as TransmitFile changes the file offset, */
      /* libeio guarantees that the file offset does not change, and windows */
      /* has no way to get an independent handle to the same file description */
      HANDLE h = TO_SOCKET (ifd);
      SetFilePointer (h, offset, 0, FILE_BEGIN);
      res = TransmitFile (TO_SOCKET (ofd), h, count, 0, 0, 0, 0);

#else
      res = EIO_ENOSYS ();
#endif

      /* we assume sendfile can copy at least 128mb in one go */
      if (res <= 128 * 1024 * 1024)
        {
          if (res > 0)
            written += res;

          if (written)
            return written;

          break;
        }
      else
        {
          /* if we requested more, then probably the kernel was lazy */
          written += res;
          offset  += res;
          count   -= res;

          if (!count)
            return written;
        }
    }

  if (res < 0
      && (errno == ENOSYS || errno == EINVAL || errno == ENOTSOCK
          /* BSDs */
#ifdef ENOTSUP /* sigh, if the steenking pile called openbsd would only try to at least compile posix code... */
          || errno == ENOTSUP
#endif
#ifdef EOPNOTSUPP /* windows */
          || errno == EOPNOTSUPP /* BSDs */
#endif
#if __solaris
          || errno == EAFNOSUPPORT || errno == EPROTOTYPE
#endif
         )
      )
    {
      /* emulate sendfile. this is a major pain in the ass */
      dBUF;

      res = 0;

      while (count)
        {
          eio_ssize_t cnt;
          
          cnt = pread (ifd, eio_buf, count > EIO_BUFSIZE ? EIO_BUFSIZE : count, offset);

          if (cnt <= 0)
            {
              if (cnt && !res) res = -1;
              break;
            }

          cnt = write (ofd, eio_buf, cnt);

          if (cnt <= 0)
            {
              if (cnt && !res) res = -1;
              break;
            }

          offset += cnt;
          res    += cnt;
          count  -= cnt;
        }

      FUBd;
    }

  return res;
}

#ifdef PAGESIZE
# define eio_pagesize() PAGESIZE
#else
static intptr_t
eio_pagesize (void)
{
  static intptr_t page;

  if (!page)
    page = sysconf (_SC_PAGESIZE);

  return page;
}
#endif

static void
eio_page_align (void **addr, size_t *length)
{
  intptr_t mask = eio_pagesize () - 1;

  /* round down addr */
  intptr_t adj = mask & (intptr_t)*addr;

  *addr   = (void *)((intptr_t)*addr - adj);
  *length += adj;

  /* round up length */
  *length = (*length + mask) & ~mask;
}

#if !_POSIX_MEMLOCK
# define eio__mlockall(a) EIO_ENOSYS ()
#else

static int
eio__mlockall (int flags)
{
  #if __GLIBC__ == 2 && __GLIBC_MINOR__ <= 7
    extern int mallopt (int, int);
    mallopt (-6, 238); /* http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=473812 */
  #endif

  if (EIO_MCL_CURRENT   != MCL_CURRENT
      || EIO_MCL_FUTURE != MCL_FUTURE)
    {
      flags = 0
         | (flags & EIO_MCL_CURRENT ? MCL_CURRENT : 0)
         | (flags & EIO_MCL_FUTURE  ? MCL_FUTURE : 0);
    }

  return mlockall (flags);
}
#endif

#if !_POSIX_MEMLOCK_RANGE
# define eio__mlock(a,b) EIO_ENOSYS ()
#else

static int
eio__mlock (void *addr, size_t length)
{
  eio_page_align (&addr, &length);

  return mlock (addr, length);
}

#endif

#if !(_POSIX_MAPPED_FILES && _POSIX_SYNCHRONIZED_IO)
# define eio__msync(a,b,c) EIO_ENOSYS ()
#else

static int
eio__msync (void *mem, size_t len, int flags)
{
  eio_page_align (&mem, &len);

  if (EIO_MS_ASYNC         != MS_SYNC
      || EIO_MS_INVALIDATE != MS_INVALIDATE
      || EIO_MS_SYNC       != MS_SYNC)
    {
      flags = 0
         | (flags & EIO_MS_ASYNC      ? MS_ASYNC : 0)
         | (flags & EIO_MS_INVALIDATE ? MS_INVALIDATE : 0)
         | (flags & EIO_MS_SYNC       ? MS_SYNC : 0);
    }

  return msync (mem, len, flags);
}

#endif

static int
eio__mtouch (eio_req *req)
{
  void *mem  = req->ptr2;
  size_t len = req->size;
  int flags  = req->int1;

  eio_page_align (&mem, &len);

  {
    intptr_t addr = (intptr_t)mem;
    intptr_t end = addr + len;
    intptr_t page = eio_pagesize ();

    if (addr < end)
      if (flags & EIO_MT_MODIFY) /* modify */
        do { *((volatile sig_atomic_t *)addr) |= 0; } while ((addr += page) < len && !EIO_CANCELLED (req));
      else
        do { *((volatile sig_atomic_t *)addr)     ; } while ((addr += page) < len && !EIO_CANCELLED (req));
  }

  return 0;
}

/*****************************************************************************/
/* requests implemented outside eio_execute, because they are so large */

static void
eio__lseek (eio_req *req)
{
  /* this usually gets optimised away completely, or your compiler sucks, */
  /* or the whence constants really are not 0, 1, 2 */
  int whence = req->int2 == EIO_SEEK_SET ? SEEK_SET
             : req->int2 == EIO_SEEK_CUR ? SEEK_CUR
             : req->int2 == EIO_SEEK_END ? SEEK_END
             : req->int2;

  req->offs   = lseek (req->int1, req->offs, whence);
  req->result = req->offs == (off_t)-1 ? -1 : 0;
}

/* result will always end up in tmpbuf, there is always space for adding a 0-byte */
static int
eio__realpath (struct tmpbuf *tmpbuf, eio_wd wd, const char *path)
{
  const char *rel = path;
  char *res;
  char *tmp1, *tmp2;
#if SYMLOOP_MAX > 32
  int symlinks = SYMLOOP_MAX;
#else
  int symlinks = 32;
#endif

  errno = EINVAL;
  if (!rel)
    return -1;

  errno = ENOENT;
  if (!*rel)
    return -1;

  res  = tmpbuf_get (tmpbuf, PATH_MAX * 3);
  tmp1 = res  + PATH_MAX;
  tmp2 = tmp1 + PATH_MAX;

