aboutsummaryrefslogtreecommitdiff
path: root/drivers/net/ethernet/packetengines/hamachi.c
blob: 94823c5f7dffcd71d73bdc444aa84de762918b39 (plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
1001
1002
1003
1004
1005
1006
1007
1008
1009
1010
1011
1012
1013
1014
1015
1016
1017
1018
1019
1020
1021
1022
1023
1024
1025
1026
1027
1028
1029
1030
1031
1032
1033
1034
1035
1036
1037
1038
1039
1040
1041
1042
1043
1044
1045
1046
1047
1048
1049
1050
1051
1052
1053
1054
1055
1056
1057
1058
1059
1060
1061
1062
1063
1064
1065
1066
1067
1068
1069
1070
1071
1072
1073
1074
1075
1076
1077
1078
1079
1080
1081
1082
1083
1084
1085
1086
1087
1088
1089
1090
1091
1092
1093
1094
1095
1096
1097
1098
1099
1100
1101
1102
1103
1104
1105
1106
1107
1108
1109
1110
1111
1112
1113
1114
1115
1116
1117
1118
1119
1120
1121
1122
1123
1124
1125
1126
1127
1128
1129
1130
1131
1132
1133
1134
1135
1136
1137
1138
1139
1140
1141
1142
1143
1144
1145
1146
1147
1148
1149
1150
1151
1152
1153
1154
1155
1156
1157
1158
1159
1160
1161
1162
1163
1164
1165
1166
1167
1168
1169
1170
1171
1172
1173
1174
1175
1176
1177
1178
1179
1180
1181
1182
1183
1184
1185
1186
1187
1188
1189
1190
1191
1192
1193
1194
1195
1196
1197
1198
1199
1200
1201
1202
1203
1204
1205
1206
1207
1208
1209
1210
1211
1212
1213
1214
1215
1216
1217
1218
1219
1220
1221
1222
1223
1224
1225
1226
1227
1228
1229
1230
1231
1232
1233
1234
1235
1236
1237
1238
1239
1240
1241
1242
1243
1244
1245
1246
1247
1248
1249
1250
1251
1252
1253
1254
1255
1256
1257
1258
1259
1260
1261
1262
1263
1264
1265
1266
1267
1268
1269
1270
1271
1272
1273
1274
1275
1276
1277
1278
1279
1280
1281
1282
1283
1284
1285
1286
1287
1288
1289
1290
1291
1292
1293
1294
1295
1296
1297
1298
1299
1300
1301
1302
1303
1304
1305
1306
1307
1308
1309
1310
1311
1312
1313
1314
1315
1316
1317
1318
1319
1320
1321
1322
1323
1324
1325
1326
1327
1328
1329
1330
1331
1332
1333
1334
1335
1336
1337
1338
1339
1340
1341
1342
1343
1344
1345
1346
1347
1348
1349
1350
1351
1352
1353
1354
1355
1356
1357
1358
1359
1360
1361
1362
1363
1364
1365
1366
1367
1368
1369
1370
1371
1372
1373
1374
1375
1376
1377
1378
1379
1380
1381
1382
1383
1384
1385
1386
1387
1388
1389
1390
1391
1392
1393
1394
1395
1396
1397
1398
1399
1400
1401
1402
1403
1404
1405
1406
1407
1408
1409
1410
1411
1412
1413
1414
1415
1416
1417
1418
1419
1420
1421
1422
1423
1424
1425
1426
1427
1428
1429
1430
1431
1432
1433
1434
1435
1436
1437
1438
1439
1440
1441
1442
1443
1444
1445
1446
1447
1448
1449
1450
1451
1452
1453
1454
1455
1456
1457
1458
1459
1460
1461
1462
1463
1464
1465
1466
1467
1468
1469
1470
1471
1472
1473
1474
1475
1476
1477
1478
1479
1480
1481
1482
1483
1484
1485
1486
1487
1488
1489
1490
1491
1492
1493
1494
1495
1496
1497
1498
1499
1500
1501
1502
1503
1504
1505
1506
1507
1508
1509
1510
1511
1512
1513
1514
1515
1516
1517
1518
1519
1520
1521
1522
1523
1524
1525
1526
1527
1528
1529
1530
1531
1532
1533
1534
1535
1536
1537
1538
1539
1540
1541
1542
1543
1544
1545
1546
1547
1548
1549
1550
1551
1552
1553
1554
1555
1556
1557
1558
1559
1560
1561
1562
1563
1564
1565
1566
1567
1568
1569
1570
1571
1572
1573
1574
1575
1576
1577
1578
1579
1580
1581
1582
1583
1584
1585
1586
1587
1588
1589
1590
1591
1592
1593
1594
1595
1596
1597
1598
1599
1600
1601
1602
1603
1604
1605
1606
1607
1608
1609
1610
1611
1612
1613
1614
1615
1616
1617
1618
1619
1620
1621
1622
1623
1624
1625
1626
1627
1628
1629
1630
1631
1632
1633
1634
1635
1636
1637
1638
1639
1640
1641
1642
1643
1644
1645
1646
1647
1648
1649
1650
1651
1652
1653
1654
1655
1656
1657
1658
1659
1660
1661
1662
1663
1664
1665
1666
1667
1668
1669
1670
1671
1672
1673
1674
1675
1676
1677
1678
1679
1680
1681
1682
1683
1684
1685
1686
1687
1688
1689
1690
1691
1692
1693
1694
1695
1696
1697
1698
1699
1700
1701
1702
1703
1704
1705
1706
1707
1708
1709
1710
1711
1712
1713
1714
1715
1716
1717
1718
1719
1720
1721
1722
1723
1724
1725
1726
1727
1728
1729
1730
1731
1732
1733
1734
1735
1736
1737
1738
1739
1740
1741
1742
1743
1744
1745
1746
1747
1748
1749
1750
1751
1752
1753
1754
1755
1756
1757
1758
1759
1760
1761
1762
1763
1764
1765
1766
1767
1768
1769
1770
1771
1772
1773
1774
1775
1776
1777
1778
1779
1780
1781
1782
1783
1784
1785
1786
1787
1788
1789
1790
1791
1792
1793
1794
1795
1796
1797
1798
1799
1800
1801
1802
1803
1804
1805
1806
1807
1808
1809
1810
1811
1812
1813
1814
1815
1816
1817
1818
1819
1820
1821
1822
1823
1824
1825
1826
1827
1828
1829
1830
1831
1832
1833
1834
1835
1836
1837
1838
1839
1840
1841
1842
1843
1844
1845
1846
1847
1848
1849
1850
1851
1852
1853
1854
1855
1856
1857
1858
1859
1860
1861
1862
1863
1864
1865
1866
1867
1868
1869
1870
1871
1872
1873
1874
1875
1876
1877
1878
1879
1880
1881
1882
1883
1884
1885
1886
1887
1888
1889
1890
1891
1892
1893
1894
1895
1896
1897
1898
1899
1900
1901
1902
1903
1904
1905
1906
1907
1908
1909
1910
1911
1912
1913
1914
1915
1916
1917
1918
1919
1920
1921
1922
1923
1924
1925
1926
1927
1928
1929
1930
1931
1932
1933
1934
1935
1936
1937
1938
1939
1940
1941
1942
1943
1944
1945
1946
1947
1948
1949
1950
1951
1952
1953
1954
1955
1956
1957
1958
1959
1960
1961
1962
1963
1964
1965
1966
/* hamachi.c: A Packet Engines GNIC-II Gigabit Ethernet driver for Linux. */
/*
	Written 1998-2000 by Donald Becker.
	Updates 2000 by Keith Underwood.

