aboutsummaryrefslogblamecommitdiffstats
path: root/erts/emulator/sys/unix/sys.c
blob: 151fa06e8e7c1c199b04e2e799f1487fdcd053d9 (plain) (tree)
1
2
3
4
5
6
7
8
9
10
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
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
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
                   

                                                        



                                                                      
  


                                                                         
  































































































































































































































































































































































                                                                               
                                 























































































                                                                            


                             


























































































































































































































































                                                                             
                                                        





































































































































































































































































































































































































































































































































































































































                                                                              
                                                                        












































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































                                                                                                                
                                                                












































































































































































































































































































































































































































































































































































                                                                               
                                                                    



































































































































































































































                                                                                




























































































































































































































                                                                                       
/*
 * %CopyrightBegin%
 *
 * Copyright Ericsson AB 1996-2010. All Rights Reserved.
 *
 * The contents of this file are subject to the Erlang Public License,
 * Version 1.1, (the "License"); you may not use this file except in
 * compliance with the License. You should have received a copy of the
 * Erlang Public License along with this software. If not, it can be
 * retrieved online at http://www.erlang.org/.
 *
 * Software distributed under the License is distributed on an "AS IS"
 * basis, WITHOUT WARRANTY OF ANY KIND, either express or implied. See
 * the License for the specific language governing rights and limitations
 * under the License.
 *
 * %CopyrightEnd%
 */

#ifdef HAVE_CONFIG_H
#  include "config.h"
#endif

#ifdef ISC32
#define _POSIX_SOURCE
#define _XOPEN_SOURCE
#endif

#include <sys/times.h>		/* ! */
#include <time.h>
#include <signal.h>
#include <sys/wait.h>
#include <sys/uio.h>
#include <termios.h>
#include <ctype.h>
#include <sys/utsname.h>

#ifdef ISC32
#include <sys/bsdtypes.h>
#endif

#include <termios.h>
#ifdef HAVE_FCNTL_H
#include <fcntl.h>
#endif
#ifdef HAVE_SYS_IOCTL_H
#include <sys/ioctl.h>
#endif

#define NEED_CHILD_SETUP_DEFINES
#define ERTS_WANT_BREAK_HANDLING
#define ERTS_WANT_GOT_SIGUSR1
#define WANT_NONBLOCKING    /* must define this to pull in defs from sys.h */
#include "sys.h"

#ifdef USE_THREADS
#include "erl_threads.h"
#endif

#include "erl_mseg.h"

extern char **environ;
static erts_smp_rwmtx_t environ_rwmtx;

#define MAX_VSIZE 16		/* Max number of entries allowed in an I/O
				 * vector sock_sendv().
				 */

/*
 * Don't need global.h, but bif_table.h (included by bif.h),
 * won't compile otherwise
 */
#include "global.h" 
#include "bif.h"

#include "erl_sys_driver.h"
#include "erl_check_io.h"

#ifndef DISABLE_VFORK
#define DISABLE_VFORK 0
#endif

#ifdef USE_THREADS
#  ifdef ENABLE_CHILD_WAITER_THREAD
#    define CHLDWTHR ENABLE_CHILD_WAITER_THREAD
#  else
#    define CHLDWTHR 0
#  endif
#else
#  define CHLDWTHR 0
#endif
/*
 * [OTP-3906]
 * Solaris signal management gets confused when threads are used and a
 * lot of child processes dies. The confusion results in that SIGCHLD
 * signals aren't delivered to the emulator which in turn results in
 * a lot of defunct processes in the system.
 *
 * The problem seems to appear when a signal is frequently
 * blocked/unblocked at the same time as the signal is frequently
 * propagated. The child waiter thread is a workaround for this problem.
 * The SIGCHLD signal is always blocked (in all threads), and the child
 * waiter thread fetches the signal by a call to sigwait(). See
 * child_waiter().
 */

typedef struct ErtsSysReportExit_ ErtsSysReportExit;
struct ErtsSysReportExit_ {
    ErtsSysReportExit *next;
    Eterm port;
    int pid;
    int ifd;
    int ofd;
#if CHLDWTHR && !defined(ERTS_SMP)
    int status;
#endif
};

static ErtsSysReportExit *report_exit_list;
#if CHLDWTHR && !defined(ERTS_SMP)
static ErtsSysReportExit *report_exit_transit_list;
#endif

extern int  check_async_ready(void);
extern int  driver_interrupt(int, int);
/*EXTERN_FUNCTION(void, increment_time, (int));*/
/*EXTERN_FUNCTION(int, next_time, (_VOID_));*/
extern void do_break(void);

extern void erl_sys_args(int*, char**);

/* The following two defs should probably be moved somewhere else */

extern void erts_sys_init_float(void);

extern void erl_crash_dump(char* file, int line, char* fmt, ...);

#define DIR_SEPARATOR_CHAR    '/'

#if defined(DEBUG)
#define ERL_BUILD_TYPE_MARKER ".debug"
#elif defined(PURIFY)
#define ERL_BUILD_TYPE_MARKER ".purify"
#elif defined(QUANTIFY)
#define ERL_BUILD_TYPE_MARKER ".quantify"
#elif defined(PURECOV)
#define ERL_BUILD_TYPE_MARKER ".purecov"
#elif defined(VALGRIND)
#define ERL_BUILD_TYPE_MARKER ".valgrind"
#else /* opt */
#define ERL_BUILD_TYPE_MARKER
#endif

#define CHILD_SETUP_PROG_NAME	"child_setup" ERL_BUILD_TYPE_MARKER
#if !DISABLE_VFORK
static char *child_setup_prog;
#endif

#ifdef DEBUG
static int debug_log = 0;
#endif

#ifdef ERTS_SMP
erts_smp_atomic_t erts_got_sigusr1;
#define ERTS_SET_GOT_SIGUSR1 \
  erts_smp_atomic_set(&erts_got_sigusr1, 1)
#define ERTS_UNSET_GOT_SIGUSR1 \
  erts_smp_atomic_set(&erts_got_sigusr1, 0)
static erts_smp_atomic_t have_prepared_crash_dump;
#define ERTS_PREPARED_CRASH_DUMP \
  ((int) erts_smp_atomic_xchg(&have_prepared_crash_dump, 1))
#else
volatile int erts_got_sigusr1;
#define ERTS_SET_GOT_SIGUSR1 (erts_got_sigusr1 = 1)
#define ERTS_UNSET_GOT_SIGUSR1 (erts_got_sigusr1 = 0)
static volatile int have_prepared_crash_dump;
#define ERTS_PREPARED_CRASH_DUMP \
  (have_prepared_crash_dump++)
#endif

static erts_smp_atomic_t sys_misc_mem_sz;

#if defined(ERTS_SMP)
static void smp_sig_notify(char c);
static int sig_notify_fds[2] = {-1, -1};
#elif defined(USE_THREADS)
static int async_fd[2];
#endif

#if CHLDWTHR || defined(ERTS_SMP)
erts_mtx_t chld_stat_mtx;
#endif
#if CHLDWTHR
static erts_tid_t child_waiter_tid;
/* chld_stat_mtx is used to protect against concurrent accesses
   of the driver_data fields pid, alive, and status. */
erts_cnd_t chld_stat_cnd;
static long children_alive;
#define CHLD_STAT_LOCK		erts_mtx_lock(&chld_stat_mtx)
#define CHLD_STAT_UNLOCK	erts_mtx_unlock(&chld_stat_mtx)
#define CHLD_STAT_WAIT		erts_cnd_wait(&chld_stat_cnd, &chld_stat_mtx)
#define CHLD_STAT_SIGNAL	erts_cnd_signal(&chld_stat_cnd)
#elif defined(ERTS_SMP) /* ------------------------------------------------- */
#define CHLD_STAT_LOCK		erts_mtx_lock(&chld_stat_mtx)
#define CHLD_STAT_UNLOCK	erts_mtx_unlock(&chld_stat_mtx)

#else /* ------------------------------------------------------------------- */
#define CHLD_STAT_LOCK
#define CHLD_STAT_UNLOCK
static volatile int children_died;
#endif


static struct fd_data {
    char  pbuf[4];   /* hold partial packet bytes */
    int   psz;       /* size of pbuf */
    char  *buf;
    char  *cpos;
    int   sz;
    int   remain;  /* for input on fd */
} *fd_data;			/* indexed by fd */

/* static FUNCTION(int, write_fill, (int, char*, int)); unused? */
static FUNCTION(void, note_child_death, (int, int));

#if CHLDWTHR
static FUNCTION(void *, child_waiter, (void *));
#endif

/********************* General functions ****************************/

/* This is used by both the drivers and general I/O, must be set early */
static int max_files = -1;

/* 
 * a few variables used by the break handler 
 */
#ifdef ERTS_SMP
erts_smp_atomic_t erts_break_requested;
#define ERTS_SET_BREAK_REQUESTED \
  erts_smp_atomic_set(&erts_break_requested, (long) 1)
#define ERTS_UNSET_BREAK_REQUESTED \
  erts_smp_atomic_set(&erts_break_requested, (long) 0)
#else
volatile int erts_break_requested = 0;
#define ERTS_SET_BREAK_REQUESTED (erts_break_requested = 1)
#define ERTS_UNSET_BREAK_REQUESTED (erts_break_requested = 0)
#endif
/* set early so the break handler has access to initial mode */
static struct termios initial_tty_mode;
static int replace_intr = 0;
/* assume yes initially, ttsl_init will clear it */
int using_oldshell = 1; 

#ifdef ERTS_ENABLE_KERNEL_POLL

int erts_use_kernel_poll = 0;

struct {
    int (*select)(ErlDrvPort, ErlDrvEvent, int, int);
    int (*event)(ErlDrvPort, ErlDrvEvent, ErlDrvEventData);
    void (*check_io_interrupt)(int);
    void (*check_io_interrupt_tmd)(int, long);
    void (*check_io)(int);
    Uint (*size)(void);
    Eterm (*info)(void *);
    int (*check_io_debug)(void);
} io_func = {0};


int
driver_select(ErlDrvPort port, ErlDrvEvent event, int mode, int on)
{
    return (*io_func.select)(port, event, mode, on);
}

int
driver_event(ErlDrvPort port, ErlDrvEvent event, ErlDrvEventData event_data)
{
    return (*io_func.event)(port, event, event_data);
}

Eterm erts_check_io_info(void *p)
{
    return (*io_func.info)(p);
}

int
erts_check_io_debug(void)
{
    return (*io_func.check_io_debug)();
}


static void
init_check_io(void)
{
    if (erts_use_kernel_poll) {
	io_func.select			= driver_select_kp;
	io_func.event			= driver_event_kp;
	io_func.check_io_interrupt	= erts_check_io_interrupt_kp;
	io_func.check_io_interrupt_tmd	= erts_check_io_interrupt_timed_kp;
	io_func.check_io		= erts_check_io_kp;
	io_func.size			= erts_check_io_size_kp;
	io_func.info			= erts_check_io_info_kp;
	io_func.check_io_debug		= erts_check_io_debug_kp;
	erts_init_check_io_kp();
	max_files = erts_check_io_max_files_kp();
    }
    else {
	io_func.select			= driver_select_nkp;
	io_func.event			= driver_event_nkp;
	io_func.check_io_interrupt	= erts_check_io_interrupt_nkp;
	io_func.check_io_interrupt_tmd	= erts_check_io_interrupt_timed_nkp;
	io_func.check_io		= erts_check_io_nkp;
	io_func.size			= erts_check_io_size_nkp;
	io_func.info			= erts_check_io_info_nkp;
	io_func.check_io_debug		= erts_check_io_debug_nkp;
	erts_init_check_io_nkp();
	max_files = erts_check_io_max_files_nkp();
    }
}

#define ERTS_CHK_IO_INTR	(*io_func.check_io_interrupt)
#define ERTS_CHK_IO_INTR_TMD	(*io_func.check_io_interrupt_tmd)
#define ERTS_CHK_IO		(*io_func.check_io)
#define ERTS_CHK_IO_SZ		(*io_func.size)

#else /* !ERTS_ENABLE_KERNEL_POLL */

static void
init_check_io(void)
{
    erts_init_check_io();
    max_files = erts_check_io_max_files();
}

#define ERTS_CHK_IO_INTR	erts_check_io_interrupt
#define ERTS_CHK_IO_INTR_TMD	erts_check_io_interrupt_timed
#define ERTS_CHK_IO		erts_check_io
#define ERTS_CHK_IO_SZ		erts_check_io_size

#endif

#ifdef ERTS_SMP
void
erts_sys_schedule_interrupt(int set)
{
    ERTS_CHK_IO_INTR(set);
}

void
erts_sys_schedule_interrupt_timed(int set, long msec)
{
    ERTS_CHK_IO_INTR_TMD(set, msec);
}
#endif

Uint
erts_sys_misc_mem_sz(void)
{
    Uint res = ERTS_CHK_IO_SZ();
    res += erts_smp_atomic_read(&sys_misc_mem_sz);
    return res;
}

/*
 * reset the terminal to the original settings on exit
 */
void sys_tty_reset(int exit_code)
{
  if (using_oldshell && !replace_intr) {
    SET_BLOCKING(0);
  }
  else if (isatty(0)) {
    tcsetattr(0,TCSANOW,&initial_tty_mode);
  }
}

#ifdef __tile__
/* Direct malloc to spread memory around the caches of multiple tiles. */
#include <malloc.h>
MALLOC_USE_HASH(1);
#endif