#if 0 /* disabled, the musl way to do things is just too racy */
#if __linux && defined(O_NONBLOCK) && defined(O_NOATIME)
  /* on linux we may be able to ask the kernel */
  {
    int fd = open (rel, O_RDONLY | O_NONBLOCK | O_NOCTTY | O_NOATIME);

    if (fd >= 0)
      {
        sprintf (tmp1, "/proc/self/fd/%d", fd);
        req->result = readlink (tmp1, res, PATH_MAX);
        close (fd);

        /* here we should probably stat the open file and the disk file, to make sure they still match */

        if (req->result > 0)
          goto done;
      }
    else if (errno == ELOOP || errno == ENAMETOOLONG || errno == ENOENT || errno == ENOTDIR || errno == EIO)
      return;
  }
#endif
#endif

  if (*rel != '/')
    {
      int len;

      errno = ENOENT;
      if (wd == EIO_INVALID_WD)
        return -1;
      
      if (wd == EIO_CWD)
        {
          if (!getcwd (res, PATH_MAX))
            return -1;

          len = strlen (res);
        }
      else
        memcpy (res, wd->str, len = wd->len);

      if (res [1]) /* only use if not / */
        res += len;
    }

  while (*rel)
    {
      eio_ssize_t len, linklen;
      const char *beg = rel;

      while (*rel && *rel != '/')
        ++rel;

      len = rel - beg;

      if (!len) /* skip slashes */
        {
          ++rel;
          continue;
        }

      if (beg [0] == '.')
        {
          if (len == 1)
            continue; /* . - nop */

          if (beg [1] == '.' && len == 2)
            {
              /* .. - back up one component, if possible */

              while (res != tmpbuf->ptr)
                if (*--res == '/')
                  break;

              continue;
            }
        }

        errno = ENAMETOOLONG;
        if (res + 1 + len + 1 >= tmp1)
          return -1;

        /* copy one component */
        *res = '/';
        memcpy (res + 1, beg, len);

        /* zero-terminate, for readlink */
        res [len + 1] = 0;

        /* now check if it's a symlink */
        linklen = readlink (tmpbuf->ptr, tmp1, PATH_MAX);

        if (linklen < 0)
          {
            if (errno != EINVAL)
              return -1;

            /* it's a normal directory. hopefully */
            res += len + 1;
          }
        else
          {
            /* yay, it was a symlink - build new path in tmp2 */
            int rellen = strlen (rel);

            errno = ENAMETOOLONG;
            if (linklen + 1 + rellen >= PATH_MAX)
              return -1;

            errno = ELOOP;
            if (!--symlinks)
              return -1;

            if (*tmp1 == '/')
              res = tmpbuf->ptr; /* symlink resolves to an absolute path */

            /* we need to be careful, as rel might point into tmp2 already */
            memmove (tmp2 + linklen + 1, rel, rellen + 1);
            tmp2 [linklen] = '/';
            memcpy (tmp2, tmp1, linklen);

            rel = tmp2;
          }
    }

  /* special case for the lone root path */
  if (res == tmpbuf->ptr)
    *res++ = '/';

  return res - (char *)tmpbuf->ptr;
}

static signed char
eio_dent_cmp (const eio_dirent *a, const eio_dirent *b)
{
  return a->score - b->score ? a->score - b->score /* works because our signed char is always 0..100 */
       : a->inode < b->inode ? -1
       : a->inode > b->inode ?  1
       :                        0;
}

#define EIO_DENT_CMP(i,op,j) eio_dent_cmp (&i, &j) op 0

#define EIO_SORT_CUTOFF 30 /* quite high, but performs well on many filesystems */
#define EIO_SORT_FAST   60 /* when to only use insertion sort */

static void
eio_dent_radix_sort (eio_dirent *dents, int size, signed char score_bits, eio_ino_t inode_bits)
{
  unsigned char bits [9 + sizeof (eio_ino_t) * 8];
  unsigned char *bit = bits;

  assert (CHAR_BIT == 8);
  assert (sizeof (eio_dirent) * 8 < 256);
  assert (offsetof (eio_dirent, inode)); /* we use bit #0 as sentinel */
  assert (offsetof (eio_dirent, score)); /* we use bit #0 as sentinel */

  if (size <= EIO_SORT_FAST)
    return;

  /* first prepare an array of bits to test in our radix sort */
  /* try to take endianness into account, as well as differences in eio_ino_t sizes */
  /* inode_bits must contain all inodes ORed together */
  /* which is used to skip bits that are 0 everywhere, which is very common */
  {
    eio_ino_t endianness;
    int i, j;

    /* we store the byte offset of byte n into byte n of "endianness" */
    for (i = 0; i < sizeof (eio_ino_t); ++i)
      ((unsigned char *)&endianness)[i] = i;

    *bit++ = 0;

    for (i = 0; i < sizeof (eio_ino_t); ++i)
      {
        /* shifting off the byte offsets out of "endianness" */
        int offs = (offsetof (eio_dirent, inode) + (endianness & 0xff)) * 8;
        endianness >>= 8;

        for (j = 0; j < 8; ++j)
          if (inode_bits & (((eio_ino_t)1) << (i * 8 + j)))
            *bit++ = offs + j;
      }

    for (j = 0; j < 8; ++j)
      if (score_bits & (1 << j))
        *bit++ = offsetof (eio_dirent, score) * 8 + j;
  }

  /* now actually do the sorting (a variant of MSD radix sort) */
  {
    eio_dirent    *base_stk [9 + sizeof (eio_ino_t) * 8], *base;
    eio_dirent    *end_stk  [9 + sizeof (eio_ino_t) * 8], *end;
    unsigned char *bit_stk  [9 + sizeof (eio_ino_t) * 8];
    int stk_idx = 0;

    base_stk [stk_idx] = dents;
    end_stk  [stk_idx] = dents + size;
    bit_stk  [stk_idx] = bit - 1;

    do
      {
        base = base_stk [stk_idx];
        end  = end_stk  [stk_idx];
        bit  = bit_stk  [stk_idx];

        for (;;)
          {
            unsigned char O = *bit >> 3;
            unsigned char M = 1 << (*bit & 7);

            eio_dirent *a = base;
            eio_dirent *b = end;

            if (b - a < EIO_SORT_CUTOFF)
              break;