	This software may be used and distributed according to the terms of
	the GNU General Public License (GPL), incorporated herein by reference.
	Drivers based on or derived from this code fall under the GPL and must
	retain the authorship, copyright and license notice.  This file is not
	a complete program and may only be used when the entire operating
	system is licensed under the GPL.

	The author may be reached as becker@scyld.com, or C/O
	Scyld Computing Corporation
	410 Severn Ave., Suite 210
	Annapolis MD 21403

	This driver is for the Packet Engines GNIC-II PCI Gigabit Ethernet
	adapter.

	Support and updates available at
	http://www.scyld.com/network/hamachi.html
	[link no longer provides useful info -jgarzik]
	or
	http://www.parl.clemson.edu/~keithu/hamachi.html

*/

#define DRV_NAME	"hamachi"
#define DRV_VERSION	"2.1"
#define DRV_RELDATE	"Sept 11, 2006"


/* A few user-configurable values. */

static int debug = 1;		/* 1 normal messages, 0 quiet .. 7 verbose.  */
#define final_version
#define hamachi_debug debug
/* Maximum events (Rx packets, etc.) to handle at each interrupt. */
static int max_interrupt_work = 40;
static int mtu;
/* Default values selected by testing on a dual processor PIII-450 */
/* These six interrupt control parameters may be set directly when loading the
 * module, or through the rx_params and tx_params variables
 */
static int max_rx_latency = 0x11;
static int max_rx_gap = 0x05;
static int min_rx_pkt = 0x18;
static int max_tx_latency = 0x00;
static int max_tx_gap = 0x00;
static int min_tx_pkt = 0x30;

/* Set the copy breakpoint for the copy-only-tiny-frames scheme.
   -Setting to > 1518 causes all frames to be copied
	-Setting to 0 disables copies
*/
static int rx_copybreak;

/* An override for the hardware detection of bus width.
	Set to 1 to force 32 bit PCI bus detection.  Set to 4 to force 64 bit.
	Add 2 to disable parity detection.
*/
static int force32;