#ifdef USE_THREADS
static void *ethr_internal_alloc(size_t size)
{
    return erts_alloc_fnf(ERTS_ALC_T_ETHR_INTERNAL, (Uint) size);
}
static void *ethr_internal_realloc(void *ptr, size_t size)
{
    return erts_realloc_fnf(ERTS_ALC_T_ETHR_INTERNAL, ptr, (Uint) size);
}
static void ethr_internal_free(void *ptr)
{
    erts_free(ERTS_ALC_T_ETHR_INTERNAL, ptr);
}

#ifdef ERTS_THR_HAVE_SIG_FUNCS
/*
 * Child thread inherits parents signal mask at creation. In order to
 * guarantee that the main thread will receive all SIGINT, SIGCHLD, and
 * SIGUSR1 signals sent to the process, we block these signals in the
 * parent thread when creating a new thread.
 */

static sigset_t thr_create_sigmask;

#endif /* #ifdef ERTS_THR_HAVE_SIG_FUNCS */

typedef struct {
#ifdef ERTS_THR_HAVE_SIG_FUNCS
    sigset_t saved_sigmask;
#endif
    int unbind_child;
} erts_thr_create_data_t;

/*
 * thr_create_prepare() is called in parent thread before thread creation.
 * Returned value is passed as argument to thr_create_cleanup().
 */
static void *
thr_create_prepare(void)
{
    erts_thr_create_data_t *tcdp;
    ErtsSchedulerData *esdp;

    tcdp = erts_alloc(ERTS_ALC_T_TMP, sizeof(erts_thr_create_data_t));

#ifdef ERTS_THR_HAVE_SIG_FUNCS
    erts_thr_sigmask(SIG_BLOCK, &thr_create_sigmask, &tcdp->saved_sigmask);
#endif
    esdp = erts_get_scheduler_data();
    tcdp->unbind_child = esdp && erts_is_scheduler_bound(esdp);

    return (void *) tcdp;
}


/* thr_create_cleanup() is called in parent thread after thread creation. */
static void
thr_create_cleanup(void *vtcdp)
{
    erts_thr_create_data_t *tcdp = (erts_thr_create_data_t *) vtcdp;

#ifdef ERTS_THR_HAVE_SIG_FUNCS
    /* Restore signalmask... */
    erts_thr_sigmask(SIG_SETMASK, &tcdp->saved_sigmask, NULL);
#endif

    erts_free(ERTS_ALC_T_TMP, tcdp);
}

static void
thr_create_prepare_child(void *vtcdp)
{
    erts_thr_create_data_t *tcdp = (erts_thr_create_data_t *) vtcdp;

#ifdef ERTS_ENABLE_LOCK_COUNT
    erts_lcnt_thread_setup();
#endif

#ifndef NO_FPE_SIGNALS
    /*
     * We do not want fp exeptions in other threads than the
     * scheduler threads. We enable fpe explicitly in the scheduler
     * threads after this.
     */
    erts_thread_disable_fpe();
#endif

    if (tcdp->unbind_child) {
	erts_smp_rwmtx_rlock(&erts_cpu_bind_rwmtx);
	erts_unbind_from_cpu(erts_cpuinfo);
	erts_smp_rwmtx_runlock(&erts_cpu_bind_rwmtx);
    }
    
}

#endif /* #ifdef USE_THREADS */

void
erts_sys_pre_init(void)
{
    erts_printf_add_cr_to_stdout = 1;
    erts_printf_add_cr_to_stderr = 1;
#ifdef USE_THREADS
    {
    erts_thr_init_data_t eid = ERTS_THR_INIT_DATA_DEF_INITER;
    eid.alloc = ethr_internal_alloc;
    eid.realloc = ethr_internal_realloc;
    eid.free = ethr_internal_free;

    eid.thread_create_child_func = thr_create_prepare_child;
    /* Before creation in parent */
    eid.thread_create_prepare_func = thr_create_prepare;
    /* After creation in parent */
    eid.thread_create_parent_func = thr_create_cleanup,

#ifdef ERTS_THR_HAVE_SIG_FUNCS
    sigemptyset(&thr_create_sigmask);
    sigaddset(&thr_create_sigmask, SIGINT);   /* block interrupt */
    sigaddset(&thr_create_sigmask, SIGCHLD);  /* block child signals */
    sigaddset(&thr_create_sigmask, SIGUSR1);  /* block user defined signal */
#endif

    erts_thr_init(&eid);

    report_exit_list = NULL;

#ifdef ERTS_ENABLE_LOCK_COUNT
    erts_lcnt_init();
#endif

#if CHLDWTHR || defined(ERTS_SMP)
    erts_mtx_init(&chld_stat_mtx, "child_status");
#endif
#if CHLDWTHR
#ifndef ERTS_SMP
    report_exit_transit_list = NULL;
#endif
    erts_cnd_init(&chld_stat_cnd);
    children_alive = 0;
#endif
    }
#ifdef ERTS_SMP
    erts_smp_atomic_init(&erts_break_requested, 0);
    erts_smp_atomic_init(&erts_got_sigusr1, 0);
    erts_smp_atomic_init(&have_prepared_crash_dump, 0);
#else
    erts_break_requested = 0;
    erts_got_sigusr1 = 0;
    have_prepared_crash_dump = 0;
#endif
#if !CHLDWTHR && !defined(ERTS_SMP)
    children_died = 0;
#endif
#endif /* USE_THREADS */
    erts_smp_atomic_init(&sys_misc_mem_sz, 0);
    erts_smp_rwmtx_init(&environ_rwmtx, "environ");
}

void
erl_sys_init(void)
{
#if !DISABLE_VFORK
    int res;
    char bindir[MAXPATHLEN];
    size_t bindirsz = sizeof(bindir);
    Uint csp_path_sz;

    res = erts_sys_getenv("BINDIR", bindir, &bindirsz);
    if (res != 0) {
	if (res < 0)
	    erl_exit(-1,
		     "Environment variable BINDIR is not set\n");
	if (res > 0)
	    erl_exit(-1,
		     "Value of environment variable BINDIR is too large\n");
    }
    if (bindir[0] != DIR_SEPARATOR_CHAR)
	erl_exit(-1,
		 "Environment variable BINDIR does not contain an"
		 " absolute path\n");
    csp_path_sz = (strlen(bindir)
		   + 1 /* DIR_SEPARATOR_CHAR */
		   + sizeof(CHILD_SETUP_PROG_NAME)
		   + 1);
    child_setup_prog = erts_alloc(ERTS_ALC_T_CS_PROG_PATH, csp_path_sz);
    erts_smp_atomic_add(&sys_misc_mem_sz, csp_path_sz);
    sprintf(child_setup_prog,
            "%s%c%s",
            bindir,
            DIR_SEPARATOR_CHAR,
            CHILD_SETUP_PROG_NAME);
#endif

#ifdef USE_SETLINEBUF
    setlinebuf(stdout);
#else
    setvbuf(stdout, (char *)NULL, _IOLBF, BUFSIZ);
#endif

    erts_sys_init_float();

    /* we save this so the break handler can set and reset it properly */
    /* also so that we can reset on exit (break handler or not) */
    if (isatty(0)) {
	tcgetattr(0,&initial_tty_mode);
    }
    tzset(); /* Required at least for NetBSD with localtime_r() */
}

/* signal handling */

#ifdef SIG_SIGSET		/* Old SysV */
RETSIGTYPE (*sys_sigset(sig, func))()
int sig;
RETSIGTYPE (*func)();
{
    return(sigset(sig, func));
}
void sys_sigblock(int sig)
{
    sighold(sig);
}
void sys_sigrelease(int sig)
{
    sigrelse(sig);
}
#else /* !SIG_SIGSET */
#ifdef SIG_SIGNAL		/* Old BSD */
RETSIGTYPE (*sys_sigset(sig, func))(int, int)
int sig;
RETSIGTYPE (*func)();
{
    return(signal(sig, func));
}
sys_sigblock(int sig)
{
    sigblock(sig);
}
sys_sigrelease(int sig)
{
    sigsetmask(sigblock(0) & ~sigmask(sig));
}
#else /* !SIG_SIGNAL */	/* The True Way - POSIX!:-) */
RETSIGTYPE (*sys_sigset(int sig, RETSIGTYPE (*func)(int)))(int)
{
    struct sigaction act, oact;

    sigemptyset(&act.sa_mask);
    act.sa_flags = 0;
    act.sa_handler = func;
    sigaction(sig, &act, &oact);
    return(oact.sa_handler);
}

#ifdef USE_THREADS
#undef  sigprocmask
#define sigprocmask erts_thr_sigmask
#endif

void sys_sigblock(int sig)
{
    sigset_t mask;

    sigemptyset(&mask);
    sigaddset(&mask, sig);
    sigprocmask(SIG_BLOCK, &mask, (sigset_t *)NULL);
}

void sys_sigrelease(int sig)
{
    sigset_t mask;

    sigemptyset(&mask);
    sigaddset(&mask, sig);
    sigprocmask(SIG_UNBLOCK, &mask, (sigset_t *)NULL);
}
#endif /* !SIG_SIGNAL */
#endif /* !SIG_SIGSET */

#if (0) /* not used? -- gordon */
static void (*break_func)();
static RETSIGTYPE break_handler(int sig)
{
#ifdef QNX
    /* Turn off SIGCHLD during break processing */
    sys_sigblock(SIGCHLD);
#endif
    (*break_func)();
#ifdef QNX
    sys_sigrelease(SIGCHLD);
#endif
}
#endif /* 0 */

static ERTS_INLINE void
prepare_crash_dump(void)
{
    int i, max;
    char env[21]; /* enough to hold any 64-bit integer */
    size_t envsz;

    if (ERTS_PREPARED_CRASH_DUMP)
	return; /* We have already been called */

    /* Make sure we unregister at epmd (unknown fd) and get at least
       one free filedescriptor (for erl_crash.dump) */
    max = max_files;
    if (max < 1024)
	max = 1024;
    for (i = 3; i < max; i++) {
#if defined(ERTS_SMP)
	/* We don't want to close the signal notification pipe... */
	if (i == sig_notify_fds[0] || i == sig_notify_fds[1])
	    continue;
#elif defined(USE_THREADS)
	/* We don't want to close the async notification pipe... */
	if (i == async_fd[0] || i == async_fd[1])
	    continue;
#endif
	close(i);
    }

    envsz = sizeof(env);
    i = erts_sys_getenv("ERL_CRASH_DUMP_NICE", env, &envsz);
    if (i >= 0) {
	int nice_val;
	nice_val = i != 0 ? 0 : atoi(env);
	if (nice_val > 39) {
	    nice_val = 39;
	}
	erts_silence_warn_unused_result(nice(nice_val));
    }
    
    envsz = sizeof(env);
    i = erts_sys_getenv("ERL_CRASH_DUMP_SECONDS", env, &envsz);
    if (i >= 0) {
	unsigned sec;
	sec = (unsigned) i != 0 ? 0 : atoi(env);
	alarm(sec);
    }

}

void
erts_sys_prepare_crash_dump(void)
{
    prepare_crash_dump();
}

static ERTS_INLINE void
break_requested(void)
{
  /*
   * just set a flag - checked for and handled by
   * scheduler threads erts_check_io() (not signal handler).
   */
#ifdef DEBUG			
  fprintf(stderr,"break!\n");
#endif
  if (ERTS_BREAK_REQUESTED)
      erl_exit(ERTS_INTR_EXIT, "");

  ERTS_SET_BREAK_REQUESTED;
  ERTS_CHK_IO_INTR(1); /* Make sure we don't sleep in poll */
}

/* set up signal handlers for break and quit */
#if (defined(SIG_SIGSET) || defined(SIG_SIGNAL))
static RETSIGTYPE request_break(void)
#else
static RETSIGTYPE request_break(int signum)
#endif
{
#ifdef ERTS_SMP
    smp_sig_notify('I');
#else
    break_requested();
#endif
}

static ERTS_INLINE void
sigusr1_exit(void)
{
   /* We do this at interrupt level, since the main reason for
      wanting to generate a crash dump in this way is that the emulator
      is hung somewhere, so it won't be able to poll any flag we set here.
      */
    ERTS_SET_GOT_SIGUSR1;
    prepare_crash_dump();
    erl_exit(1, "Received SIGUSR1\n");
}

#ifdef ETHR_UNUSABLE_SIGUSRX
#warning "Unusable SIGUSR1 & SIGUSR2. Disabling use of these signals"
#endif

#ifndef ETHR_UNUSABLE_SIGUSRX

#if (defined(SIG_SIGSET) || defined(SIG_SIGNAL))
static RETSIGTYPE user_signal1(void)
#else
static RETSIGTYPE user_signal1(int signum)
#endif
{
#ifdef ERTS_SMP
   smp_sig_notify('1');
#else
   sigusr1_exit();
#endif
}

#ifdef QUANTIFY
#if (defined(SIG_SIGSET) || defined(SIG_SIGNAL))
static RETSIGTYPE user_signal2(void)
#else
static RETSIGTYPE user_signal2(int signum)
#endif
{
#ifdef ERTS_SMP
   smp_sig_notify('2');
#else
   quantify_save_data();
#endif
}
#endif

#endif /* #ifndef ETHR_UNUSABLE_SIGUSRX */

static void
quit_requested(void)
{
    erl_exit(ERTS_INTR_EXIT, "");
}

#if (defined(SIG_SIGSET) || defined(SIG_SIGNAL))
static RETSIGTYPE do_quit(void)
#else
static RETSIGTYPE do_quit(int signum)
#endif
{
#ifdef ERTS_SMP
    smp_sig_notify('Q');
#else
    quit_requested();
#endif
}