            /* now bit-partition the array on the bit */
            /* this ugly asymmetric loop seems to perform much better than typical */
            /* partition algos found in the literature */
            do
              if (!(((unsigned char *)a)[O] & M))
                ++a;
              else if (!(((unsigned char *)--b)[O] & M))
                {
                  eio_dirent tmp = *a; *a = *b; *b = tmp;
                  ++a;
                }
            while (b > a);

            /* next bit, or stop, if no bits left in this path */
            if (!*--bit)
              break;

            base_stk [stk_idx] = a;
            end_stk  [stk_idx] = end;
            bit_stk  [stk_idx] = bit;
            ++stk_idx;

            end = a;
          }
      }
    while (stk_idx--);
  }
}

static void
eio_dent_insertion_sort (eio_dirent *dents, int size)
{
  /* first move the smallest element to the front, to act as a sentinel */
  {
    int i;
    eio_dirent *min = dents;
    
    /* the radix pre-pass ensures that the minimum element is in the first EIO_SORT_CUTOFF + 1 elements */
    for (i = size > EIO_SORT_FAST ? EIO_SORT_CUTOFF + 1 : size; --i; )
      if (EIO_DENT_CMP (dents [i], <, *min))
        min = &dents [i];

    /* swap elements 0 and j (minimum) */
    {
      eio_dirent tmp = *dents; *dents = *min; *min = tmp;
    }
  }

  /* then do standard insertion sort, assuming that all elements are >= dents [0] */
  {
    eio_dirent *i, *j;

    for (i = dents + 1; i < dents + size; ++i)
      {
        eio_dirent value = *i;

        for (j = i - 1; EIO_DENT_CMP (*j, >, value); --j)
          j [1] = j [0];

        j [1] = value;
      }
  }
}

static void
eio_dent_sort (eio_dirent *dents, int size, signed char score_bits, eio_ino_t inode_bits)
{
  if (size <= 1)
    return; /* our insertion sort relies on size > 0 */

  /* first we use a radix sort, but only for dirs >= EIO_SORT_FAST */
  /* and stop sorting when the partitions are <= EIO_SORT_CUTOFF */
  eio_dent_radix_sort (dents, size, score_bits, inode_bits);

  /* use an insertion sort at the end, or for small arrays, */
  /* as insertion sort is more efficient for small partitions */
  eio_dent_insertion_sort (dents, size);
}

/* read a full directory */
static void
eio__scandir (eio_req *req, etp_worker *self)
{
  char *name, *names;
  int namesalloc = 4096 - sizeof (void *) * 4;
  int namesoffs = 0;
  int flags = req->int1;
  eio_dirent *dents = 0;
  int dentalloc = 128;
  int dentoffs = 0;
  eio_ino_t inode_bits = 0;
#ifdef _WIN32
  HANDLE dirp;
  WIN32_FIND_DATA entp;
#else
  DIR *dirp;
  EIO_STRUCT_DIRENT *entp;
#endif

  req->result = -1;

  if (!(flags & EIO_READDIR_DENTS))
    flags &= ~(EIO_READDIR_DIRS_FIRST | EIO_READDIR_STAT_ORDER);

#ifdef _WIN32
  {
    int len = strlen ((const char *)req->ptr1);
    char *path = malloc (MAX_PATH);
    const char *fmt;
    const char *reqpath = wd_expand (&self->tmpbuf, req->wd, req->ptr1);

    if (!len)
      fmt = "./*";
    else if (reqpath[len - 1] == '/' || reqpath[len - 1] == '\\')
      fmt = "%s*";
    else
      fmt = "%s/*";

    _snprintf (path, MAX_PATH, fmt, reqpath);
    dirp = FindFirstFile (path, &entp);
    free (path);

    if (dirp == INVALID_HANDLE_VALUE)
     {
       /* should steal _dosmaperr */
       switch (GetLastError ())
         {
           case ERROR_FILE_NOT_FOUND:
             req->result = 0;
             break;

           case ERROR_INVALID_NAME:
           case ERROR_PATH_NOT_FOUND:
           case ERROR_NO_MORE_FILES:
             errno = ENOENT;
             break;

           case ERROR_NOT_ENOUGH_MEMORY:
             errno = ENOMEM;
             break;

           default:
             errno = EINVAL;
             break;
         }

       return;
     }
  }
#else
  #if HAVE_AT
    if (req->wd)
      {
        int fd = openat (WD2FD (req->wd), req->ptr1, O_CLOEXEC | O_SEARCH | O_DIRECTORY);

        if (fd < 0)
          return;

        dirp = fdopendir (fd);

        if (!dirp)
          close (fd);
      }
    else
      dirp = opendir (req->ptr1);
  #else
    dirp = opendir (wd_expand (&self->tmpbuf, req->wd, req->ptr1));
  #endif

  if (!dirp)
    return;
#endif

  if (req->flags & EIO_FLAG_PTR1_FREE)
    free (req->ptr1);

  req->flags |= EIO_FLAG_PTR1_FREE | EIO_FLAG_PTR2_FREE;
  req->ptr1 = dents = flags ? malloc (dentalloc * sizeof (eio_dirent)) : 0;
  req->ptr2 = names = malloc (namesalloc);

  if (!names || (flags && !dents))
    return;

  for (;;)
    {
      int done;

#ifdef _WIN32
      done = !dirp;
#else
      errno = 0;
      entp = readdir (dirp);
      done = !entp;
#endif

      if (done)
        {
#ifndef _WIN32
          int old_errno = errno;
          closedir (dirp);
          errno = old_errno;

          if (errno)
            break;
#endif

          /* sort etc. */
          req->int1   = flags;
          req->result = dentoffs;

          if (flags & EIO_READDIR_STAT_ORDER)
            eio_dent_sort (dents, dentoffs, flags & EIO_READDIR_DIRS_FIRST ? 7 : 0, inode_bits);
          else if (flags & EIO_READDIR_DIRS_FIRST)
            if (flags & EIO_READDIR_FOUND_UNKNOWN)
              eio_dent_sort (dents, dentoffs, 7, inode_bits); /* sort by score and inode */
            else
              {
                /* in this case, all is known, and we just put dirs first and sort them */
                eio_dirent *oth = dents + dentoffs;
                eio_dirent *dir = dents;

                /* now partition dirs to the front, and non-dirs to the back */
                /* by walking from both sides and swapping if necessary */
                while (oth > dir)
                  {
                    if (dir->type == EIO_DT_DIR)
                      ++dir;
                    else if ((--oth)->type == EIO_DT_DIR)
                      {
                        eio_dirent tmp = *dir; *dir = *oth; *oth = tmp;