/* Used to pass the media type, etc.
   These exist for driver interoperability.
   No media types are currently defined.
		- The lower 4 bits are reserved for the media type.
		- The next three bits may be set to one of the following:
			0x00000000 : Autodetect PCI bus
			0x00000010 : Force 32 bit PCI bus
			0x00000020 : Disable parity detection
			0x00000040 : Force 64 bit PCI bus
			Default is autodetect
		- The next bit can be used to force half-duplex.  This is a bad
		  idea since no known implementations implement half-duplex, and,
		  in general, half-duplex for gigabit ethernet is a bad idea.
			0x00000080 : Force half-duplex
			Default is full-duplex.
		- In the original driver, the ninth bit could be used to force
		  full-duplex.  Maintain that for compatibility
		   0x00000200 : Force full-duplex
*/
#define MAX_UNITS 8				/* More are supported, limit only on options */
static int options[MAX_UNITS] = {-1, -1, -1, -1, -1, -1, -1, -1};
static int full_duplex[MAX_UNITS] = {-1, -1, -1, -1, -1, -1, -1, -1};
/* The Hamachi chipset supports 3 parameters each for Rx and Tx
 * interruput management.  Parameters will be loaded as specified into
 * the TxIntControl and RxIntControl registers.
 *
 * The registers are arranged as follows:
 *     23 - 16   15 -  8   7    -    0
 *    _________________________________
 *   | min_pkt | max_gap | max_latency |
 *    ---------------------------------
 *   min_pkt      : The minimum number of packets processed between
 *                  interrupts.
 *   max_gap      : The maximum inter-packet gap in units of 8.192 us
 *   max_latency  : The absolute time between interrupts in units of 8.192 us
 *
 */
static int rx_params[MAX_UNITS] = {-1, -1, -1, -1, -1, -1, -1, -1};
static int tx_params[MAX_UNITS] = {-1, -1, -1, -1, -1, -1, -1, -1};

/* Operational parameters that are set at compile time. */

/* Keep the ring sizes a power of two for compile efficiency.
	The compiler will convert <unsigned>'%'<2^N> into a bit mask.
   Making the Tx ring too large decreases the effectiveness of channel
   bonding and packet priority.
   There are no ill effects from too-large receive rings, except for
	excessive memory usage */
/* Empirically it appears that the Tx ring needs to be a little bigger
   for these Gbit adapters or you get into an overrun condition really
   easily.  Also, things appear to work a bit better in back-to-back
   configurations if the Rx ring is 8 times the size of the Tx ring
*/
#define TX_RING_SIZE	64
#define RX_RING_SIZE	512
#define TX_TOTAL_SIZE	TX_RING_SIZE*sizeof(struct hamachi_desc)
#define RX_TOTAL_SIZE	RX_RING_SIZE*sizeof(struct hamachi_desc)

/*
 * Enable netdev_ioctl.  Added interrupt coalescing parameter adjustment.
 * 2/19/99 Pete Wyckoff <wyckoff@ca.sandia.gov>
 */

/* play with 64-bit addrlen; seems to be a teensy bit slower  --pw */
/* #define ADDRLEN 64 */

/*
 * RX_CHECKSUM turns on card-generated receive checksum generation for
 *   TCP and UDP packets.  Otherwise the upper layers do the calculation.
 * 3/10/1999 Pete Wyckoff <wyckoff@ca.sandia.gov>
 */
#define RX_CHECKSUM

/* Operational parameters that usually are not changed. */
/* Time in jiffies before concluding the transmitter is hung. */
#define TX_TIMEOUT  (5*HZ)

#include <linux/capability.h>
#include <linux/module.h>
#include <linux/kernel.h>
#include <linux/string.h>
#include <linux/timer.h>
#include <linux/time.h>
#include <linux/errno.h>
#include <linux/ioport.h>
#include <linux/interrupt.h>
#include <linux/pci.h>
#include <linux/init.h>
#include <linux/ethtool.h>
#include <linux/mii.h>
#include <linux/netdevice.h>
#include <linux/etherdevice.h>
#include <linux/skbuff.h>
#include <linux/ip.h>
#include <linux/delay.h>
#include <linux/bitops.h>

#include <linux/uaccess.h>
#include <asm/processor.h>	/* Processor type for cache alignment. */
#include <asm/io.h>
#include <asm/unaligned.h>
#include <asm/cache.h>

static const char version[] =
KERN_INFO DRV_NAME ".c:v" DRV_VERSION " " DRV_RELDATE "  Written by Donald Becker\n"
"   Some modifications by Eric kasten <kasten@nscl.msu.edu>\n"
"   Further modifications by Keith Underwood <keithu@parl.clemson.edu>\n";


/* IP_MF appears to be only defined in <netinet/ip.h>, however,
   we need it for hardware checksumming support.  FYI... some of
   the definitions in <netinet/ip.h> conflict/duplicate those in
   other linux headers causing many compiler warnings.
*/
#ifndef IP_MF
  #define IP_MF 0x2000   /* IP more frags from <netinet/ip.h> */
#endif