/* Disable break */
void erts_set_ignore_break(void) {
    sys_sigset(SIGINT,  SIG_IGN);
    sys_sigset(SIGQUIT, SIG_IGN);
    sys_sigset(SIGTSTP, SIG_IGN);
}

/* Don't use ctrl-c for break handler but let it be 
   used by the shell instead (see user_drv.erl) */
void erts_replace_intr(void) {
  struct termios mode;

  if (isatty(0)) {
    tcgetattr(0, &mode);
    
    /* here's an example of how to replace ctrl-c with ctrl-u */
    /* mode.c_cc[VKILL] = 0;
       mode.c_cc[VINTR] = CKILL; */
    
    mode.c_cc[VINTR] = 0;	/* disable ctrl-c */
    tcsetattr(0, TCSANOW, &mode);
    replace_intr = 1;
  }
}

void init_break_handler(void)
{
   sys_sigset(SIGINT, request_break);
#ifndef ETHR_UNUSABLE_SIGUSRX
   sys_sigset(SIGUSR1, user_signal1);
#ifdef QUANTIFY
   sys_sigset(SIGUSR2, user_signal2);
#endif
#endif /* #ifndef ETHR_UNUSABLE_SIGUSRX */
   sys_sigset(SIGQUIT, do_quit);
}

int sys_max_files(void)
{
   return(max_files);
}

static void block_signals(void)
{
#if !CHLDWTHR
   sys_sigblock(SIGCHLD);
#endif
#ifndef ERTS_SMP
   sys_sigblock(SIGINT);
#ifndef ETHR_UNUSABLE_SIGUSRX
   sys_sigblock(SIGUSR1);
#endif
#endif
}

static void unblock_signals(void)
{
    /* Update erl_child_setup.c if changed */
#if !CHLDWTHR
   sys_sigrelease(SIGCHLD);
#endif
#ifndef ERTS_SMP
   sys_sigrelease(SIGINT);
#ifndef ETHR_UNUSABLE_SIGUSRX
   sys_sigrelease(SIGUSR1);
#endif /* #ifndef ETHR_UNUSABLE_SIGUSRX */
#endif
}
/************************** Time stuff **************************/
#ifdef HAVE_GETHRTIME
#ifdef GETHRTIME_WITH_CLOCK_GETTIME

SysHrTime sys_gethrtime(void)
{
    struct timespec ts;
    long long result;
    if (clock_gettime(CLOCK_MONOTONIC,&ts) != 0) {
	erl_exit(1,"Fatal, could not get clock_monotonic value!, "
		 "errno = %d\n", errno);
    }
    result = ((long long) ts.tv_sec) * 1000000000LL + 
	((long long) ts.tv_nsec);
    return (SysHrTime) result;
}
#endif
#endif

/************************** OS info *******************************/

/* Used by erlang:info/1. */
/* (This code was formerly in drv.XXX/XXX_os_drv.c) */

char os_type[] = "unix";

static int
get_number(char **str_ptr)
{
    char* s = *str_ptr;		/* Pointer to beginning of string. */
    char* dot;			/* Pointer to dot in string or NULL. */

    if (!isdigit((int) *s))
	return 0;
    if ((dot = strchr(s, '.')) == NULL) {
	*str_ptr = s+strlen(s);
	return atoi(s);
    } else {
	*dot = '\0';
	*str_ptr = dot+1;
	return atoi(s);
    }
}

void
os_flavor(char* namebuf, 	/* Where to return the name. */
	  unsigned size) 	/* Size of name buffer. */
{
    static int called = 0;
    static struct utsname uts;	/* Information about the system. */

    if (!called) {
	char* s;

	(void) uname(&uts);
	called = 1;
	for (s = uts.sysname; *s; s++) {
	    if (isupper((int) *s)) {
		*s = tolower((int) *s);
	    }
	}
    }
    strcpy(namebuf, uts.sysname);
}

void
os_version(pMajor, pMinor, pBuild)
int* pMajor;			/* Pointer to major version. */
int* pMinor;			/* Pointer to minor version. */
int* pBuild;			/* Pointer to build number. */
{
    struct utsname uts;		/* Information about the system. */
    char* release;		/* Pointer to the release string:
				 * X.Y or X.Y.Z.
				 */

    (void) uname(&uts);
    release = uts.release;
    *pMajor = get_number(&release);
    *pMinor = get_number(&release);
    *pBuild = get_number(&release);
}

void init_getenv_state(GETENV_STATE *state)
{
   erts_smp_rwmtx_rlock(&environ_rwmtx);
   *state = NULL;
}

char *getenv_string(GETENV_STATE *state0)
{
   char **state = (char **) *state0;
   char *cp;

   ERTS_SMP_LC_ASSERT(erts_smp_lc_rwmtx_is_rlocked(&environ_rwmtx));

   if (state == NULL)
      state = environ;

   cp = *state++;
   *state0 = (GETENV_STATE) state;

   return cp;
}

void fini_getenv_state(GETENV_STATE *state)
{
   *state = NULL;
   erts_smp_rwmtx_runlock(&environ_rwmtx);
}


/************************** Port I/O *******************************/



/* I. Common stuff */

/*
 * Decreasing the size of it below 16384 is not allowed.
 */

/* II. The spawn/fd/vanilla drivers */

#define ERTS_SYS_READ_BUF_SZ (64*1024)

/* This data is shared by these drivers - initialized by spawn_init() */
static struct driver_data {
    int port_num, ofd, packet_bytes;
    ErtsSysReportExit *report_exit;
    int pid;
    int alive;
    int status;
} *driver_data;			/* indexed by fd */

/* Driver interfaces */
static ErlDrvData spawn_start(ErlDrvPort, char*, SysDriverOpts*);
static ErlDrvData fd_start(ErlDrvPort, char*, SysDriverOpts*);
static int fd_control(ErlDrvData, unsigned int, char *, int, char **, int);
static ErlDrvData vanilla_start(ErlDrvPort, char*, SysDriverOpts*);
static int spawn_init(void);
static void fd_stop(ErlDrvData);
static void stop(ErlDrvData);
static void ready_input(ErlDrvData, ErlDrvEvent);
static void ready_output(ErlDrvData, ErlDrvEvent);
static void output(ErlDrvData, char*, int);
static void outputv(ErlDrvData, ErlIOVec*);
static void stop_select(ErlDrvEvent, void*);

struct erl_drv_entry spawn_driver_entry = {
    spawn_init,
    spawn_start,
    stop,
    output,
    ready_input,
    ready_output,
    "spawn",
    NULL,
    NULL,
    NULL,
    NULL,
    NULL,
    NULL,
    NULL,
    NULL,
    NULL,
    ERL_DRV_EXTENDED_MARKER,
    ERL_DRV_EXTENDED_MAJOR_VERSION,
    ERL_DRV_EXTENDED_MINOR_VERSION,
    ERL_DRV_FLAG_USE_PORT_LOCKING,
    NULL, NULL,
    stop_select
};
struct erl_drv_entry fd_driver_entry = {
    NULL,
    fd_start,
    fd_stop,
    output,
    ready_input,
    ready_output, 
    "fd",
    NULL,
    NULL,
    fd_control,
    NULL,
    outputv,
    NULL, /* ready_async */
    NULL, /* flush */
    NULL, /* call */
    NULL, /* event */
    ERL_DRV_EXTENDED_MARKER,
    ERL_DRV_EXTENDED_MAJOR_VERSION,
    ERL_DRV_EXTENDED_MINOR_VERSION,
    0, /* ERL_DRV_FLAGs */
    NULL, /* handle2 */
    NULL, /* process_exit */
    stop_select
};
struct erl_drv_entry vanilla_driver_entry = {
    NULL,
    vanilla_start,
    stop,
    output,
    ready_input,
    ready_output,
    "vanilla",
    NULL,
    NULL,
    NULL,
    NULL,
    NULL,
    NULL,
    NULL, /* flush */
    NULL, /* call */
    NULL, /* event */
    ERL_DRV_EXTENDED_MARKER,
    ERL_DRV_EXTENDED_MAJOR_VERSION,
    ERL_DRV_EXTENDED_MINOR_VERSION,
    0, /* ERL_DRV_FLAGs */
    NULL, /* handle2 */
    NULL, /* process_exit */
    stop_select
};

#if defined(USE_THREADS) && !defined(ERTS_SMP)
static int  async_drv_init(void);
static ErlDrvData async_drv_start(ErlDrvPort, char*, SysDriverOpts*);
static void async_drv_stop(ErlDrvData);
static void async_drv_input(ErlDrvData, ErlDrvEvent);

/* INTERNAL use only */

struct erl_drv_entry async_driver_entry = {
    async_drv_init,
    async_drv_start,
    async_drv_stop,
    NULL,
    async_drv_input,
    NULL,
    "async",
    NULL,
    NULL,
    NULL,
    NULL,
    NULL,
    NULL
};
#endif

/* Handle SIGCHLD signals. */
#if (defined(SIG_SIGSET) || defined(SIG_SIGNAL))
static RETSIGTYPE onchld(void)
#else
static RETSIGTYPE onchld(int signum)
#endif
{
#if CHLDWTHR
    ASSERT(0); /* We should *never* catch a SIGCHLD signal */
#elif defined(ERTS_SMP)
    smp_sig_notify('C');
#else
    children_died = 1;
    ERTS_CHK_IO_INTR(1); /* Make sure we don't sleep in poll */
#endif
}

static int set_driver_data(int port_num,
			   int ifd,
			   int ofd,
			   int packet_bytes,
			   int read_write,
			   int exit_status,
			   int pid)
{
    ErtsSysReportExit *report_exit;

    if (!exit_status)
	report_exit = NULL;
    else {
	report_exit = erts_alloc(ERTS_ALC_T_PRT_REP_EXIT,
				 sizeof(ErtsSysReportExit));
	report_exit->next = report_exit_list;
	report_exit->port = erts_port[port_num].id;
	report_exit->pid = pid;
	report_exit->ifd = read_write & DO_READ ? ifd : -1;
	report_exit->ofd = read_write & DO_WRITE ? ofd : -1;
#if CHLDWTHR && !defined(ERTS_SMP)
	report_exit->status = 0;
#endif
	report_exit_list = report_exit;
    }

    if (read_write & DO_READ) {
	driver_data[ifd].packet_bytes = packet_bytes;
	driver_data[ifd].port_num = port_num;
	driver_data[ifd].report_exit = report_exit;
	driver_data[ifd].pid = pid;
	driver_data[ifd].alive = 1;
	driver_data[ifd].status = 0;
	if (read_write & DO_WRITE) {
	    driver_data[ifd].ofd = ofd;
	    if (ifd != ofd)
		driver_data[ofd] = driver_data[ifd];  /* structure copy */
	} else {		/* DO_READ only */
	    driver_data[ifd].ofd = -1;
	}
	(void) driver_select(port_num, ifd, (ERL_DRV_READ|ERL_DRV_USE), 1);
	return(ifd);
    } else {			/* DO_WRITE only */
	driver_data[ofd].packet_bytes = packet_bytes;
	driver_data[ofd].port_num = port_num;
	driver_data[ofd].report_exit = report_exit;
	driver_data[ofd].ofd = ofd;
	driver_data[ofd].pid = pid;
	driver_data[ofd].alive = 1;
	driver_data[ofd].status = 0;
	return(ofd);
    }
}

static int spawn_init()
{
   int i;
#if CHLDWTHR
   erts_thr_opts_t thr_opts = ERTS_THR_OPTS_DEFAULT_INITER;
   thr_opts.detached = 0;
   thr_opts.suggested_stack_size = 0; /* Smallest possible */
#endif

   sys_sigset(SIGPIPE, SIG_IGN); /* Ignore - we'll handle the write failure */
   driver_data = (struct driver_data *)
       erts_alloc(ERTS_ALC_T_DRV_TAB, max_files * sizeof(struct driver_data));
   erts_smp_atomic_add(&sys_misc_mem_sz,
		       max_files * sizeof(struct driver_data));

   for (i = 0; i < max_files; i++)
      driver_data[i].pid = -1;

#if CHLDWTHR
   sys_sigblock(SIGCHLD);
#endif

   sys_sigset(SIGCHLD, onchld); /* Reap children */

#if CHLDWTHR
   erts_thr_create(&child_waiter_tid, child_waiter, NULL, &thr_opts);
#endif

   return 1;
}

static void close_pipes(int ifd[2], int ofd[2], int read_write)
{
    if (read_write & DO_READ) {
	(void) close(ifd[0]);
	(void) close(ifd[1]);
    }
    if (read_write & DO_WRITE) {
	(void) close(ofd[0]);
	(void) close(ofd[1]);
    }
}

static void init_fd_data(int fd, int prt)
{
    fd_data[fd].buf = NULL;
    fd_data[fd].cpos = NULL;
    fd_data[fd].remain = 0;
    fd_data[fd].sz = 0;
    fd_data[fd].psz = 0;
}

static char **build_unix_environment(char *block)
{
    int i;
    int j;
    int len;
    char *cp;
    char **cpp;
    char** old_env;
    