                        ++dir;
                      }
                  }

                /* now sort the dirs only (dirs all have the same score) */
                eio_dent_sort (dents, dir - dents, 0, inode_bits);
              }

          break;
        }

      /* now add the entry to our list(s) */
      name = D_NAME (entp);

      /* skip . and .. entries */
      if (name [0] != '.' || (name [1] && (name [1] != '.' || name [2])))
        {
          int len = D_NAMLEN (entp) + 1;

          while (ecb_expect_false (namesoffs + len > namesalloc))
            {
              namesalloc *= 2;
              req->ptr2 = names = realloc (names, namesalloc);

              if (!names)
                break;
            }

          memcpy (names + namesoffs, name, len);

          if (dents)
            {
              struct eio_dirent *ent;

              if (ecb_expect_false (dentoffs == dentalloc))
                {
                  dentalloc *= 2;
                  req->ptr1 = dents = realloc (dents, dentalloc * sizeof (eio_dirent));

                  if (!dents)
                    break;
                }

              ent = dents + dentoffs;

              ent->nameofs = namesoffs; /* rather dirtily we store the offset in the pointer */
              ent->namelen = len - 1;
              ent->inode   = D_INO (entp);

              inode_bits |= ent->inode;

              switch (D_TYPE (entp))
                {
                  default:
                    ent->type = EIO_DT_UNKNOWN;
                    flags |= EIO_READDIR_FOUND_UNKNOWN;
                    break;

                  #ifdef DT_FIFO
                    case DT_FIFO: ent->type = EIO_DT_FIFO; break;
                  #endif
                  #ifdef DT_CHR
                    case DT_CHR:  ent->type = EIO_DT_CHR;  break;
                  #endif          
                  #ifdef DT_MPC
                    case DT_MPC:  ent->type = EIO_DT_MPC;  break;
                  #endif          
                  #ifdef DT_DIR
                    case DT_DIR:  ent->type = EIO_DT_DIR;  break;
                  #endif          
                  #ifdef DT_NAM
                    case DT_NAM:  ent->type = EIO_DT_NAM;  break;
                  #endif          
                  #ifdef DT_BLK
                    case DT_BLK:  ent->type = EIO_DT_BLK;  break;
                  #endif          
                  #ifdef DT_MPB
                    case DT_MPB:  ent->type = EIO_DT_MPB;  break;
                  #endif          
                  #ifdef DT_REG
                    case DT_REG:  ent->type = EIO_DT_REG;  break;
                  #endif          
                  #ifdef DT_NWK
                    case DT_NWK:  ent->type = EIO_DT_NWK;  break;
                  #endif          
                  #ifdef DT_CMP
                    case DT_CMP:  ent->type = EIO_DT_CMP;  break;
                  #endif          
                  #ifdef DT_LNK
                    case DT_LNK:  ent->type = EIO_DT_LNK;  break;
                  #endif
                  #ifdef DT_SOCK
                    case DT_SOCK: ent->type = EIO_DT_SOCK; break;
                  #endif
                  #ifdef DT_DOOR
                    case DT_DOOR: ent->type = EIO_DT_DOOR; break;
                  #endif
                  #ifdef DT_WHT
                    case DT_WHT:  ent->type = EIO_DT_WHT;  break;
                  #endif
                }

              ent->score = 7;

              if (flags & EIO_READDIR_DIRS_FIRST)
                {
                  if (ent->type == EIO_DT_UNKNOWN)
                    {
                      if (*name == '.') /* leading dots are likely directories, and, in any case, rare */
                        ent->score = 1;
                      else if (!strchr (name, '.')) /* absence of dots indicate likely dirs */
                        ent->score = len <= 2 ? 4 - len : len <= 4 ? 4 : len <= 7 ? 5 : 6; /* shorter == more likely dir, but avoid too many classes */
                    }
                  else if (ent->type == EIO_DT_DIR)
                    ent->score = 0;
                }
            }

          namesoffs += len;
          ++dentoffs;
        }

      if (EIO_CANCELLED (req))
        {
          errno = ECANCELED;
          break;
        }

#ifdef _WIN32
      if (!FindNextFile (dirp, &entp))
        {
          FindClose (dirp);
          dirp = 0;
        }
#endif
    }
}

/*****************************************************************************/
/* working directory stuff */
/* various deficiencies in the posix 2008 api force us to */
/* keep the absolute path in string form at all times */
/* fuck yeah. */

#if !HAVE_AT

/* a bit like realpath, but usually faster because it doesn'T have to return */
/* an absolute or canonical path */
static const char *
wd_expand (struct tmpbuf *tmpbuf, eio_wd wd, const char *path)
{
  if (!wd || *path == '/')
    return path;

  if (path [0] == '.' && !path [1])
    return wd->str;

  {
    int l1 = wd->len;
    int l2 = strlen (path);

    char *res = tmpbuf_get (tmpbuf, l1 + l2 + 2);

    memcpy (res, wd->str, l1);
    res [l1] = '/';
    memcpy (res + l1 + 1, path, l2 + 1);

    return res;
  }
}

#endif

static eio_wd
eio__wd_open_sync (struct tmpbuf *tmpbuf, eio_wd wd, const char *path)
{
  int fd;
  eio_wd res;
  int len = eio__realpath (tmpbuf, wd, path);

  if (len < 0)
    return EIO_INVALID_WD;

#if HAVE_AT
  fd = openat (WD2FD (wd), path, O_CLOEXEC | O_SEARCH | O_DIRECTORY);

  if (fd < 0)
    return EIO_INVALID_WD;
#endif

  res = malloc (sizeof (*res) + len); /* one extra 0-byte */

#if HAVE_AT
  res->fd = fd;
#endif

  res->len = len;
  memcpy (res->str, tmpbuf->ptr, len);
  res->str [len] = 0;

  return res;
}

eio_wd
eio_wd_open_sync (eio_wd wd, const char *path)
{
  struct tmpbuf tmpbuf = { 0 };
  wd = eio__wd_open_sync (&tmpbuf, wd, path);
  free (tmpbuf.ptr);

  return wd;
}

void
eio_wd_close_sync (eio_wd wd)
{
  if (wd != EIO_INVALID_WD && wd != EIO_CWD)
    {
      #if HAVE_AT
      close (wd->fd);
      #endif
      free (wd);
    }
}