/* Define IP_OFFSET to be IPOPT_OFFSET */
#ifndef IP_OFFSET
  #ifdef IPOPT_OFFSET
    #define IP_OFFSET IPOPT_OFFSET
  #else
    #define IP_OFFSET 2
  #endif
#endif

#define RUN_AT(x) (jiffies + (x))

#ifndef ADDRLEN
#define ADDRLEN 32
#endif

/* Condensed bus+endian portability operations. */
#if ADDRLEN == 64
#define cpu_to_leXX(addr)	cpu_to_le64(addr)
#define leXX_to_cpu(addr)	le64_to_cpu(addr)
#else
#define cpu_to_leXX(addr)	cpu_to_le32(addr)
#define leXX_to_cpu(addr)	le32_to_cpu(addr)
#endif


/*
				Theory of Operation

I. Board Compatibility

This device driver is designed for the Packet Engines "Hamachi"
Gigabit Ethernet chip.  The only PCA currently supported is the GNIC-II 64-bit
66Mhz PCI card.

II. Board-specific settings

No jumpers exist on the board.  The chip supports software correction of
various motherboard wiring errors, however this driver does not support
that feature.

III. Driver operation

IIIa. Ring buffers

The Hamachi uses a typical descriptor based bus-master architecture.
The descriptor list is similar to that used by the Digital Tulip.
This driver uses two statically allocated fixed-size descriptor lists
formed into rings by a branch from the final descriptor to the beginning of
the list.  The ring sizes are set at compile time by RX/TX_RING_SIZE.

This driver uses a zero-copy receive and transmit scheme similar my other
network drivers.
The driver allocates full frame size skbuffs for the Rx ring buffers at
open() time and passes the skb->data field to the Hamachi as receive data
buffers.  When an incoming frame is less than RX_COPYBREAK bytes long,
a fresh skbuff is allocated and the frame is copied to the new skbuff.
When the incoming frame is larger, the skbuff is passed directly up the
protocol stack and replaced by a newly allocated skbuff.

The RX_COPYBREAK value is chosen to trade-off the memory wasted by
using a full-sized skbuff for small frames vs. the copying costs of larger
frames.  Gigabit cards are typically used on generously configured machines
and the underfilled buffers have negligible impact compared to the benefit of
a single allocation size, so the default value of zero results in never
copying packets.

IIIb/c. Transmit/Receive Structure

The Rx and Tx descriptor structure are straight-forward, with no historical
baggage that must be explained.  Unlike the awkward DBDMA structure, there
are no unused fields or option bits that had only one allowable setting.

Two details should be noted about the descriptors: The chip supports both 32
bit and 64 bit address structures, and the length field is overwritten on
the receive descriptors.  The descriptor length is set in the control word
for each channel. The development driver uses 32 bit addresses only, however
64 bit addresses may be enabled for 64 bit architectures e.g. the Alpha.

IIId. Synchronization

This driver is very similar to my other network drivers.
The driver runs as two independent, single-threaded flows of control.  One
is the send-packet routine, which enforces single-threaded use by the
dev->tbusy flag.  The other thread is the interrupt handler, which is single
threaded by the hardware and other software.

The send packet thread has partial control over the Tx ring and 'dev->tbusy'
flag.  It sets the tbusy flag whenever it's queuing a Tx packet. If the next
queue slot is empty, it clears the tbusy flag when finished otherwise it sets
the 'hmp->tx_full' flag.

The interrupt handler has exclusive control over the Rx ring and records stats
from the Tx ring.  After reaping the stats, it marks the Tx queue entry as
empty by incrementing the dirty_tx mark. Iff the 'hmp->tx_full' flag is set, it
clears both the tx_full and tbusy flags.

IV. Notes

Thanks to Kim Stearns of Packet Engines for providing a pair of GNIC-II boards.

IVb. References

Hamachi Engineering Design Specification, 5/15/97
(Note: This version was marked "Confidential".)

IVc. Errata

None noted.

V.  Recent Changes

01/15/1999 EPK  Enlargement of the TX and RX ring sizes.  This appears
    to help avoid some stall conditions -- this needs further research.

01/15/1999 EPK  Creation of the hamachi_tx function.  This function cleans
    the Tx ring and is called from hamachi_start_xmit (this used to be
    called from hamachi_interrupt but it tends to delay execution of the
    interrupt handler and thus reduce bandwidth by reducing the latency
    between hamachi_rx()'s).  Notably, some modification has been made so
    that the cleaning loop checks only to make sure that the DescOwn bit
    isn't set in the status flag since the card is not required
    to set the entire flag to zero after processing.

01/15/1999 EPK In the hamachi_start_tx function, the Tx ring full flag is
    checked before attempting to add a buffer to the ring.  If the ring is full
    an attempt is made to free any dirty buffers and thus find space for
    the new buffer or the function returns non-zero which should case the
    scheduler to reschedule the buffer later.

01/15/1999 EPK Some adjustments were made to the chip initialization.
    End-to-end flow control should now be fully active and the interrupt
    algorithm vars have been changed.  These could probably use further tuning.