    ERTS_SMP_LC_ASSERT(erts_smp_lc_rwmtx_is_rlocked(&environ_rwmtx));

    cp = block;
    len = 0;
    while (*cp != '\0') {
	cp += strlen(cp) + 1;
	len++;
    }
    old_env = environ;
    while (*old_env++ != NULL) {
	len++;
    }
    
    cpp = (char **) erts_alloc_fnf(ERTS_ALC_T_ENVIRONMENT,
				   sizeof(char *) * (len+1));
    if (cpp == NULL) {
	return NULL;
    }

    cp = block;
    len = 0;
    while (*cp != '\0') {
	cpp[len] = cp;
	cp += strlen(cp) + 1;
	len++;
    }
    
    i = len;
    for (old_env = environ; *old_env; old_env++) {
	char* old = *old_env;

	for (j = 0; j < len; j++) {
	    char *s, *t;

	    s = cpp[j];
	    t = old;
	    while (*s == *t && *s != '=') {
		s++, t++;
	    }
	    if (*s == '=' && *t == '=') {
		break;
	    }
	}

	if (j == len) {		/* New version not found */
	    cpp[len++] = old;
	}
    }

    for (j = 0; j < i; j++) {
        size_t last = strlen(cpp[j])-1;
	if (cpp[j][last] == '=' && strchr(cpp[j], '=') == cpp[j]+last) {
	    cpp[j] = cpp[--len];
	}
    }

    cpp[len] = NULL;
    return cpp;
}

/*
  [arndt] In most Unix systems, including Solaris 2.5, 'fork' allocates memory
  in swap space for the child of a 'fork', whereas 'vfork' does not do this.
  The natural call to use here is therefore 'vfork'. Due to a bug in
  'vfork' in Solaris 2.5 (apparently fixed in 2.6), using 'vfork'
  can be dangerous in what seems to be these circumstances:
      If the child code under a vfork sets the signal action to SIG_DFL
      (or SIG_IGN)
      for any signal which was previously set to a signal handler, the
      state of the parent is clobbered, so that the later arrival of
      such a signal yields a sigsegv in the parent. If the signal was
      not set to a signal handler, but ignored, all seems to work.
  If you change the forking code below, beware of this.
 */

static ErlDrvData spawn_start(ErlDrvPort port_num, char* name, SysDriverOpts* opts)
{
#define CMD_LINE_PREFIX_STR "exec "
#define CMD_LINE_PREFIX_STR_SZ (sizeof(CMD_LINE_PREFIX_STR) - 1)

    int ifd[2], ofd[2], len, pid, i;
    char **volatile new_environ; /* volatile since a vfork() then cannot
				    cause 'new_environ' to be clobbered
				    in the parent process. */
    int saved_errno;
    long res;
    char *cmd_line;
#ifndef QNX
    int unbind;
#endif
#if !DISABLE_VFORK
    int no_vfork;
    size_t no_vfork_sz = sizeof(no_vfork);

    no_vfork = (erts_sys_getenv("ERL_NO_VFORK",
				(char *) &no_vfork,
				&no_vfork_sz) >= 0);
#endif

    switch (opts->read_write) {
    case DO_READ:
	if (pipe(ifd) < 0)
	    return ERL_DRV_ERROR_ERRNO;
	if (ifd[0] >= max_files) {
	    close_pipes(ifd, ofd, opts->read_write);
	    errno = EMFILE;
	    return ERL_DRV_ERROR_ERRNO;
	}
	ofd[1] = -1;		/* keep purify happy */
	break;
    case DO_WRITE:
	if (pipe(ofd) < 0) return ERL_DRV_ERROR_ERRNO;
	if (ofd[1] >= max_files) {
	    close_pipes(ifd, ofd, opts->read_write);
	    errno = EMFILE;
	    return ERL_DRV_ERROR_ERRNO;
	}
	ifd[0] = -1;		/* keep purify happy */
	break;
    case DO_READ|DO_WRITE:
	if (pipe(ifd) < 0) return ERL_DRV_ERROR_ERRNO;
	errno = EMFILE;		/* default for next two conditions */
	if (ifd[0] >= max_files || pipe(ofd) < 0) {
	    close_pipes(ifd, ofd, DO_READ);
	    return ERL_DRV_ERROR_ERRNO;
	}
	if (ofd[1] >= max_files) {
	    close_pipes(ifd, ofd, opts->read_write);
	    errno = EMFILE;
	    return ERL_DRV_ERROR_ERRNO;
	}
	break;
    default:
	ASSERT(0);
	return ERL_DRV_ERROR_GENERAL;
    }

    if (opts->spawn_type == ERTS_SPAWN_EXECUTABLE) {
	/* started with spawn_executable, not with spawn */
	len = strlen(name);
	cmd_line = (char *) erts_alloc_fnf(ERTS_ALC_T_TMP, len + 1);
	if (!cmd_line) {
	    close_pipes(ifd, ofd, opts->read_write);
	    errno = ENOMEM;
	    return ERL_DRV_ERROR_ERRNO;
	}
	memcpy((void *) cmd_line,(void *) name, len);
	cmd_line[len] = '\0';
	if (access(cmd_line,X_OK) != 0) {
	    int save_errno = errno;
	    erts_free(ERTS_ALC_T_TMP, cmd_line);
	    errno = save_errno;
	    return ERL_DRV_ERROR_ERRNO;
	}
    } else {
	/* make the string suitable for giving to "sh" */
	len = strlen(name);
	cmd_line = (char *) erts_alloc_fnf(ERTS_ALC_T_TMP,
					   CMD_LINE_PREFIX_STR_SZ + len + 1);
	if (!cmd_line) {
	    close_pipes(ifd, ofd, opts->read_write);
	    errno = ENOMEM;
	    return ERL_DRV_ERROR_ERRNO;
	}
	memcpy((void *) cmd_line,
	       (void *) CMD_LINE_PREFIX_STR,
	       CMD_LINE_PREFIX_STR_SZ);
	memcpy((void *) (cmd_line + CMD_LINE_PREFIX_STR_SZ), (void *) name, len);
	cmd_line[CMD_LINE_PREFIX_STR_SZ + len] = '\0';
    }

    erts_smp_rwmtx_rlock(&environ_rwmtx);

    if (opts->envir == NULL) {
	new_environ = environ;
    } else if ((new_environ = build_unix_environment(opts->envir)) == NULL) {
	erts_smp_rwmtx_runlock(&environ_rwmtx);
	erts_free(ERTS_ALC_T_TMP, (void *) cmd_line);
	errno = ENOMEM;
	return ERL_DRV_ERROR_ERRNO;
    } 

#ifndef QNX
    /* Block child from SIGINT and SIGUSR1. Must be before fork()
       to be safe. */
    block_signals();

    CHLD_STAT_LOCK;

    unbind = erts_is_scheduler_bound(NULL);
    if (unbind)
	erts_smp_rwmtx_rlock(&erts_cpu_bind_rwmtx);

#if !DISABLE_VFORK
    /* See fork/vfork discussion before this function. */
    if (no_vfork) {
#endif

	DEBUGF(("Using fork\n"));
	pid = fork();

	if (pid == 0) {
	    /* The child! Setup child... */

	    if (unbind && erts_unbind_from_cpu(erts_cpuinfo) != 0)
		goto child_error;

	    /* OBSERVE!
	     * Keep child setup after vfork() (implemented below and in
	     * erl_child_setup.c) up to date if changes are made here.
	     */

	    if (opts->use_stdio) {
		if (opts->read_write & DO_READ) {
		    /* stdout for process */
		    if (dup2(ifd[1], 1) < 0)
			goto child_error;
		    if(opts->redir_stderr)
			/* stderr for process */
			if (dup2(ifd[1], 2) < 0)
			    goto child_error;
		}
		if (opts->read_write & DO_WRITE)
		    /* stdin for process */
		    if (dup2(ofd[0], 0) < 0)
			goto child_error;
	    }
	    else {	/* XXX will fail if ofd[0] == 4 (unlikely..) */
		if (opts->read_write & DO_READ)
		    if (dup2(ifd[1], 4) < 0)
			goto child_error;
		if (opts->read_write & DO_WRITE)
		    if (dup2(ofd[0], 3) < 0)
			goto child_error;
	    }

	    for (i = opts->use_stdio ? 3 : 5; i < max_files; i++)
		(void) close(i);
	    
	    if (opts->wd && chdir(opts->wd) < 0)
		goto child_error;

#if defined(USE_SETPGRP_NOARGS)		/* SysV */
	    (void) setpgrp();
#elif defined(USE_SETPGRP)		/* BSD */
	    (void) setpgrp(0, getpid());
#else					/* POSIX */
	    (void) setsid();
#endif
	    
	    unblock_signals();

	    if (opts->spawn_type == ERTS_SPAWN_EXECUTABLE) {
		if (opts->argv == NULL) {
		    execle(cmd_line,cmd_line,(char *) NULL, new_environ);
		} else {
		    if (opts->argv[0] == erts_default_arg0) {
			opts->argv[0] = cmd_line;
		    }
		    execve(cmd_line, opts->argv, new_environ);
		    if (opts->argv[0] == cmd_line) {
			opts->argv[0] = erts_default_arg0;
		    }
		}
	    } else {
		execle("/bin/sh", "sh", "-c", cmd_line, (char *) NULL, new_environ);
	    }
	child_error:
	    _exit(1);
	}
#if !DISABLE_VFORK
    }
    else { /* Use vfork() */
	char **cs_argv= erts_alloc(ERTS_ALC_T_TMP,(CS_ARGV_NO_OF_ARGS + 1)*
				   sizeof(char *));
	char fd_close_range[44];                  /* 44 bytes are enough to  */
	char dup2_op[CS_ARGV_NO_OF_DUP2_OPS][44]; /* hold any "%d:%d" string */
                                                  /* on a 64-bit machine.    */

	/* Setup argv[] for the child setup program (implemented in
	   erl_child_setup.c) */
	i = 0;
	if (opts->use_stdio) {
	    if (opts->read_write & DO_READ){
		/* stdout for process */
		sprintf(&dup2_op[i++][0], "%d:%d", ifd[1], 1);
		if(opts->redir_stderr)
		    /* stderr for process */
		    sprintf(&dup2_op[i++][0], "%d:%d", ifd[1], 2);
	    }
	    if (opts->read_write & DO_WRITE)
		/* stdin for process */
		sprintf(&dup2_op[i++][0], "%d:%d", ofd[0], 0);
	} else {	/* XXX will fail if ofd[0] == 4 (unlikely..) */
	    if (opts->read_write & DO_READ)
		sprintf(&dup2_op[i++][0], "%d:%d", ifd[1], 4);
	    if (opts->read_write & DO_WRITE)
		sprintf(&dup2_op[i++][0], "%d:%d", ofd[0], 3);
	}
	for (; i < CS_ARGV_NO_OF_DUP2_OPS; i++)
	    strcpy(&dup2_op[i][0], "-");
	sprintf(fd_close_range, "%d:%d", opts->use_stdio ? 3 : 5, max_files-1);

	cs_argv[CS_ARGV_PROGNAME_IX] = child_setup_prog;
	cs_argv[CS_ARGV_WD_IX] = opts->wd ? opts->wd : ".";
	cs_argv[CS_ARGV_UNBIND_IX]
	    = (unbind ? erts_get_unbind_from_cpu_str(erts_cpuinfo) : "false");
	cs_argv[CS_ARGV_FD_CR_IX] = fd_close_range;
	for (i = 0; i < CS_ARGV_NO_OF_DUP2_OPS; i++)
	    cs_argv[CS_ARGV_DUP2_OP_IX(i)] = &dup2_op[i][0];

	if (opts->spawn_type == ERTS_SPAWN_EXECUTABLE) {
	    int num = 0;
	    int j = 0;
	    if (opts->argv != NULL) {
		for(; opts->argv[num] != NULL; ++num)
		    ;
	    }
	    cs_argv = erts_realloc(ERTS_ALC_T_TMP,cs_argv, (CS_ARGV_NO_OF_ARGS + 1 + num + 1) * sizeof(char *));
	    cs_argv[CS_ARGV_CMD_IX] = "-";
	    cs_argv[CS_ARGV_NO_OF_ARGS] = cmd_line;
	    if (opts->argv != NULL) {
		for (;opts->argv[j] != NULL; ++j) {
		    if (opts->argv[j] == erts_default_arg0) {
			cs_argv[CS_ARGV_NO_OF_ARGS + 1 + j] = cmd_line;
		    } else {
			cs_argv[CS_ARGV_NO_OF_ARGS + 1 + j] = opts->argv[j];
		    }
		}
	    }
	    cs_argv[CS_ARGV_NO_OF_ARGS + 1 + j] = NULL;
	} else {
	    cs_argv[CS_ARGV_CMD_IX] = cmd_line; /* Command */
	    cs_argv[CS_ARGV_NO_OF_ARGS] = NULL;
	}
	DEBUGF(("Using vfork\n"));
	pid = vfork();

	if (pid == 0) {
	    /* The child! */

	    /* Observe!
	     * OTP-4389: The child setup program (implemented in
	     * erl_child_setup.c) will perform the necessary setup of the
	     * child before it execs to the user program. This because
	     * vfork() only allow an *immediate* execve() or _exit() in the
	     * child.
	     */
	    execve(child_setup_prog, cs_argv, new_environ);
	    _exit(1);
	}
	erts_free(ERTS_ALC_T_TMP,cs_argv);
    }
#endif

    if (unbind)
	erts_smp_rwmtx_runlock(&erts_cpu_bind_rwmtx);