#if HAVE_AT

/* they forgot these */

static int
eio__truncateat (int dirfd, const char *path, off_t length)
{
  int fd = openat (dirfd, path, O_WRONLY | O_CLOEXEC);
  int res;

  if (fd < 0)
    return fd;

  res = ftruncate (fd, length);
  close (fd);
  return res;
}

static int
eio__statvfsat (int dirfd, const char *path, struct statvfs *buf)
{
  int fd = openat (dirfd, path, O_SEARCH | O_CLOEXEC);
  int res;

  if (fd < 0)
    return fd;

  res = fstatvfs (fd, buf);
  close (fd);
  return res;

}

#endif

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

#define ALLOC(len)				\
  if (!req->ptr2)				\
    {						\
      X_LOCK (wrklock);				\
      req->flags |= EIO_FLAG_PTR2_FREE;		\
      X_UNLOCK (wrklock);			\
      req->ptr2 = malloc (len);			\
      if (!req->ptr2)				\
        {					\
          errno       = ENOMEM;			\
          req->result = -1;			\
          break;				\
        }					\
    }

static void ecb_noinline ecb_cold
etp_proc_init (void)
{
#if HAVE_PRCTL_SET_NAME
  /* provide a more sensible "thread name" */
  char name[16 + 1];
  const int namelen = sizeof (name) - 1;
  int len;

  prctl (PR_GET_NAME, (unsigned long)name, 0, 0, 0);
  name [namelen] = 0;
  len = strlen (name);
  strcpy (name + (len <= namelen - 4 ? len : namelen - 4), "/eio");
  prctl (PR_SET_NAME, (unsigned long)name, 0, 0, 0);
#endif
}

X_THREAD_PROC (etp_proc)
{
  ETP_REQ *req;
  struct timespec ts;
  etp_worker *self = (etp_worker *)thr_arg;

  etp_proc_init ();

  /* try to distribute timeouts somewhat evenly */
  ts.tv_nsec = ((unsigned long)self & 1023UL) * (1000000000UL / 1024UL);

  for (;;)
    {
      ts.tv_sec = 0;

      X_LOCK (reqlock);

      for (;;)
        {
          req = reqq_shift (&req_queue);

          if (req)
            break;

          if (ts.tv_sec == 1) /* no request, but timeout detected, let's quit */
            {
              X_UNLOCK (reqlock);
              X_LOCK (wrklock);
              --started;
              X_UNLOCK (wrklock);
              goto quit;
            }

          ++idle;

          if (idle <= max_idle)
            /* we are allowed to idle, so do so without any timeout */
            X_COND_WAIT (reqwait, reqlock);
          else
            {
              /* initialise timeout once */
              if (!ts.tv_sec)
                ts.tv_sec = time (0) + idle_timeout;

              if (X_COND_TIMEDWAIT (reqwait, reqlock, ts) == ETIMEDOUT)
                ts.tv_sec = 1; /* assuming this is not a value computed above.,.. */
            }

          --idle;
        }

      --nready;

      X_UNLOCK (reqlock);
     
      if (req->type < 0)
        goto quit;

      ETP_EXECUTE (self, req);

      X_LOCK (reslock);

      ++npending;

      if (!reqq_push (&res_queue, req) && want_poll_cb)
        want_poll_cb ();

      etp_worker_clear (self);

      X_UNLOCK (reslock);
    }

quit:
  free (req);

  X_LOCK (wrklock);
  etp_worker_free (self);
  X_UNLOCK (wrklock);
}

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

int ecb_cold
eio_init (void (*want_poll)(void), void (*done_poll)(void))
{
  return etp_init (want_poll, done_poll);
}

ecb_inline void
eio_api_destroy (eio_req *req)
{
  free (req);
}

#define REQ(rtype)						\
  eio_req *req;                                                 \
                                                                \
  req = (eio_req *)calloc (1, sizeof *req);                     \
  if (!req)                                                     \
    return 0;                                                   \
                                                                \
  req->type    = rtype;                                         \
  req->pri     = pri;						\
  req->finish  = cb;						\
  req->data    = data;						\
  req->destroy = eio_api_destroy;

#define SEND eio_submit (req); return req

#define PATH							\
  req->flags |= EIO_FLAG_PTR1_FREE;				\
  req->ptr1 = strdup (path);					\
  if (!req->ptr1)						\
    {								\
      eio_api_destroy (req);					\
      return 0;							\
    }

static void
eio_execute (etp_worker *self, eio_req *req)
{
#if HAVE_AT
  int dirfd;
#else
  const char *path;
#endif

  if (ecb_expect_false (EIO_CANCELLED (req)))
    {
      req->result  = -1;
      req->errorno = ECANCELED;
      return;
    }

  if (ecb_expect_false (req->wd == EIO_INVALID_WD))
    {
      req->result  = -1;
      req->errorno = ENOENT;
      return;
    }

  if (req->type >= EIO_OPEN)
    {
      #if HAVE_AT
        dirfd = WD2FD (req->wd);
      #else
        path = wd_expand (&self->tmpbuf, req->wd, req->ptr1);
      #endif
    }

  switch (req->type)
    {
      case EIO_WD_OPEN:   req->wd = eio__wd_open_sync (&self->tmpbuf, req->wd, req->ptr1);
                          req->result = req->wd == EIO_INVALID_WD ? -1 : 0;
                          break;
      case EIO_WD_CLOSE:  req->result = 0;
                          eio_wd_close_sync (req->wd); break;

      case EIO_SEEK:      eio__lseek (req); break;
      case EIO_READ:      ALLOC (req->size);
                          req->result = req->offs >= 0
                                      ? pread     (req->int1, req->ptr2, req->size, req->offs)
                                      : read      (req->int1, req->ptr2, req->size); break;
      case EIO_WRITE:     req->result = req->offs >= 0
                                      ? pwrite    (req->int1, req->ptr2, req->size, req->offs)
                                      : write     (req->int1, req->ptr2, req->size); break;

      case EIO_READAHEAD: req->result = readahead     (req->int1, req->offs, req->size); break;
      case EIO_SENDFILE:  req->result = eio__sendfile (req->int1, req->int2, req->offs, req->size); break;