01/15/1999 EPK Added the max_{rx,tx}_latency options.  These are used to
    set the rx and tx latencies for the Hamachi interrupts. If you're having
    problems with network stalls, try setting these to higher values.
    Valid values are 0x00 through 0xff.

01/15/1999 EPK In general, the overall bandwidth has increased and
    latencies are better (sometimes by a factor of 2).  Stalls are rare at
    this point, however there still appears to be a bug somewhere between the
    hardware and driver.  TCP checksum errors under load also appear to be
    eliminated at this point.

01/18/1999 EPK Ensured that the DescEndRing bit was being set on both the
    Rx and Tx rings.  This appears to have been affecting whether a particular
    peer-to-peer connection would hang under high load.  I believe the Rx
    rings was typically getting set correctly, but the Tx ring wasn't getting
    the DescEndRing bit set during initialization. ??? Does this mean the
    hamachi card is using the DescEndRing in processing even if a particular
    slot isn't in use -- hypothetically, the card might be searching the
    entire Tx ring for slots with the DescOwn bit set and then processing
    them.  If the DescEndRing bit isn't set, then it might just wander off
    through memory until it hits a chunk of data with that bit set
    and then looping back.

02/09/1999 EPK Added Michel Mueller's TxDMA Interrupt and Tx-timeout
    problem (TxCmd and RxCmd need only to be set when idle or stopped.

02/09/1999 EPK Added code to check/reset dev->tbusy in hamachi_interrupt.
    (Michel Mueller pointed out the ``permanently busy'' potential
    problem here).

02/22/1999 EPK Added Pete Wyckoff's ioctl to control the Tx/Rx latencies.

02/23/1999 EPK Verified that the interrupt status field bits for Tx were
    incorrectly defined and corrected (as per Michel Mueller).

02/23/1999 EPK Corrected the Tx full check to check that at least 4 slots
    were available before resetting the tbusy and tx_full flags
    (as per Michel Mueller).

03/11/1999 EPK Added Pete Wyckoff's hardware checksumming support.

12/31/1999 KDU Cleaned up assorted things and added Don's code to force
32 bit.

02/20/2000 KDU Some of the control was just plain odd.  Cleaned up the
hamachi_start_xmit() and hamachi_interrupt() code.  There is still some
re-structuring I would like to do.

03/01/2000 KDU Experimenting with a WIDE range of interrupt mitigation
parameters on a dual P3-450 setup yielded the new default interrupt
mitigation parameters.  Tx should interrupt VERY infrequently due to
Eric's scheme.  Rx should be more often...

03/13/2000 KDU Added a patch to make the Rx Checksum code interact
nicely with non-linux machines.

03/13/2000 KDU Experimented with some of the configuration values:

	-It seems that enabling PCI performance commands for descriptors
	(changing RxDMACtrl and TxDMACtrl lower nibble from 5 to D) has minimal
	performance impact for any of my tests. (ttcp, netpipe, netperf)  I will
	leave them that way until I hear further feedback.

	-Increasing the PCI_LATENCY_TIMER to 130
	(2 + (burst size of 128 * (0 wait states + 1))) seems to slightly
	degrade performance.  Leaving default at 64 pending further information.

03/14/2000 KDU Further tuning:

	-adjusted boguscnt in hamachi_rx() to depend on interrupt
	mitigation parameters chosen.

	-Selected a set of interrupt parameters based on some extensive testing.
	These may change with more testing.

TO DO:

-Consider borrowing from the acenic driver code to check PCI_COMMAND for
PCI_COMMAND_INVALIDATE.  Set maximum burst size to cache line size in
that case.

-fix the reset procedure.  It doesn't quite work.
*/

/* A few values that may be tweaked. */
/* Size of each temporary Rx buffer, calculated as:
 * 1518 bytes (ethernet packet) + 2 bytes (to get 8 byte alignment for
 * the card) + 8 bytes of status info + 8 bytes for the Rx Checksum
 */
#define PKT_BUF_SZ		1536

/* For now, this is going to be set to the maximum size of an ethernet
 * packet.  Eventually, we may want to make it a variable that is
 * related to the MTU
 */
#define MAX_FRAME_SIZE  1518

/* The rest of these values should never change. */

static void hamachi_timer(struct timer_list *t);

enum capability_flags {CanHaveMII=1, };
static const struct chip_info {
	u16	vendor_id, device_id, device_id_mask, pad;
	const char *name;
	void (*media_timer)(struct timer_list *t);
	int flags;
} chip_tbl[] = {
	{0x1318, 0x0911, 0xffff, 0, "Hamachi GNIC-II", hamachi_timer, 0},
	{0,},
};