    if (pid == -1) {
        saved_errno = errno;
	CHLD_STAT_UNLOCK;
	erts_smp_rwmtx_runlock(&environ_rwmtx);
	erts_free(ERTS_ALC_T_TMP, (void *) cmd_line);
        unblock_signals();
        close_pipes(ifd, ofd, opts->read_write);
	errno = saved_errno;
	return ERL_DRV_ERROR_ERRNO;
    }
#else /* QNX */
    if (opts->use_stdio) {
	if (opts->read_write & DO_READ)
	    qnx_spawn_options.iov[1] = ifd[1];  /* stdout for process */
	if (opts->read_write & DO_WRITE)
	    qnx_spawn_options.iov[0] = ofd[0];  /* stdin for process */
	} 
    else {
	if (opts->read_write & DO_READ)
	    qnx_spawn_options.iov[4] = ifd[1];
	if (opts->read_write & DO_WRITE)
	    qnx_spawn_options.iov[3] = ofd[0];
    }
    /* Close fds on exec */
    for (i = 3; i < max_files; i++)
	fcntl(i, F_SETFD, 1);

    qnx_spawn_options.flags = _SPAWN_SETSID;
    if ((pid = spawnl(P_NOWAIT, "/bin/sh", "/bin/sh", "-c", cmd_line, 
                      (char *) 0)) < 0) {
	erts_free(ERTS_ALC_T_TMP, (void *) cmd_line);
        reset_qnx_spawn();
	erts_smp_rwmtx_runlock(&environ_rwmtx);
	close_pipes(ifd, ofd, opts->read_write);
	return ERL_DRV_ERROR_GENERAL;
    }
    reset_qnx_spawn();
#endif /* QNX */

    erts_free(ERTS_ALC_T_TMP, (void *) cmd_line);

    if (new_environ != environ)
	erts_free(ERTS_ALC_T_ENVIRONMENT, (void *) new_environ);

    if (opts->read_write & DO_READ) 
	(void) close(ifd[1]);
    if (opts->read_write & DO_WRITE)
	(void) close(ofd[0]);
	
    if (opts->read_write & DO_READ) {
	SET_NONBLOCKING(ifd[0]);
	init_fd_data(ifd[0], port_num);
    }
    if (opts->read_write & DO_WRITE) {
	SET_NONBLOCKING(ofd[1]);
        init_fd_data(ofd[1], port_num);
    }

    res = set_driver_data(port_num, ifd[0], ofd[1], opts->packet_bytes,
			  opts->read_write, opts->exit_status, pid);
    /* Don't unblock SIGCHLD until now, since the call above must
       first complete putting away the info about our new subprocess. */
    unblock_signals();

#if CHLDWTHR
    ASSERT(children_alive >= 0);

    if (!(children_alive++))
	CHLD_STAT_SIGNAL; /* Wake up child waiter thread if no children
			     was alive before we fork()ed ... */
#endif
    /* Don't unlock chld_stat_mtx until now of the same reason as above */
    CHLD_STAT_UNLOCK;

    erts_smp_rwmtx_runlock(&environ_rwmtx);

    return (ErlDrvData)res;
#undef CMD_LINE_PREFIX_STR
#undef CMD_LINE_PREFIX_STR_SZ
}

#ifdef QNX
static reset_qnx_spawn()
{
    int i;

    /* Reset qnx_spawn_options */
    qnx_spawn_options.flags = 0; 
    qnx_spawn_options.iov[0] = 0xff;
    qnx_spawn_options.iov[1] = 0xff;
    qnx_spawn_options.iov[2] = 0xff;
    qnx_spawn_options.iov[3] = 0xff;
}
#endif

#define FD_DEF_HEIGHT 24
#define FD_DEF_WIDTH 80
/* Control op */
#define FD_CTRL_OP_GET_WINSIZE 100

static int fd_get_window_size(int fd, Uint32 *width, Uint32 *height)
{
#ifdef TIOCGWINSZ 
    struct winsize ws;
    if (ioctl(fd,TIOCGWINSZ,&ws) == 0) {
	*width = (Uint32) ws.ws_col;
	*height = (Uint32) ws.ws_row;
	return 0;
    }
#endif
    return -1;
}

static int fd_control(ErlDrvData drv_data,
		      unsigned int command,
		      char *buf, int len,
		      char **rbuf, int rlen)
{
    int fd = (int)(long)drv_data;
    char resbuff[2*sizeof(Uint32)];
    switch (command) {
    case FD_CTRL_OP_GET_WINSIZE:
	{
	    Uint32 w,h;
	    if (fd_get_window_size(fd,&w,&h)) 
		return 0;
	    memcpy(resbuff,&w,sizeof(Uint32));
	    memcpy(resbuff+sizeof(Uint32),&h,sizeof(Uint32));
	}
	break;
    default:
	return 0;
    }
    if (rlen < 2*sizeof(Uint32)) {
	*rbuf = driver_alloc(2*sizeof(Uint32));
    }
    memcpy(*rbuf,resbuff,2*sizeof(Uint32));
    return 2*sizeof(Uint32);
}

static ErlDrvData fd_start(ErlDrvPort port_num, char* name,
			   SysDriverOpts* opts)
{
    ErlDrvData res;

    if (((opts->read_write & DO_READ) && opts->ifd >= max_files) ||
	((opts->read_write & DO_WRITE) && opts->ofd >= max_files))
	return ERL_DRV_ERROR_GENERAL;

    /*
     * Historical:
     *
     * "Note about nonblocking I/O.
     *
     * At least on Solaris, setting the write end of a TTY to nonblocking,
     * will set the input end to nonblocking as well (and vice-versa).
     * If erl is run in a pipeline like this:  cat | erl
     * the input end of the TTY will be the standard input of cat.
     * And cat is not prepared to handle nonblocking I/O."
     *
     * Actually, the reason for this is not that the tty itself gets set
     * in non-blocking mode, but that the "input end" (cat's stdin) and
     * the "output end" (erlang's stdout) are typically the "same" file
     * descriptor, dup()'ed from a single fd by one of this process'
     * ancestors.
     *
     * The workaround for this problem used to be a rather bad kludge,
     * interposing an extra process ("internal cat") between erlang's
     * stdout and the original stdout, allowing erlang to set its stdout
     * in non-blocking mode without affecting the stdin of the preceding
     * process in the pipeline - and being a kludge, it caused all kinds
     * of weird problems.
     *
     * So, this is the current logic:
     *
     * The only reason to set non-blocking mode on the output fd at all is
     * if it's something that can cause a write() to block, of course,
     * i.e. primarily if it points to a tty, socket, pipe, or fifo. 
     *
     * If we don't set non-blocking mode when we "should" have, and output
     * becomes blocked, the entire runtime system will be suspended - this
     * is normally bad of course, and can happen fairly "easily" - e.g. user
     * hits ^S on tty - but doesn't necessarily happen.
     * 
     * If we do set non-blocking mode when we "shouldn't" have, the runtime
     * system will end up seeing EOF on the input fd (due to the preceding
     * process dying), which typically will cause the entire runtime system
     * to terminate immediately (due to whatever erlang process is seeing
     * the EOF taking it as a signal to halt the system). This is *very* bad.
     * 
     * I.e. we should take a conservative approach, and only set non-
     * blocking mode when we a) need to, and b) are reasonably certain
     * that it won't be a problem. And as in the example above, the problem
     * occurs when input fd and output fd point to different "things".
     *
     * However, determining that they are not just the same "type" of
     * "thing", but actually the same instance of that type of thing, is
     * unreasonably complex in many/most cases.
     *
     * Also, with pipes, sockets, and fifos it's far from obvious that the
     * user *wants* non-blocking output: If you're running erlang inside
     * some complex pipeline, you're probably not running a real-time system
     * that must never stop, but rather *want* it to suspend if the output
     * channel is "full".
     *
     * So, the bottom line: We will only set the output fd non-blocking if
     * it points to a tty, and either a) the input fd also points to a tty,
     * or b) we can make sure that setting the output fd non-blocking
     * doesn't interfere with someone else's input, via a somewhat milder
     * kludge than the above.
     *
     * Also keep in mind that while this code is almost exclusively run as
     * a result of an erlang open_port({fd,0,1}, ...), that isn't the only
     * case - it can be called with any old pre-existing file descriptors,
     * the relations between which (if they're even two) we can only guess
     * at - still, we try our best...
     */

    if (opts->read_write & DO_READ) {
	init_fd_data(opts->ifd, port_num);
    }
    if (opts->read_write & DO_WRITE) {
	init_fd_data(opts->ofd, port_num);

	/* If we don't have a read end, all bets are off - no non-blocking. */
	if (opts->read_write & DO_READ) {

	    if (isatty(opts->ofd)) { /* output fd is a tty:-) */

		if (isatty(opts->ifd)) { /* input fd is also a tty */

		    /* To really do this "right", we should also check that
		       input and output fd point to the *same* tty - but
		       this seems like overkill; ttyname() isn't for free,
		       and this is a very common case - and it's hard to
		       imagine a scenario where setting non-blocking mode
		       here would cause problems - go ahead and do it. */

		    SET_NONBLOCKING(opts->ofd);

		} else {	/* output fd is a tty, input fd isn't */

		    /* This is a "problem case", but also common (see the
		       example above) - i.e. it makes sense to try a bit
		       harder before giving up on non-blocking mode: Try to
		       re-open the tty that the output fd points to, and if
		       successful replace the original one with the "new" fd
		       obtained this way, and set *that* one in non-blocking
		       mode. (Yes, this is a kludge.)

		       However, re-opening the tty may fail in a couple of
		       (unusual) cases:

		       1) The name of the tty (or an equivalent one, i.e.
			  same major/minor number) can't be found, because
			  it actually lives somewhere other than /dev (or
			  wherever ttyname() looks for it), and isn't
			  equivalent to any of those that do live in the
			  "standard" place - this should be *very* unusual.

		       2) Permissions on the tty don't allow us to open it -
			  it's perfectly possible to have an fd open to an
			  object whose permissions wouldn't allow us to open
			  it. This is not as unusual as it sounds, one case
			  is if the user has su'ed to someone else (not
			  root) - we have a read/write fd open to the tty
			  (because it has been inherited all the way down
			  here), but we have neither read nor write
			  permission for the tty.

		       In these cases, we finally give up, and don't set the
		       output fd in non-blocking mode. */

		    char *tty;
		    int nfd;

		    if ((tty = ttyname(opts->ofd)) != NULL &&
			(nfd = open(tty, O_WRONLY)) != -1) {
			dup2(nfd, opts->ofd);
			close(nfd);
			SET_NONBLOCKING(opts->ofd);
		    }
		}
	    }
	}
    }
    CHLD_STAT_LOCK;
    res = (ErlDrvData)(long)set_driver_data(port_num, opts->ifd, opts->ofd,
				      opts->packet_bytes,
				      opts->read_write, 0, -1);
    CHLD_STAT_UNLOCK;
    return res;
}

static void clear_fd_data(int fd) 
{
    if (fd_data[fd].sz > 0) {
	erts_free(ERTS_ALC_T_FD_ENTRY_BUF, (void *) fd_data[fd].buf);
	ASSERT(erts_smp_atomic_read(&sys_misc_mem_sz) >= fd_data[fd].sz);
	erts_smp_atomic_add(&sys_misc_mem_sz, -1*fd_data[fd].sz);
    }
    fd_data[fd].buf = NULL;
    fd_data[fd].sz = 0;
    fd_data[fd].remain = 0;
    fd_data[fd].cpos = NULL;
    fd_data[fd].psz = 0;
}

static void nbio_stop_fd(int prt, int fd)
{
    driver_select(prt,fd,DO_READ|DO_WRITE,0);
    clear_fd_data(fd);
    SET_BLOCKING(fd);
}

static void fd_stop(ErlDrvData fd)  /* Does not close the fds */
{
    int ofd;
    
    nbio_stop_fd(driver_data[(int)(long)fd].port_num, (int)(long)fd);
    ofd = driver_data[(int)(long)fd].ofd;
    if (ofd != (int)(long)fd && ofd != -1) 
	nbio_stop_fd(driver_data[(int)(long)fd].port_num, (int)(long)ofd);
}

static ErlDrvData vanilla_start(ErlDrvPort port_num, char* name,
				SysDriverOpts* opts)
{
    int flags, fd;
    ErlDrvData res;

    flags = (opts->read_write == DO_READ ? O_RDONLY :
	     opts->read_write == DO_WRITE ? O_WRONLY|O_CREAT|O_TRUNC :
	     O_RDWR|O_CREAT);
    if ((fd = open(name, flags, 0666)) < 0)
	return ERL_DRV_ERROR_GENERAL;
    if (fd >= max_files) {
	close(fd);
	return ERL_DRV_ERROR_GENERAL;
    }
    SET_NONBLOCKING(fd);
    init_fd_data(fd, port_num);

    CHLD_STAT_LOCK;
    res = (ErlDrvData)(long)set_driver_data(port_num, fd, fd,
				      opts->packet_bytes,
				      opts->read_write, 0, -1);
    CHLD_STAT_UNLOCK;
    return res;
}

/* Note that driver_data[fd].ifd == fd if the port was opened for reading, */
/* otherwise (i.e. write only) driver_data[fd].ofd = fd.  */

static void stop(ErlDrvData fd)
{
    int prt, ofd;

    prt = driver_data[(int)(long)fd].port_num;
    nbio_stop_fd(prt, (int)(long)fd);

    ofd = driver_data[(int)(long)fd].ofd;
    if (ofd != (int)(long)fd && (int)(long)ofd != -1)
	nbio_stop_fd(prt, ofd);
    else
	ofd = -1;