#if HAVE_AT

      case EIO_STAT:      ALLOC (sizeof (EIO_STRUCT_STAT));
                          req->result = fstatat   (dirfd, req->ptr1, (EIO_STRUCT_STAT *)req->ptr2, 0); break;
      case EIO_LSTAT:     ALLOC (sizeof (EIO_STRUCT_STAT));
                          req->result = fstatat   (dirfd, req->ptr1, (EIO_STRUCT_STAT *)req->ptr2, AT_SYMLINK_NOFOLLOW); break;
      case EIO_CHOWN:     req->result = fchownat  (dirfd, req->ptr1, req->int2, req->int3, 0); break;
      case EIO_CHMOD:     req->result = fchmodat  (dirfd, req->ptr1, (mode_t)req->int2, 0); break;
      case EIO_TRUNCATE:  req->result = eio__truncateat (dirfd, req->ptr1, req->offs); break;
      case EIO_OPEN:      req->result = openat    (dirfd, req->ptr1, req->int1, (mode_t)req->int2); break;

      case EIO_UNLINK:    req->result = unlinkat  (dirfd, req->ptr1, 0); break;
      case EIO_RMDIR:     req->result = unlinkat  (dirfd, req->ptr1, AT_REMOVEDIR); break;
      case EIO_MKDIR:     req->result = mkdirat   (dirfd, req->ptr1, (mode_t)req->int2); break;
      case EIO_RENAME:    req->result = renameat  (dirfd, req->ptr1, WD2FD ((eio_wd)req->int3), req->ptr2); break;
      case EIO_LINK:      req->result = linkat    (dirfd, req->ptr1, WD2FD ((eio_wd)req->int3), req->ptr2, 0); break;
      case EIO_SYMLINK:   req->result = symlinkat (req->ptr1, dirfd, req->ptr2); break;
      case EIO_MKNOD:     req->result = mknodat   (dirfd, req->ptr1, (mode_t)req->int2, (dev_t)req->offs); break;
      case EIO_READLINK:  ALLOC (PATH_MAX);
                          req->result = readlinkat (dirfd, req->ptr1, req->ptr2, PATH_MAX); break;
      case EIO_STATVFS:   ALLOC (sizeof (EIO_STRUCT_STATVFS));
                          req->result = eio__statvfsat (dirfd, req->ptr1, (EIO_STRUCT_STATVFS *)req->ptr2); break;
      case EIO_UTIME:
      case EIO_FUTIME:
        {
          struct timespec ts[2];
          struct timespec *times;

          if (req->nv1 != -1. || req->nv2 != -1.)
            {
              ts[0].tv_sec  = req->nv1;
              ts[0].tv_nsec = (req->nv1 - ts[0].tv_sec) * 1e9;
              ts[1].tv_sec  = req->nv2;
              ts[1].tv_nsec = (req->nv2 - ts[1].tv_sec) * 1e9;

              times = ts;
            }
          else
            times = 0;

          req->result = req->type == EIO_FUTIME
                        ? futimens  (req->int1, times)
                        : utimensat (dirfd, req->ptr1, times, 0);
        }
        break;

#else

      case EIO_STAT:      ALLOC (sizeof (EIO_STRUCT_STAT));
                          req->result = stat      (path     , (EIO_STRUCT_STAT *)req->ptr2); break;
      case EIO_LSTAT:     ALLOC (sizeof (EIO_STRUCT_STAT));
                          req->result = lstat     (path     , (EIO_STRUCT_STAT *)req->ptr2); break;
      case EIO_CHOWN:     req->result = chown     (path     , req->int2, req->int3); break;
      case EIO_CHMOD:     req->result = chmod     (path     , (mode_t)req->int2); break;
      case EIO_TRUNCATE:  req->result = truncate  (path     , req->offs); break;
      case EIO_OPEN:      req->result = open      (path     , req->int1, (mode_t)req->int2); break;

      case EIO_UNLINK:    req->result = unlink    (path     ); break;
      case EIO_RMDIR:     req->result = rmdir     (path     ); break;
      case EIO_MKDIR:     req->result = mkdir     (path     , (mode_t)req->int2); break;
      case EIO_RENAME:    req->result = rename    (path     , req->ptr2); break;
      case EIO_LINK:      req->result = link      (path     , req->ptr2); break;
      case EIO_SYMLINK:   req->result = symlink   (path     , req->ptr2); break;
      case EIO_MKNOD:     req->result = mknod     (path     , (mode_t)req->int2, (dev_t)req->offs); break;
      case EIO_READLINK:  ALLOC (PATH_MAX);
                          req->result = readlink  (path, req->ptr2, PATH_MAX); break;
      case EIO_STATVFS:   ALLOC (sizeof (EIO_STRUCT_STATVFS));
                          req->result = statvfs   (path     , (EIO_STRUCT_STATVFS *)req->ptr2); break;

      case EIO_UTIME:
      case EIO_FUTIME:
        {
          struct timeval tv[2];
          struct timeval *times;

          if (req->nv1 != -1. || req->nv2 != -1.)
            {
              tv[0].tv_sec  = req->nv1;
              tv[0].tv_usec = (req->nv1 - tv[0].tv_sec) * 1e6;
              tv[1].tv_sec  = req->nv2;
              tv[1].tv_usec = (req->nv2 - tv[1].tv_sec) * 1e6;

              times = tv;
            }
          else
            times = 0;

          req->result = req->type == EIO_FUTIME
                        ? futimes (req->int1, times)
                        : utimes  (req->ptr1, times);
        }
        break;

#endif

      case EIO_REALPATH:  if (0 <= (req->result = eio__realpath (&self->tmpbuf, req->wd, req->ptr1)))
                            {
                              ALLOC (req->result);
                              memcpy (req->ptr2, self->tmpbuf.ptr, req->result);
                            }
                           break;

      case EIO_FSTAT:     ALLOC (sizeof (EIO_STRUCT_STAT));
                          req->result = fstat     (req->int1, (EIO_STRUCT_STAT *)req->ptr2); break;

      case EIO_FSTATVFS:  ALLOC (sizeof (EIO_STRUCT_STATVFS));
                          req->result = fstatvfs  (req->int1, (EIO_STRUCT_STATVFS *)req->ptr2); break;

      case EIO_FCHOWN:    req->result = fchown    (req->int1, req->int2, req->int3); break;
      case EIO_FCHMOD:    req->result = fchmod    (req->int1, (mode_t)req->int2); break;
      case EIO_FTRUNCATE: req->result = ftruncate (req->int1, req->offs); break;