/* Offsets to the Hamachi registers.  Various sizes. */
enum hamachi_offsets {
	TxDMACtrl=0x00, TxCmd=0x04, TxStatus=0x06, TxPtr=0x08, TxCurPtr=0x10,
	RxDMACtrl=0x20, RxCmd=0x24, RxStatus=0x26, RxPtr=0x28, RxCurPtr=0x30,
	PCIClkMeas=0x060, MiscStatus=0x066, ChipRev=0x68, ChipReset=0x06B,
	LEDCtrl=0x06C, VirtualJumpers=0x06D, GPIO=0x6E,
	TxChecksum=0x074, RxChecksum=0x076,
	TxIntrCtrl=0x078, RxIntrCtrl=0x07C,
	InterruptEnable=0x080, InterruptClear=0x084, IntrStatus=0x088,
	EventStatus=0x08C,
	MACCnfg=0x0A0, FrameGap0=0x0A2, FrameGap1=0x0A4,
	/* See enum MII_offsets below. */
	MACCnfg2=0x0B0, RxDepth=0x0B8, FlowCtrl=0x0BC, MaxFrameSize=0x0CE,
	AddrMode=0x0D0, StationAddr=0x0D2,
	/* Gigabit AutoNegotiation. */
	ANCtrl=0x0E0, ANStatus=0x0E2, ANXchngCtrl=0x0E4, ANAdvertise=0x0E8,
	ANLinkPartnerAbility=0x0EA,
	EECmdStatus=0x0F0, EEData=0x0F1, EEAddr=0x0F2,
	FIFOcfg=0x0F8,
};

/* Offsets to the MII-mode registers. */
enum MII_offsets {
	MII_Cmd=0xA6, MII_Addr=0xA8, MII_Wr_Data=0xAA, MII_Rd_Data=0xAC,
	MII_Status=0xAE,
};

/* Bits in the interrupt status/mask registers. */
enum intr_status_bits {
	IntrRxDone=0x01, IntrRxPCIFault=0x02, IntrRxPCIErr=0x04,
	IntrTxDone=0x100, IntrTxPCIFault=0x200, IntrTxPCIErr=0x400,
	LinkChange=0x10000, NegotiationChange=0x20000, StatsMax=0x40000, };

/* The Hamachi Rx and Tx buffer descriptors. */
struct hamachi_desc {
	__le32 status_n_length;
#if ADDRLEN == 64
	u32 pad;
	__le64 addr;
#else
	__le32 addr;
#endif
};

/* Bits in hamachi_desc.status_n_length */
enum desc_status_bits {
	DescOwn=0x80000000, DescEndPacket=0x40000000, DescEndRing=0x20000000,
	DescIntr=0x10000000,
};

#define PRIV_ALIGN	15  			/* Required alignment mask */
#define MII_CNT		4
struct hamachi_private {
	/* Descriptor rings first for alignment.  Tx requires a second descriptor
	   for status. */
	struct hamachi_desc *rx_ring;
	struct hamachi_desc *tx_ring;
	struct sk_buff* rx_skbuff[RX_RING_SIZE];
	struct sk_buff* tx_skbuff[TX_RING_SIZE];
	dma_addr_t tx_ring_dma;
	dma_addr_t rx_ring_dma;
	struct timer_list timer;		/* Media selection timer. */
	/* Frequently used and paired value: keep adjacent for cache effect. */
	spinlock_t lock;
	int chip_id;
	unsigned int cur_rx, dirty_rx;		/* Producer/consumer ring indices */
	unsigned int cur_tx, dirty_tx;
	unsigned int rx_buf_sz;			/* Based on MTU+slack. */
	unsigned int tx_full:1;			/* The Tx queue is full. */
	unsigned int duplex_lock:1;
	unsigned int default_port:4;		/* Last dev->if_port value. */
	/* MII transceiver section. */
	int mii_cnt;								/* MII device addresses. */
	struct mii_if_info mii_if;		/* MII lib hooks/info */
	unsigned char phys[MII_CNT];		/* MII device addresses, only first one used. */
	u32 rx_int_var, tx_int_var;	/* interrupt control variables */
	u32 option;							/* Hold on to a copy of the options */
	struct pci_dev *pci_dev;
	void __iomem *base;
};

MODULE_AUTHOR("Donald Becker <becker@scyld.com>, Eric Kasten <kasten@nscl.msu.edu>, Keith Underwood <keithu@parl.clemson.edu>");
MODULE_DESCRIPTION("Packet Engines 'Hamachi' GNIC-II Gigabit Ethernet driver");
MODULE_LICENSE("GPL");