    CHLD_STAT_LOCK;

    /* Mark as unused. Maybe resetting the 'port_num' slot is better? */
    driver_data[(int)(long)fd].pid = -1;

    CHLD_STAT_UNLOCK;

    /* SMP note: Close has to be last thing done (open file descriptors work
       as locks on driver_data[] entries) */
    driver_select(prt, (int)(long)fd, ERL_DRV_USE, 0);  /* close(fd); */
    if (ofd >= 0) {
	driver_select(prt, (int)(long)ofd, ERL_DRV_USE, 0);  /* close(ofd); */
    }
}

static void outputv(ErlDrvData e, ErlIOVec* ev)
{
    int fd = (int)(long)e;
    int ix = driver_data[fd].port_num;
    int pb = driver_data[fd].packet_bytes;
    int ofd = driver_data[fd].ofd;
    int n;
    int sz;
    char lb[4];
    char* lbp;
    int len = ev->size;

    /* (len > ((unsigned long)-1 >> (4-pb)*8)) */
    if (((pb == 2) && (len > 0xffff)) || (pb == 1 && len > 0xff)) {
	driver_failure_posix(ix, EINVAL);
	return; /* -1; */
    }
    put_int32(len, lb);
    lbp = lb + (4-pb);

    ev->iov[0].iov_base = lbp;
    ev->iov[0].iov_len = pb;
    ev->size += pb;
    if ((sz = driver_sizeq(ix)) > 0) {
	driver_enqv(ix, ev, 0);
	if (sz + ev->size >= (1 << 13))
	    set_busy_port(ix, 1);
    }
    else {
	int vsize = ev->vsize > MAX_VSIZE ? MAX_VSIZE : ev->vsize;

	n = writev(ofd, (const void *) (ev->iov), vsize);
	if (n == ev->size)
	    return; /* 0;*/
	if (n < 0) {
	    if ((errno != EINTR) && (errno != ERRNO_BLOCK)) {
		driver_failure_posix(ix, errno);
		return; /* -1;*/
	    }
	    n = 0;
	}
	driver_enqv(ix, ev, n);  /* n is the skip value */
	driver_select(ix, ofd, ERL_DRV_WRITE|ERL_DRV_USE, 1);
    }
    /* return 0;*/
}


static void output(ErlDrvData e, char* buf, int len)
{
    int fd = (int)(long)e;
    int ix = driver_data[fd].port_num;
    int pb = driver_data[fd].packet_bytes;
    int ofd = driver_data[fd].ofd;
    int n;
    int sz;
    char lb[4];
    char* lbp;
    struct iovec iv[2];

    /* (len > ((unsigned long)-1 >> (4-pb)*8)) */
    if (((pb == 2) && (len > 0xffff)) || (pb == 1 && len > 0xff)) {
	driver_failure_posix(ix, EINVAL);
	return; /* -1; */
    }
    put_int32(len, lb);
    lbp = lb + (4-pb);

    if ((sz = driver_sizeq(ix)) > 0) {
	driver_enq(ix, lbp, pb);
	driver_enq(ix, buf, len);
	if (sz + len + pb >= (1 << 13))
	    set_busy_port(ix, 1);
    }
    else {
	iv[0].iov_base = lbp;
	iv[0].iov_len = pb;  /* should work for pb=0 */
	iv[1].iov_base = buf;
	iv[1].iov_len = len;
	n = writev(ofd, iv, 2);
	if (n == pb+len)
	    return; /* 0; */
	if (n < 0) {
	    if ((errno != EINTR) && (errno != ERRNO_BLOCK)) {
		driver_failure_posix(ix, errno);
		return; /* -1; */
	    }
	    n = 0;
	}
	if (n < pb) {
	    driver_enq(ix, lbp+n, pb-n);
	    driver_enq(ix, buf, len);
	}
	else {
	    n -= pb;
	    driver_enq(ix, buf+n, len-n);
	}
	driver_select(ix, ofd, ERL_DRV_WRITE|ERL_DRV_USE, 1);
    }
    return; /* 0; */
}

static int port_inp_failure(int port_num, int ready_fd, int res)
				/* Result: 0 (eof) or -1 (error) */
{
    int err = errno;

    ASSERT(res <= 0);
    (void) driver_select(port_num, ready_fd, ERL_DRV_READ|ERL_DRV_WRITE, 0); 
    clear_fd_data(ready_fd);
    if (res == 0) {
	if (driver_data[ready_fd].report_exit) {
	    CHLD_STAT_LOCK;

	    if (driver_data[ready_fd].alive) {
		/*
		 * We have eof and want to report exit status, but the process
		 * hasn't exited yet. When it does report_exit_status() will
		 * driver_select() this fd which will make sure that we get
		 * back here with driver_data[ready_fd].alive == 0 and
		 * driver_data[ready_fd].status set.
		 */
		CHLD_STAT_UNLOCK;
		return 0;
	    }
	    else {
		int status = driver_data[ready_fd].status;
		CHLD_STAT_UNLOCK;

		/* We need not be prepared for stopped/continued processes. */
		if (WIFSIGNALED(status))
		    status = 128 + WTERMSIG(status);
		else
		    status = WEXITSTATUS(status);

		driver_report_exit(driver_data[ready_fd].port_num, status);
	    }
       }
       driver_failure_eof(port_num);
    } else {
	driver_failure_posix(port_num, err);
    }
    return 0;
}

/* fd is the drv_data that is returned from the */
/* initial start routine                        */
/* ready_fd is the descriptor that is ready to read */

static void ready_input(ErlDrvData e, ErlDrvEvent ready_fd)
{
    int fd = (int)(long)e;
    int port_num;
    int packet_bytes;
    int res;
    Uint h;

    port_num = driver_data[fd].port_num;
    packet_bytes = driver_data[fd].packet_bytes;

    if (packet_bytes == 0) {
	byte *read_buf = (byte *) erts_alloc(ERTS_ALC_T_SYS_READ_BUF,
					     ERTS_SYS_READ_BUF_SZ);
	res = read(ready_fd, read_buf, ERTS_SYS_READ_BUF_SZ);
	if (res < 0) {
	    if ((errno != EINTR) && (errno != ERRNO_BLOCK))
		port_inp_failure(port_num, ready_fd, res);
	}
	else if (res == 0)
	    port_inp_failure(port_num, ready_fd, res);
	else 
	    driver_output(port_num, (char*) read_buf, res);
	erts_free(ERTS_ALC_T_SYS_READ_BUF, (void *) read_buf);
    }
    else if (fd_data[ready_fd].remain > 0) { /* We try to read the remainder */
	/* space is allocated in buf */
	res = read(ready_fd, fd_data[ready_fd].cpos, 
		   fd_data[ready_fd].remain);
	if (res < 0) {
	    if ((errno != EINTR) && (errno != ERRNO_BLOCK))
		port_inp_failure(port_num, ready_fd, res);
	}
	else if (res == 0) {
	    port_inp_failure(port_num, ready_fd, res);
	}
	else if (res == fd_data[ready_fd].remain) { /* we're done  */
	    driver_output(port_num, fd_data[ready_fd].buf, 
			  fd_data[ready_fd].sz);
	    clear_fd_data(ready_fd);
	}
	else { /*  if (res < fd_data[ready_fd].remain) */
	    fd_data[ready_fd].cpos += res;
	    fd_data[ready_fd].remain -= res;
	}
    }
    else if (fd_data[ready_fd].remain == 0) { /* clean fd */
	byte *read_buf = (byte *) erts_alloc(ERTS_ALC_T_SYS_READ_BUF,
					     ERTS_SYS_READ_BUF_SZ);
	/* We make one read attempt and see what happens */
	res = read(ready_fd, read_buf, ERTS_SYS_READ_BUF_SZ);
	if (res < 0) {  
	    if ((errno != EINTR) && (errno != ERRNO_BLOCK))
		port_inp_failure(port_num, ready_fd, res);
	}
	else if (res == 0) {     	/* eof */
	    port_inp_failure(port_num, ready_fd, res);
	} 
	else if (res < packet_bytes - fd_data[ready_fd].psz) { 
	    memcpy(fd_data[ready_fd].pbuf+fd_data[ready_fd].psz,
		   read_buf, res);
	    fd_data[ready_fd].psz += res;
	}
	else  { /* if (res >= packet_bytes) */
	    unsigned char* cpos = read_buf;
	    int bytes_left = res;

	    while (1) {
		int psz = fd_data[ready_fd].psz;
		char* pbp = fd_data[ready_fd].pbuf + psz;

		while(bytes_left && (psz < packet_bytes)) {
		    *pbp++ = *cpos++;
		    bytes_left--;
		    psz++;
		}
		
		if (psz < packet_bytes) {
		    fd_data[ready_fd].psz = psz;
		    break;
		}
		fd_data[ready_fd].psz = 0;

		switch (packet_bytes) {
		case 1: h = get_int8(fd_data[ready_fd].pbuf);  break;
		case 2: h = get_int16(fd_data[ready_fd].pbuf); break;
		case 4: h = get_int32(fd_data[ready_fd].pbuf); break;
		default: ASSERT(0); return; /* -1; */
		}

		if (h <= (bytes_left)) {
		    driver_output(port_num, (char*) cpos, h);
		    cpos += h;
		    bytes_left -= h;
		    continue;
		}
		else {		/* The last message we got was split */
		        char *buf = erts_alloc_fnf(ERTS_ALC_T_FD_ENTRY_BUF, h);
		    if (!buf) {
			errno = ENOMEM;
			port_inp_failure(port_num, ready_fd, -1);
		    }
		    else {
			erts_smp_atomic_add(&sys_misc_mem_sz, h);
			sys_memcpy(buf, cpos, bytes_left);
			fd_data[ready_fd].buf = buf;
			fd_data[ready_fd].sz = h;
			fd_data[ready_fd].remain = h - bytes_left;
			fd_data[ready_fd].cpos = buf + bytes_left;
		    }
		    break;
		}
	    }
	}
	erts_free(ERTS_ALC_T_SYS_READ_BUF, (void *) read_buf);
    }
}


/* fd is the drv_data that is returned from the */
/* initial start routine                        */
/* ready_fd is the descriptor that is ready to read */

static void ready_output(ErlDrvData e, ErlDrvEvent ready_fd)
{
    int fd = (int)(long)e;
    int ix = driver_data[fd].port_num;
    int n;
    struct iovec* iv;
    int vsize;
    

    if ((iv = (struct iovec*) driver_peekq(ix, &vsize)) == NULL) {
	driver_select(ix, ready_fd, ERL_DRV_WRITE, 0);
	return; /* 0; */
    }
    vsize = vsize > MAX_VSIZE ? MAX_VSIZE : vsize;
    if ((n = writev(ready_fd, iv, vsize)) > 0) {
	if (driver_deq(ix, n) == 0)
	    set_busy_port(ix, 0);
    }
    else if (n < 0) {
	if (errno == ERRNO_BLOCK || errno == EINTR)
	    return; /* 0; */
	else {
	    int res = errno;
	    driver_select(ix, ready_fd, ERL_DRV_WRITE, 0);
	    driver_failure_posix(ix, res);
	    return; /* -1; */
	}
    }
    return; /* 0; */
}

static void stop_select(ErlDrvEvent fd, void* _)
{
    close((int)fd);
}

/*
** Async opertation support
*/
#if defined(USE_THREADS) && !defined(ERTS_SMP)
static void
sys_async_ready_failed(int fd, int r, int err)
{
    char buf[120];
    sprintf(buf, "sys_async_ready(): Fatal error: fd=%d, r=%d, errno=%d\n",
	     fd, r, err);
    erts_silence_warn_unused_result(write(2, buf, strlen(buf)));
    abort();
}

/* called from threads !! */
void sys_async_ready(int fd)
{
    int r;
    while (1) {
	r = write(fd, "0", 1);  /* signal main thread fd MUST be async_fd[1] */
	if (r == 1) {
	    DEBUGF(("sys_async_ready(): r = 1\r\n"));
	    break;
	}
	if (r < 0 && errno == EINTR) {
	    DEBUGF(("sys_async_ready(): r = %d\r\n", r));
	    continue;
	}
	sys_async_ready_failed(fd, r, errno);
    }
}

static int async_drv_init(void)
{
    async_fd[0] = -1;
    async_fd[1] = -1;
    return 0;
}

static ErlDrvData async_drv_start(ErlDrvPort port_num,
				  char* name, SysDriverOpts* opts)
{
    if (async_fd[0] != -1)
	return ERL_DRV_ERROR_GENERAL;
    if (pipe(async_fd) < 0)
	return ERL_DRV_ERROR_GENERAL;

    DEBUGF(("async_drv_start: %d\r\n", port_num));

    SET_NONBLOCKING(async_fd[0]);
    driver_select(port_num, async_fd[0], ERL_DRV_READ, 1);

    if (init_async(async_fd[1]) < 0)
	return ERL_DRV_ERROR_GENERAL;
    return (ErlDrvData)port_num;
}

static void async_drv_stop(ErlDrvData e)
{
    int port_num = (int)(long)e;