      case EIO_CLOSE:     req->result = close     (req->int1); break;
      case EIO_DUP2:      req->result = dup2      (req->int1, req->int2); break;
      case EIO_SYNC:      req->result = 0; sync (); break;
      case EIO_FSYNC:     req->result = fsync     (req->int1); break;
      case EIO_FDATASYNC: req->result = fdatasync (req->int1); break;
      case EIO_SYNCFS:    req->result = eio__syncfs (req->int1); break;
      case EIO_SYNC_FILE_RANGE: req->result = eio__sync_file_range (req->int1, req->offs, req->size, req->int2); break;
      case EIO_MSYNC:     req->result = eio__msync (req->ptr2, req->size, req->int1); break;
      case EIO_MTOUCH:    req->result = eio__mtouch (req); break;
      case EIO_MLOCK:     req->result = eio__mlock (req->ptr2, req->size); break;
      case EIO_MLOCKALL:  req->result = eio__mlockall (req->int1); break;
      case EIO_FALLOCATE: req->result = eio__fallocate (req->int1, req->int2, req->offs, req->size); break;

      case EIO_READDIR:   eio__scandir (req, self); break;

      case EIO_BUSY:
#ifdef _WIN32
	Sleep (req->nv1 * 1e3);
#else
        {
          struct timeval tv;

          tv.tv_sec  = req->nv1;
          tv.tv_usec = (req->nv1 - tv.tv_sec) * 1e6;

          req->result = select (0, 0, 0, 0, &tv);
        }
#endif
        break;

      case EIO_GROUP:
        abort (); /* handled in eio_request */

      case EIO_NOP:
        req->result = 0;
        break;

      case EIO_CUSTOM:
        req->feed (req);
        break;

      default:
        req->result = EIO_ENOSYS ();
        break;
    }

  req->errorno = errno;
}

#ifndef EIO_NO_WRAPPERS

eio_req *eio_wd_open (const char *path, int pri, eio_cb cb, void *data)
{
  REQ (EIO_WD_OPEN); PATH; SEND;
}

eio_req *eio_wd_close (eio_wd wd, int pri, eio_cb cb, void *data)
{
  REQ (EIO_WD_CLOSE); req->wd = wd; SEND;
}

eio_req *eio_nop (int pri, eio_cb cb, void *data)
{
  REQ (EIO_NOP); SEND;
}

eio_req *eio_busy (double delay, int pri, eio_cb cb, void *data)
{
  REQ (EIO_BUSY); req->nv1 = delay; SEND;
}

eio_req *eio_sync (int pri, eio_cb cb, void *data)
{
  REQ (EIO_SYNC); SEND;
}

eio_req *eio_fsync (int fd, int pri, eio_cb cb, void *data)
{
  REQ (EIO_FSYNC); req->int1 = fd; SEND;
}

eio_req *eio_msync (void *addr, size_t length, int flags, int pri, eio_cb cb, void *data)
{
  REQ (EIO_MSYNC); req->ptr2 = addr; req->size = length; req->int1 = flags; SEND;
}

eio_req *eio_fdatasync (int fd, int pri, eio_cb cb, void *data)
{
  REQ (EIO_FDATASYNC); req->int1 = fd; SEND;
}

eio_req *eio_syncfs (int fd, int pri, eio_cb cb, void *data)
{
  REQ (EIO_SYNCFS); req->int1 = fd; SEND;
}

eio_req *eio_sync_file_range (int fd, off_t offset, size_t nbytes, unsigned int flags, int pri, eio_cb cb, void *data)
{
  REQ (EIO_SYNC_FILE_RANGE); req->int1 = fd; req->offs = offset; req->size = nbytes; req->int2 = flags; SEND;
}

eio_req *eio_mtouch (void *addr, size_t length, int flags, int pri, eio_cb cb, void *data)
{
  REQ (EIO_MTOUCH); req->ptr2 = addr; req->size = length; req->int1 = flags; SEND;
}

eio_req *eio_mlock (void *addr, size_t length, int pri, eio_cb cb, void *data)
{
  REQ (EIO_MLOCK); req->ptr2 = addr; req->size = length; SEND;
}

eio_req *eio_mlockall (int flags, int pri, eio_cb cb, void *data)
{
  REQ (EIO_MLOCKALL); req->int1 = flags; SEND;
}

eio_req *eio_fallocate (int fd, int mode, off_t offset, size_t len, int pri, eio_cb cb, void *data)
{
  REQ (EIO_FALLOCATE); req->int1 = fd; req->int2 = mode; req->offs = offset; req->size = len; SEND;
}

eio_req *eio_close (int fd, int pri, eio_cb cb, void *data)
{
  REQ (EIO_CLOSE); req->int1 = fd; SEND;
}

eio_req *eio_readahead (int fd, off_t offset, size_t length, int pri, eio_cb cb, void *data)
{
  REQ (EIO_READAHEAD); req->int1 = fd; req->offs = offset; req->size = length; SEND;
}

eio_req *eio_seek (int fd, off_t offset, int whence, int pri, eio_cb cb, void *data)
{
  REQ (EIO_SEEK); req->int1 = fd; req->offs = offset; req->int2 = whence; SEND;
}

eio_req *eio_read (int fd, void *buf, size_t length, off_t offset, int pri, eio_cb cb, void *data)
{
  REQ (EIO_READ); req->int1 = fd; req->offs = offset; req->size = length; req->ptr2 = buf; SEND;
}

eio_req *eio_write (int fd, void *buf, size_t length, off_t offset, int pri, eio_cb cb, void *data)
{
  REQ (EIO_WRITE); req->int1 = fd; req->offs = offset; req->size = length; req->ptr2 = buf; SEND;
}

eio_req *eio_fstat (int fd, int pri, eio_cb cb, void *data)
{
  REQ (EIO_FSTAT); req->int1 = fd; SEND;
}

eio_req *eio_fstatvfs (int fd, int pri, eio_cb cb, void *data)
{
  REQ (EIO_FSTATVFS); req->int1 = fd; SEND;
}

eio_req *eio_futime (int fd, double atime, double mtime, int pri, eio_cb cb, void *data)
{
  REQ (EIO_FUTIME); req->int1 = fd; req->nv1 = atime; req->nv2 = mtime; SEND;
}

eio_req *eio_ftruncate (int fd, off_t offset, int pri, eio_cb cb, void *data)
{
  REQ (EIO_FTRUNCATE); req->int1 = fd; req->offs = offset; SEND;
}