module_param(max_interrupt_work, int, 0);
module_param(mtu, int, 0);
module_param(debug, int, 0);
module_param(min_rx_pkt, int, 0);
module_param(max_rx_gap, int, 0);
module_param(max_rx_latency, int, 0);
module_param(min_tx_pkt, int, 0);
module_param(max_tx_gap, int, 0);
module_param(max_tx_latency, int, 0);
module_param(rx_copybreak, int, 0);
module_param_array(rx_params, int, NULL, 0);
module_param_array(tx_params, int, NULL, 0);
module_param_array(options, int, NULL, 0);
module_param_array(full_duplex, int, NULL, 0);
module_param(force32, int, 0);
MODULE_PARM_DESC(max_interrupt_work, "GNIC-II maximum events handled per interrupt");
MODULE_PARM_DESC(mtu, "GNIC-II MTU (all boards)");
MODULE_PARM_DESC(debug, "GNIC-II debug level (0-7)");
MODULE_PARM_DESC(min_rx_pkt, "GNIC-II minimum Rx packets processed between interrupts");
MODULE_PARM_DESC(max_rx_gap, "GNIC-II maximum Rx inter-packet gap in 8.192 microsecond units");
MODULE_PARM_DESC(max_rx_latency, "GNIC-II time between Rx interrupts in 8.192 microsecond units");
MODULE_PARM_DESC(min_tx_pkt, "GNIC-II minimum Tx packets processed between interrupts");
MODULE_PARM_DESC(max_tx_gap, "GNIC-II maximum Tx inter-packet gap in 8.192 microsecond units");
MODULE_PARM_DESC(max_tx_latency, "GNIC-II time between Tx interrupts in 8.192 microsecond units");
MODULE_PARM_DESC(rx_copybreak, "GNIC-II copy breakpoint for copy-only-tiny-frames");
MODULE_PARM_DESC(rx_params, "GNIC-II min_rx_pkt+max_rx_gap+max_rx_latency");
MODULE_PARM_DESC(tx_params, "GNIC-II min_tx_pkt+max_tx_gap+max_tx_latency");
MODULE_PARM_DESC(options, "GNIC-II Bits 0-3: media type, bits 4-6: as force32, bit 7: half duplex, bit 9 full duplex");
MODULE_PARM_DESC(full_duplex, "GNIC-II full duplex setting(s) (1)");
MODULE_PARM_DESC(force32, "GNIC-II: Bit 0: 32 bit PCI, bit 1: disable parity, bit 2: 64 bit PCI (all boards)");

static int read_eeprom(void __iomem *ioaddr, int location);
static int mdio_read(struct net_device *dev, int phy_id, int location);
static void mdio_write(struct net_device *dev, int phy_id, int location, int value);
static int hamachi_open(struct net_device *dev);
static int hamachi_ioctl(struct net_device *dev, struct ifreq *rq, int cmd);
static int hamachi_siocdevprivate(struct net_device *dev, struct ifreq *rq,
				  void __user *data, int cmd);
static void hamachi_timer(struct timer_list *t);
static void hamachi_tx_timeout(struct net_device *dev, unsigned int txqueue);
static void hamachi_init_ring(struct net_device *dev);
static netdev_tx_t hamachi_start_xmit(struct sk_buff *skb,
				      struct net_device *dev);
static irqreturn_t hamachi_interrupt(int irq, void *dev_instance);
static int hamachi_rx(struct net_device *dev);
static inline int hamachi_tx(struct net_device *dev);
static void hamachi_error(struct net_device *dev, int intr_status);
static int hamachi_close(struct net_device *dev);
static struct net_device_stats *hamachi_get_stats(struct net_device *dev);
static void set_rx_mode(struct net_device *dev);
static const struct ethtool_ops ethtool_ops;
static const struct ethtool_ops ethtool_ops_no_mii;

static const struct net_device_ops hamachi_netdev_ops = {
	.ndo_open		= hamachi_open,
	.ndo_stop		= hamachi_close,
	.ndo_start_xmit		= hamachi_start_xmit,
	.ndo_get_stats		= hamachi_get_stats,
	.ndo_set_rx_mode	= set_rx_mode,
	.ndo_validate_addr	= eth_validate_addr,
	.ndo_set_mac_address 	= eth_mac_addr,
	.ndo_tx_timeout		= hamachi_tx_timeout,
	.ndo_do_ioctl		= hamachi_ioctl,
	.ndo_siocdevprivate	= hamachi_siocdevprivate,
};


static int hamachi_init_one(struct pci_dev *pdev,
			    const struct pci_device_id *ent)
{
	struct hamachi_private *hmp;
	int option, i, rx_int_var, tx_int_var, boguscnt;
	int chip_id = ent->driver_data;
	int irq;
	void __iomem *ioaddr;
	unsigned long base;
	static int card_idx;
	struct net_device *dev;
	void *ring_space;
	dma_addr_t ring_dma;
	int ret = -ENOMEM;

/* when built into the kernel, we only print version if device is found */
#ifndef MODULE
	static int printed_version;
	if (!printed_version++)
		printk(version);
#endif

	if (pci_enable_device(pdev)) {
		ret = -EIO;
		goto err_out;
	}

	base = pci_resource_start(pdev, 0);
#ifdef __alpha__				/* Really "64 bit addrs" */
	base |= (pci_resource_start(pdev, 1) << 32);
#endif

	pci_set_master(pdev);

	i = pci_request_regions(pdev, DRV_NAME);
	if (i)
		return i;

	irq = pdev->irq;
	ioaddr = ioremap(base, 0x400);
	if (!ioaddr)
		goto err_out_release;

	dev = alloc_etherdev(sizeof(struct hamachi_private));
	if (!dev)
		goto err_out_iounmap;