    DEBUGF(("async_drv_stop: %d\r\n", port_num));

    exit_async();

    driver_select(port_num, async_fd[0], ERL_DRV_READ, 0);

    close(async_fd[0]);
    close(async_fd[1]);
    async_fd[0] = async_fd[1] = -1;
}


static void async_drv_input(ErlDrvData e, ErlDrvEvent fd)
{
    char *buf[32];
    DEBUGF(("async_drv_input\r\n"));
    while (read((int) fd, (void *) buf, 32) > 0); /* fd MUST be async_fd[0] */
    check_async_ready();  /* invoke all async_ready */
}
#endif

void erts_do_break_handling(void)
{
    struct termios temp_mode;
    int saved = 0;
    
    /*
     * Most functions that do_break() calls are intentionally not thread safe;
     * therefore, make sure that all threads but this one are blocked before
     * proceeding!
     */
    erts_smp_block_system(0);
    /*
     * NOTE: since we allow gc we are not allowed to lock
     *       (any) process main locks while blocking system...
     */

    /* during break we revert to initial settings */
    /* this is done differently for oldshell */
    if (using_oldshell && !replace_intr) {
      SET_BLOCKING(1);
    }
    else if (isatty(0)) {
      tcgetattr(0,&temp_mode);
      tcsetattr(0,TCSANOW,&initial_tty_mode);
      saved = 1;
    }
    
    /* call the break handling function, reset the flag */
    do_break();

    ERTS_UNSET_BREAK_REQUESTED;

    fflush(stdout);
    
    /* after break we go back to saved settings */
    if (using_oldshell && !replace_intr) {
      SET_NONBLOCKING(1);
    }
    else if (saved) {
      tcsetattr(0,TCSANOW,&temp_mode);
    }

    erts_smp_release_system();
}

/* Fills in the systems representation of the jam/beam process identifier.
** The Pid is put in STRING representation in the supplied buffer,
** no interpretatione of this should be done by the rest of the
** emulator. The buffer should be at least 21 bytes long.
*/
void sys_get_pid(char *buffer){
    pid_t p = getpid();
    /* Assume the pid is scalar and can rest in an unsigned long... */
    sprintf(buffer,"%lu",(unsigned long) p);
}

int
erts_sys_putenv(char *buffer, int sep_ix)
{
    int res;
    char *env;
#ifdef HAVE_COPYING_PUTENV
    env = buffer;
#else
    Uint sz = strlen(buffer)+1;
    env = erts_alloc(ERTS_ALC_T_PUTENV_STR, sz);
    erts_smp_atomic_add(&sys_misc_mem_sz, sz);
    strcpy(env,buffer);
#endif
    erts_smp_rwmtx_rwlock(&environ_rwmtx);
    res = putenv(env);
    erts_smp_rwmtx_rwunlock(&environ_rwmtx);
    return res;
}

int
erts_sys_getenv(char *key, char *value, size_t *size)
{
    char *orig_value;
    int res;
    erts_smp_rwmtx_rlock(&environ_rwmtx);
    orig_value = getenv(key);
    if (!orig_value)
	res = -1;
    else {
	size_t len = sys_strlen(orig_value);
	if (len >= *size) {
	    *size = len + 1;
	    res = 1;
	}
	else {
	    *size = len;
	    sys_memcpy((void *) value, (void *) orig_value, len+1);
	    res = 0;
	}
    }
    erts_smp_rwmtx_runlock(&environ_rwmtx);
    return res;
}

void
sys_init_io(void)
{
    fd_data = (struct fd_data *)
	erts_alloc(ERTS_ALC_T_FD_TAB, max_files * sizeof(struct fd_data));
    erts_smp_atomic_add(&sys_misc_mem_sz,
			max_files * sizeof(struct fd_data));

#ifdef USE_THREADS
#ifdef ERTS_SMP
    if (init_async(-1) < 0)
	erl_exit(1, "Failed to initialize async-threads\n");
#else
    {
	/* This is speical stuff, starting a driver from the 
	 * system routines, but is a nice way of handling stuff
	 * the erlang way
	 */
	SysDriverOpts dopts;
	int ret;

	sys_memset((void*)&dopts, 0, sizeof(SysDriverOpts));
	add_driver_entry(&async_driver_entry);
	ret = erts_open_driver(NULL, NIL, "async", &dopts, NULL);
	DEBUGF(("open_driver = %d\n", ret));
	if (ret < 0)
	    erl_exit(1, "Failed to open async driver\n");
	erts_port[ret].status |= ERTS_PORT_SFLG_IMMORTAL;
    }
#endif
#endif

}

#if (0) /* unused? */
static int write_fill(fd, buf, len)
int fd, len;
char *buf;
{
    int i, done = 0;
    
    do {
	if ((i = write(fd, buf+done, len-done)) < 0) {
	    if (errno != EINTR)
		return (i);
	    i = 0;
	}
	done += i;
    } while (done < len);
    return (len);
}
#endif

extern const char pre_loaded_code[];
extern Preload pre_loaded[];

void erts_sys_alloc_init(void)
{
    elib_ensure_initialized();
}

void *erts_sys_alloc(ErtsAlcType_t t, void *x, Uint sz)
{
    void *res = malloc((size_t) sz);
#if HAVE_ERTS_MSEG
    if (!res) {
	erts_mseg_clear_cache();
	return malloc((size_t) sz);
    }
#endif
    return res;
}

void *erts_sys_realloc(ErtsAlcType_t t, void *x, void *p, Uint sz)
{
    void *res = realloc(p, (size_t) sz);
#if HAVE_ERTS_MSEG
    if (!res) {
	erts_mseg_clear_cache();
	return realloc(p, (size_t) sz);
    }
#endif
    return res;
}

void erts_sys_free(ErtsAlcType_t t, void *x, void *p)
{
    free(p);
}

/* Return a pointer to a vector of names of preloaded modules */

Preload*
sys_preloaded(void)
{
    return pre_loaded;
}

/* Return a pointer to preloaded code for module "module" */
unsigned char*
sys_preload_begin(Preload* p)
{
    return p->code;
}

/* Clean up if allocated */
void sys_preload_end(Preload* p)
{
    /* Nothing */
}

/* Read a key from console (?) */

int sys_get_key(fd)
int fd;
{
    int c;
    unsigned char rbuf[64];

    fflush(stdout);		/* Flush query ??? */

    if ((c = read(fd,rbuf,64)) <= 0) {
      return c; 
    }

    return rbuf[0]; 
}


#ifdef DEBUG

extern int erts_initialized;
void
erl_assert_error(char* expr, char* file, int line)
{   
    fflush(stdout);
    fprintf(stderr, "Assertion failed: %s in %s, line %d\n",
	    expr, file, line);
    fflush(stderr);
#if !defined(ERTS_SMP) && 0
    /* Writing a crashdump from a failed assertion when smp support
     * is enabled almost a guaranteed deadlocking, don't even bother.
     *
     * It could maybe be useful (but I'm not convinced) to write the
     * crashdump if smp support is disabled...
     */
    if (erts_initialized)
	erl_crash_dump(file, line, "Assertion failed: %s\n", expr);
#endif
    abort();
}

void
erl_debug(char* fmt, ...)
{
    char sbuf[1024];		/* Temporary buffer. */
    va_list va;
    
    if (debug_log) {
	va_start(va, fmt);
	vsprintf(sbuf, fmt, va);
	va_end(va);
	fprintf(stderr, "%s", sbuf);
    }
}

#endif /* DEBUG */

static ERTS_INLINE void
report_exit_status(ErtsSysReportExit *rep, int status)
{
    Port *pp;
#ifdef ERTS_SMP
    CHLD_STAT_UNLOCK;
#endif
    pp = erts_id2port_sflgs(rep->port,
			    NULL,
			    0,
			    ERTS_PORT_SFLGS_INVALID_DRIVER_LOOKUP);
#ifdef ERTS_SMP
    CHLD_STAT_LOCK;
#endif
    if (pp) {
	if (rep->ifd >= 0) {
	    driver_data[rep->ifd].alive = 0;
	    driver_data[rep->ifd].status = status;
	    (void) driver_select((ErlDrvPort) internal_port_index(pp->id),
				 rep->ifd,
				 (ERL_DRV_READ|ERL_DRV_USE),
				 1);
	}
	if (rep->ofd >= 0) {
	    driver_data[rep->ofd].alive = 0;
	    driver_data[rep->ofd].status = status;
	    (void) driver_select((ErlDrvPort) internal_port_index(pp->id),
				 rep->ofd,
				 (ERL_DRV_WRITE|ERL_DRV_USE),
				 1);
	}
	erts_port_release(pp);
    }
    erts_free(ERTS_ALC_T_PRT_REP_EXIT, rep);
}

#if !CHLDWTHR  /* ---------------------------------------------------------- */

#define ERTS_REPORT_EXIT_STATUS report_exit_status

static int check_children(void)
{
    int res = 0;
    int pid;
    int status;

#ifndef ERTS_SMP
    if (children_died)
#endif
    {
	sys_sigblock(SIGCHLD);
	CHLD_STAT_LOCK;
	while ((pid = waitpid(-1, &status, WNOHANG)) > 0)
	    note_child_death(pid, status);
#ifndef ERTS_SMP
	children_died = 0;
#endif
	CHLD_STAT_UNLOCK;
	sys_sigrelease(SIGCHLD);
	res = 1;
    }
    return res;
}

#ifdef ERTS_SMP

void
erts_check_children(void)
{
    (void) check_children();
}

#endif

#elif CHLDWTHR && defined(ERTS_SMP) /* ------------------------------------- */

#define ERTS_REPORT_EXIT_STATUS report_exit_status

#define check_children() (0)


#else /* CHLDWTHR && !defined(ERTS_SMP) ------------------------------------ */

#define ERTS_REPORT_EXIT_STATUS initiate_report_exit_status

static ERTS_INLINE void
initiate_report_exit_status(ErtsSysReportExit *rep, int status)
{
    rep->next = report_exit_transit_list;
    rep->status = status;
    report_exit_transit_list = rep;
    /*
     * We need the scheduler thread to call check_children().
     * If the scheduler thread is sleeping in a poll with a
     * timeout, we need to wake the scheduler thread. We use the
     * functionality of the async driver to do this, instead of
     * implementing yet another driver doing the same thing. A
     * little bit ugly, but it works...
     */
    sys_async_ready(async_fd[1]);
}

static int check_children(void)
{
    int res;
    ErtsSysReportExit *rep;
    CHLD_STAT_LOCK;
    rep = report_exit_transit_list;
    res = rep != NULL;
    while (rep) {
	ErtsSysReportExit *curr_rep = rep;
	rep = rep->next;
	report_exit_status(curr_rep, curr_rep->status);
    }
    report_exit_transit_list = NULL;
    CHLD_STAT_UNLOCK;
    return res;
}

#endif /* ------------------------------------------------------------------ */

static void note_child_death(int pid, int status)
{
    ErtsSysReportExit **repp = &report_exit_list;
    ErtsSysReportExit *rep = report_exit_list;

    while (rep) {
	if (pid == rep->pid) {
	    *repp = rep->next;
	    ERTS_REPORT_EXIT_STATUS(rep, status);
	    break;
	}
	repp = &rep->next;
	rep = rep->next;
    }
}

#if CHLDWTHR

static void *
child_waiter(void *unused)
{
  int pid;
  int status;

#ifdef ERTS_ENABLE_LOCK_CHECK
  erts_lc_set_thread_name("child waiter");
#endif

  while(1) {
#ifdef DEBUG
      int waitpid_errno;
#endif
      pid = waitpid(-1, &status, 0);
#ifdef DEBUG
      waitpid_errno = errno;
#endif
      CHLD_STAT_LOCK;
      if (pid < 0) {
	  ASSERT(waitpid_errno == ECHILD);
      }
      else {
	  children_alive--;
	  ASSERT(children_alive >= 0);
	  note_child_death(pid, status);
      }
      while (!children_alive)
	  CHLD_STAT_WAIT; /* Wait for children to wait on... :) */
      CHLD_STAT_UNLOCK;
  }

  return NULL;
}

#endif

/*
 * Called from schedule() when it runs out of runnable processes,
 * or when Erlang code has performed INPUT_REDUCTIONS reduction
 * steps. runnable == 0 iff there are no runnable Erlang processes.
 */
void
erl_sys_schedule(int runnable)
{
#ifdef ERTS_SMP
    ERTS_CHK_IO(!runnable);
    ERTS_SMP_LC_ASSERT(!ERTS_LC_IS_BLOCKING);
#else
    ERTS_CHK_IO_INTR(0);
    if (runnable) {
	ERTS_CHK_IO(0);		/* Poll for I/O */
	check_async_ready();	/* Check async completions */
    } else {
	int wait_for_io = !check_async_ready();
	if (wait_for_io)
	    wait_for_io = !check_children();
	ERTS_CHK_IO(wait_for_io);
    }
    (void) check_children();
#endif
}