eio_req *eio_fchmod (int fd, mode_t mode, int pri, eio_cb cb, void *data)
{
  REQ (EIO_FCHMOD); req->int1 = fd; req->int2 = (long)mode; SEND;
}

eio_req *eio_fchown (int fd, eio_uid_t uid, eio_gid_t gid, int pri, eio_cb cb, void *data)
{
  REQ (EIO_FCHOWN); req->int1 = fd; req->int2 = (long)uid; req->int3 = (long)gid; SEND;
}

eio_req *eio_dup2 (int fd, int fd2, int pri, eio_cb cb, void *data)
{
  REQ (EIO_DUP2); req->int1 = fd; req->int2 = fd2; SEND;
}

eio_req *eio_sendfile (int out_fd, int in_fd, off_t in_offset, size_t length, int pri, eio_cb cb, void *data)
{
  REQ (EIO_SENDFILE); req->int1 = out_fd; req->int2 = in_fd; req->offs = in_offset; req->size = length; SEND;
}

eio_req *eio_open (const char *path, int flags, mode_t mode, int pri, eio_cb cb, void *data)
{
  REQ (EIO_OPEN); PATH; req->int1 = flags; req->int2 = (long)mode; SEND;
}

eio_req *eio_utime (const char *path, double atime, double mtime, int pri, eio_cb cb, void *data)
{
  REQ (EIO_UTIME); PATH; req->nv1 = atime; req->nv2 = mtime; SEND;
}

eio_req *eio_truncate (const char *path, off_t offset, int pri, eio_cb cb, void *data)
{
  REQ (EIO_TRUNCATE); PATH; req->offs = offset; SEND;
}

eio_req *eio_chown (const char *path, eio_uid_t uid, eio_gid_t gid, int pri, eio_cb cb, void *data)
{
  REQ (EIO_CHOWN); PATH; req->int2 = (long)uid; req->int3 = (long)gid; SEND;
}

eio_req *eio_chmod (const char *path, mode_t mode, int pri, eio_cb cb, void *data)
{
  REQ (EIO_CHMOD); PATH; req->int2 = (long)mode; SEND;
}

eio_req *eio_mkdir (const char *path, mode_t mode, int pri, eio_cb cb, void *data)
{
  REQ (EIO_MKDIR); PATH; req->int2 = (long)mode; SEND;
}

static eio_req *
eio__1path (int type, const char *path, int pri, eio_cb cb, void *data)
{
  REQ (type); PATH; SEND;
}

eio_req *eio_readlink (const char *path, int pri, eio_cb cb, void *data)
{
  return eio__1path (EIO_READLINK, path, pri, cb, data);
}

eio_req *eio_realpath (const char *path, int pri, eio_cb cb, void *data)
{
  return eio__1path (EIO_REALPATH, path, pri, cb, data);
}

eio_req *eio_stat (const char *path, int pri, eio_cb cb, void *data)
{
  return eio__1path (EIO_STAT, path, pri, cb, data);
}

eio_req *eio_lstat (const char *path, int pri, eio_cb cb, void *data)
{
  return eio__1path (EIO_LSTAT, path, pri, cb, data);
}

eio_req *eio_statvfs (const char *path, int pri, eio_cb cb, void *data)
{
  return eio__1path (EIO_STATVFS, path, pri, cb, data);
}

eio_req *eio_unlink (const char *path, int pri, eio_cb cb, void *data)
{
  return eio__1path (EIO_UNLINK, path, pri, cb, data);
}

eio_req *eio_rmdir (const char *path, int pri, eio_cb cb, void *data)
{
  return eio__1path (EIO_RMDIR, path, pri, cb, data);
}

eio_req *eio_readdir (const char *path, int flags, int pri, eio_cb cb, void *data)
{
  REQ (EIO_READDIR); PATH; req->int1 = flags; SEND;
}

eio_req *eio_mknod (const char *path, mode_t mode, dev_t dev, int pri, eio_cb cb, void *data)
{
  REQ (EIO_MKNOD); PATH; req->int2 = (long)mode; req->offs = (off_t)dev; SEND;
}

static eio_req *
eio__2path (int type, const char *path, const char *new_path, int pri, eio_cb cb, void *data)
{
  REQ (type); PATH;

  req->flags |= EIO_FLAG_PTR2_FREE;
  req->ptr2 = strdup (new_path);
  if (!req->ptr2)
    {
      eio_api_destroy (req);
      return 0;
    }

  SEND;
}

eio_req *eio_link (const char *path, const char *new_path, int pri, eio_cb cb, void *data)
{
  return eio__2path (EIO_LINK, path, new_path, pri, cb, data);
}

eio_req *eio_symlink (const char *path, const char *new_path, int pri, eio_cb cb, void *data)
{
  return eio__2path (EIO_SYMLINK, path, new_path, pri, cb, data);
}

eio_req *eio_rename (const char *path, const char *new_path, int pri, eio_cb cb, void *data)
{
  return eio__2path (EIO_RENAME, path, new_path, pri, cb, data);
}

eio_req *eio_custom (void (*execute)(eio_req *), int pri, eio_cb cb, void *data)
{
  REQ (EIO_CUSTOM); req->feed = execute; SEND;
}

#endif

eio_req *eio_grp (eio_cb cb, void *data)
{
  const int pri = EIO_PRI_MAX;

  REQ (EIO_GROUP); SEND;
}

#undef REQ
#undef PATH
#undef SEND

/*****************************************************************************/
/* grp functions */

void
eio_grp_feed (eio_req *grp, void (*feed)(eio_req *req), int limit)
{
  grp->int2 = limit;
  grp->feed = feed;

  grp_try_feed (grp);
}

void
eio_grp_limit (eio_req *grp, int limit)
{
  grp->int2 = limit;

  grp_try_feed (grp);
}

void
eio_grp_add (eio_req *grp, eio_req *req)
{
  assert (("cannot add requests to IO::AIO::GRP after the group finished", grp->int1 != 2));

  grp->flags |= EIO_FLAG_GROUPADD;

  ++grp->size;
  req->grp = grp;

  req->grp_prev = 0;
  req->grp_next = grp->grp_first;

  if (grp->grp_first)
    grp->grp_first->grp_prev = req;

  grp->grp_first = req;
}

/*****************************************************************************/
/* misc garbage */

eio_ssize_t
eio_sendfile_sync (int ofd, int ifd, off_t offset, size_t count)
{
  return eio__sendfile (ofd, ifd, offset, count);
}