#ifdef ERTS_SMP

static erts_smp_tid_t sig_dispatcher_tid;

static void
smp_sig_notify(char c)
{
    int res;
    do {
	/* write() is async-signal safe (according to posix) */
	res = write(sig_notify_fds[1], &c, 1);
    } while (res < 0 && errno == EINTR);
    if (res != 1) {
	char msg[] =
	    "smp_sig_notify(): Failed to notify signal-dispatcher thread "
	    "about received signal";
	erts_silence_warn_unused_result(write(2, msg, sizeof(msg)));
	abort();
    }
}

static void *
signal_dispatcher_thread_func(void *unused)
{
    int initialized = 0;
#if !CHLDWTHR
    int notify_check_children = 0;
#endif
#ifdef ERTS_ENABLE_LOCK_CHECK
    erts_lc_set_thread_name("signal_dispatcher");
#endif
    while (1) {
	char buf[32];
	int res, i;
	/* Block on read() waiting for a signal notification to arrive... */
	res = read(sig_notify_fds[0], (void *) &buf[0], 32);
	if (res < 0) {
	    if (errno == EINTR)
		continue;
	    erl_exit(ERTS_ABORT_EXIT,
		     "signal-dispatcher thread got unexpected error: %s (%d)\n",
		     erl_errno_id(errno),
		     errno);
	}
	for (i = 0; i < res; i++) {
	    /*
	     * NOTE 1: The signal dispatcher thread should not do work
	     *         that takes a substantial amount of time (except
	     *         perhaps in test and debug builds). It needs to
	     *         be responsive, i.e, it should only dispatch work
	     *         to other threads.
	     *
	     * NOTE 2: The signal dispatcher thread is not a blockable
	     *         thread (i.e., it hasn't called 
	     *         erts_register_blockable_thread()). This is
	     *         intentional. We want to be able to interrupt
	     *         writing of a crash dump by hitting C-c twice.
	     *         Since it isn't a blockable thread it is important
	     *         that it doesn't change the state of any data that
	     *         a blocking thread expects to have exclusive access
	     *         to (unless the signal dispatcher itself explicitly
	     *         is blocking all blockable threads).
	     */
	    switch (buf[i]) {
	    case 0: /* Emulator initialized */
		initialized = 1;
#if !CHLDWTHR
		if (!notify_check_children)
#endif
		    break;
#if !CHLDWTHR
	    case 'C': /* SIGCHLD */
		if (initialized)
		    erts_smp_notify_check_children_needed();
		else
		    notify_check_children = 1;
		break;
#endif
	    case 'I': /* SIGINT */
		break_requested();
		break;
	    case 'Q': /* SIGQUIT */
		quit_requested();
		break;
	    case '1': /* SIGUSR1 */
		sigusr1_exit();
		break;
#ifdef QUANTIFY
	    case '2': /* SIGUSR2 */
		quantify_save_data(); /* Might take a substantial amount of
					 time, but this is a test/debug
					 build */
		break;
#endif
	    default:
		erl_exit(ERTS_ABORT_EXIT,
			 "signal-dispatcher thread received unknown "
			 "signal notification: '%c'\n",
			 buf[i]);
	    }
	}
	ERTS_SMP_LC_ASSERT(!ERTS_LC_IS_BLOCKING);
    }
    return NULL;
}

static void
init_smp_sig_notify(void)
{
    erts_smp_thr_opts_t thr_opts = ERTS_SMP_THR_OPTS_DEFAULT_INITER;
    thr_opts.detached = 1;

    if (pipe(sig_notify_fds) < 0) {
	erl_exit(ERTS_ABORT_EXIT,
		 "Failed to create signal-dispatcher pipe: %s (%d)\n",
		 erl_errno_id(errno),
		 errno);
    }

    /* Start signal handler thread */
    erts_smp_thr_create(&sig_dispatcher_tid,
			signal_dispatcher_thread_func,
			NULL,
			&thr_opts);
}

void
erts_sys_main_thread(void)
{
    erts_thread_disable_fpe();
    /* Become signal receiver thread... */
#ifdef ERTS_ENABLE_LOCK_CHECK
    erts_lc_set_thread_name("signal_receiver");
#endif

    smp_sig_notify(0); /* Notify initialized */
    while (1) {
	/* Wait for a signal to arrive... */
#ifdef DEBUG
	int res =
#else
	(void)
#endif
	    select(0, NULL, NULL, NULL, NULL);
	ASSERT(res < 0);
	ASSERT(errno == EINTR);
    }
}

#endif /* ERTS_SMP */

#ifdef ERTS_ENABLE_KERNEL_POLL /* get_value() is currently only used when
				  kernel-poll is enabled */

/* Get arg marks argument as handled by
   putting NULL in argv */
static char *
get_value(char* rest, char** argv, int* ip)
{
    char *param = argv[*ip]+1;
    argv[*ip] = NULL;
    if (*rest == '\0') {
	char *next = argv[*ip + 1];
	if (next[0] == '-'
	    && next[1] == '-'
	    &&  next[2] == '\0') {
	    erts_fprintf(stderr, "bad \"%s\" value: \n", param);
	    erts_usage();
	}
	(*ip)++;
	argv[*ip] = NULL;
	return next;
    }
    return rest;
}

#endif /* ERTS_ENABLE_KERNEL_POLL */

void
erl_sys_args(int* argc, char** argv)
{
    int i, j;

    i = 1;

    ASSERT(argc && argv);

    while (i < *argc) {
	if(argv[i][0] == '-') {
	    switch (argv[i][1]) {
#ifdef ERTS_ENABLE_KERNEL_POLL
	    case 'K': {
		char *arg = get_value(argv[i] + 2, argv, &i);
		if (strcmp("true", arg) == 0) {
		    erts_use_kernel_poll = 1;
		}
		else if (strcmp("false", arg) == 0) {
		    erts_use_kernel_poll = 0;
		}
		else {
		    erts_fprintf(stderr, "bad \"K\" value: %s\n", arg);
		    erts_usage();
		}
		break;
	    }
#endif
	    case '-':
		goto done_parsing;
	    default:
		break;
	    }
	}
	i++;
    }

 done_parsing:

#ifdef ERTS_ENABLE_KERNEL_POLL
    if (erts_use_kernel_poll) {
	char no_kp[10];
	size_t no_kp_sz = sizeof(no_kp);
	int res = erts_sys_getenv("ERL_NO_KERNEL_POLL", no_kp, &no_kp_sz);
	if (res > 0
	    || (res == 0
		&& sys_strcmp("false", no_kp) != 0
		&& sys_strcmp("FALSE", no_kp) != 0)) {
	    erts_use_kernel_poll = 0;
	}
    }
#endif

    init_check_io();

#ifdef ERTS_SMP
    init_smp_sig_notify();
#endif

    /* Handled arguments have been marked with NULL. Slide arguments
       not handled towards the beginning of argv. */
    for (i = 0, j = 0; i < *argc; i++) {
	if (argv[i])
	    argv[j++] = argv[i];
    }
    *argc = j;
}

#ifdef ERTS_TIMER_THREAD

/*
 * Interruptible-wait facility: low-level synchronisation state
 * and methods that are implementation dependent.
 *
 * Constraint: Every implementation must define 'struct erts_iwait'
 * with a field 'erts_smp_atomic_t state;'.
 */

/* values for struct erts_iwait's state field */
#define IWAIT_WAITING	0
#define IWAIT_AWAKE	1
#define IWAIT_INTERRUPT	2

#if 0	/* XXX: needs feature test in erts/configure.in */

/*
 * This is an implementation of the interruptible wait facility on
 * top of Linux-specific futexes.
 */
#include <asm/unistd.h>
#define FUTEX_WAIT		0
#define FUTEX_WAKE		1
static int sys_futex(void *futex, int op, int val, const struct timespec *timeout)
{
    return syscall(__NR_futex, futex, op, val, timeout);
}

struct erts_iwait {
    erts_smp_atomic_t state; /* &state.counter is our futex */
};

static void iwait_lowlevel_init(struct erts_iwait *iwait) { /* empty */ }

static void iwait_lowlevel_wait(struct erts_iwait *iwait, struct timeval *delay)
{
    struct timespec timeout;
    int res;

    timeout.tv_sec = delay->tv_sec;
    timeout.tv_nsec = delay->tv_usec * 1000;
    res = sys_futex((void*)&iwait->state.counter, FUTEX_WAIT, IWAIT_WAITING, &timeout);
    if (res < 0 && errno != ETIMEDOUT && errno != EWOULDBLOCK && errno != EINTR)
	perror("FUTEX_WAIT");
}

static void iwait_lowlevel_interrupt(struct erts_iwait *iwait)
{
    int res = sys_futex((void*)&iwait->state.counter, FUTEX_WAKE, 1, NULL);
    if (res < 0)
	perror("FUTEX_WAKE");
}

#else	/* using poll() or select() */

/*
 * This is an implementation of the interruptible wait facility on
 * top of pipe(), poll() or select(), read(), and write().
 */
struct erts_iwait {
    erts_smp_atomic_t state;
    int read_fd;	/* wait polls and reads this fd */
    int write_fd;	/* interrupt writes this fd */
};

static void iwait_lowlevel_init(struct erts_iwait *iwait)
{
    int fds[2];

    if (pipe(fds) < 0) {
	perror("pipe()");
	exit(1);
    }
    iwait->read_fd = fds[0];
    iwait->write_fd = fds[1];
}

#if defined(ERTS_USE_POLL)

#include <sys/poll.h>
#define PERROR_POLL "poll()"

static int iwait_lowlevel_poll(int read_fd, struct timeval *delay)
{
    struct pollfd pollfd;
    int timeout;

    pollfd.fd = read_fd;
    pollfd.events = POLLIN;
    pollfd.revents = 0;
    timeout = delay->tv_sec * 1000 + delay->tv_usec / 1000;
    return poll(&pollfd, 1, timeout);
}

#else	/* !ERTS_USE_POLL */

#include <sys/select.h>
#define PERROR_POLL "select()"

static int iwait_lowlevel_poll(int read_fd, struct timeval *delay)
{
    fd_set readfds;

    FD_ZERO(&readfds);
    FD_SET(read_fd, &readfds);
    return select(read_fd + 1, &readfds, NULL, NULL, delay);
}

#endif	/* !ERTS_USE_POLL */

static void iwait_lowlevel_wait(struct erts_iwait *iwait, struct timeval *delay)
{
    int res;
    char buf[64];

    res = iwait_lowlevel_poll(iwait->read_fd, delay);
    if (res > 0)
	(void)read(iwait->read_fd, buf, sizeof buf);
    else if (res < 0 && errno != EINTR)
	perror(PERROR_POLL);
}

static void iwait_lowlevel_interrupt(struct erts_iwait *iwait)
{
    int res = write(iwait->write_fd, "!", 1);
    if (res < 0)
	perror("write()");
}

#endif	/* using poll() or select() */

#if 0	/* not using poll() or select() */
/*
 * This is an implementation of the interruptible wait facility on
 * top of pthread_cond_timedwait(). This has two problems:
 * 1. pthread_cond_timedwait() requires an absolute time point,
 *    so the relative delay must be converted to absolute time.
 *    Worse, this breaks if the machine's time is adjusted while
 *    we're preparing to wait.
 * 2. Each cond operation requires additional mutex lock/unlock operations.
 *
 * Problem 2 is probably not too bad on Linux (they'll just become
 * relatively cheap futex operations), but problem 1 is the real killer.
 * Only use this implementation if no better alternatives are available!
 */
struct erts_iwait {
    erts_smp_atomic_t state;
    pthread_cond_t cond;
    pthread_mutex_t mutex;
};

static void iwait_lowlevel_init(struct erts_iwait *iwait)
{
    iwait->cond = (pthread_cond_t) PTHREAD_COND_INITIALIZER;
    iwait->mutex = (pthread_mutex_t) PTHREAD_MUTEX_INITIALIZER;
}

static void iwait_lowlevel_wait(struct erts_iwait *iwait, struct timeval *delay)
{
    struct timeval tmp;
    struct timespec timeout;

    /* Due to pthread_cond_timedwait()'s use of absolute
       time, this must be the real gettimeofday(), _not_
       the "smoothed" one beam/erl_time_sup.c implements. */
    gettimeofday(&tmp, NULL);

    tmp.tv_sec += delay->tv_sec;
    tmp.tv_usec += delay->tv_usec;
    if (tmp.tv_usec >= 1000*1000) {
	tmp.tv_usec -= 1000*1000;
	tmp.tv_sec += 1;
    }
    timeout.tv_sec = tmp.tv_sec;
    timeout.tv_nsec = tmp.tv_usec * 1000;
    pthread_mutex_lock(&iwait->mutex);
    pthread_cond_timedwait(&iwait->cond, &iwait->mutex, &timeout);
    pthread_mutex_unlock(&iwait->mutex);
}

static void iwait_lowlevel_interrupt(struct erts_iwait *iwait)
{
    pthread_mutex_lock(&iwait->mutex);
    pthread_cond_signal(&iwait->cond);
    pthread_mutex_unlock(&iwait->mutex);
}

#endif /* not using POLL */

/*
 * Interruptible-wait facility. This is just a wrapper around the
 * low-level synchronisation code, where we maintain our logical
 * state in order to suppress some state transitions.
 */

struct erts_iwait *erts_iwait_init(void)
{
    struct erts_iwait *iwait = malloc(sizeof *iwait);
    if (!iwait) {
	perror("malloc");
	exit(1);
    }
    iwait_lowlevel_init(iwait);
    erts_smp_atomic_init(&iwait->state, IWAIT_AWAKE);
    return iwait;
}

void erts_iwait_wait(struct erts_iwait *iwait, struct timeval *delay)
{
    if (erts_smp_atomic_xchg(&iwait->state, IWAIT_WAITING) != IWAIT_INTERRUPT)
	iwait_lowlevel_wait(iwait, delay);
    erts_smp_atomic_set(&iwait->state, IWAIT_AWAKE);
}

void erts_iwait_interrupt(struct erts_iwait *iwait)
{
    if (erts_smp_atomic_xchg(&iwait->state, IWAIT_INTERRUPT) == IWAIT_WAITING)
	iwait_lowlevel_interrupt(iwait);
}

#endif /* ERTS_TIMER_THREAD */