	SET_NETDEV_DEV(dev, &pdev->dev);

	for (i = 0; i < 6; i++)
		dev->dev_addr[i] = 1 ? read_eeprom(ioaddr, 4 + i)
			: readb(ioaddr + StationAddr + i);

#if ! defined(final_version)
	if (hamachi_debug > 4)
		for (i = 0; i < 0x10; i++)
			printk("%2.2x%s",
				   read_eeprom(ioaddr, i), i % 16 != 15 ? " " : "\n");
#endif

	hmp = netdev_priv(dev);
	spin_lock_init(&hmp->lock);

	hmp->mii_if.dev = dev;
	hmp->mii_if.mdio_read = mdio_read;
	hmp->mii_if.mdio_write = mdio_write;
	hmp->mii_if.phy_id_mask = 0x1f;
	hmp->mii_if.reg_num_mask = 0x1f;

	ring_space = dma_alloc_coherent(&pdev->dev, TX_TOTAL_SIZE, &ring_dma,
					GFP_KERNEL);
	if (!ring_space)
		goto err_out_cleardev;
	hmp->tx_ring = ring_space;
	hmp->tx_ring_dma = ring_dma;

	ring_space = dma_alloc_coherent(&pdev->dev, RX_TOTAL_SIZE, &ring_dma,
					GFP_KERNEL);
	if (!ring_space)
		goto err_out_unmap_tx;
	hmp->rx_ring = ring_space;
	hmp->rx_ring_dma = ring_dma;

	/* Check for options being passed in */
	option = card_idx < MAX_UNITS ? options[card_idx] : 0;
	if (dev->mem_start)
		option = dev->mem_start;

	/* If the bus size is misidentified, do the following. */
	force32 = force32 ? force32 :
		((option  >= 0) ? ((option & 0x00000070) >> 4) : 0 );
	if (force32)
		writeb(force32, ioaddr + VirtualJumpers);

	/* Hmmm, do we really need to reset the chip???. */
	writeb(0x01, ioaddr + ChipReset);

	/* After a reset, the clock speed measurement of the PCI bus will not
	 * be valid for a moment.  Wait for a little while until it is.  If
	 * it takes more than 10ms, forget it.
	 */
	udelay(10);
	i = readb(ioaddr + PCIClkMeas);
	for (boguscnt = 0; (!(i & 0x080)) && boguscnt < 1000; boguscnt++){
		udelay(10);
		i = readb(ioaddr + PCIClkMeas);
	}

	hmp->base = ioaddr;
	pci_set_drvdata(pdev, dev);

	hmp->chip_id = chip_id;
	hmp->pci_dev = pdev;

	/* The lower four bits are the media type. */
	if (option > 0) {
		hmp->option = option;
		if (option & 0x200)
			hmp->mii_if.full_duplex = 1;
		else if (option & 0x080)
			hmp->mii_if.full_duplex = 0;
		hmp->default_port = option & 15;
		if (hmp->default_port)
			hmp->mii_if.force_media = 1;
	}
	if (card_idx < MAX_UNITS  &&  full_duplex[card_idx] > 0)
		hmp->mii_if.full_duplex = 1;

	/* lock the duplex mode if someone specified a value */
	if (hmp->mii_if.full_duplex || (option & 0x080))
		hmp->duplex_lock = 1;

	/* Set interrupt tuning parameters */
	max_rx_latency = max_rx_latency & 0x00ff;
	max_rx_gap = max_rx_gap & 0x00ff;
	min_rx_pkt = min_rx_pkt & 0x00ff;
	max_tx_latency = max_tx_latency & 0x00ff;
	max_tx_gap = max_tx_gap & 0x00ff;
	min_tx_pkt = min_tx_pkt & 0x00ff;

	rx_int_var = card_idx < MAX_UNITS ? rx_params[card_idx] : -1;
	tx_int_var = card_idx < MAX_UNITS ? tx_params[card_idx] : -1;
	hmp->rx_int_var = rx_int_var >= 0 ? rx_int_var :
		(min_rx_pkt << 16 | max_rx_gap << 8 | max_rx_latency);
	hmp->tx_int_var = tx_int_var >= 0 ? tx_int_var :
		(min_tx_pkt << 16 | max_tx_gap << 8 | max_tx_latency);


	/* The Hamachi-specific entries in the device structure. */
	dev->netdev_ops = &hamachi_netdev_ops;
	dev->ethtool_ops = (chip_tbl[hmp->chip_id].flags & CanHaveMII) ?
		&ethtool_ops : &ethtool_ops_no_mii;
	dev->watchdog_timeo = TX_TIMEOUT;
	if (mtu)
		dev->mtu = mtu;

	i = register_netdev(dev);
	if (i) {
		ret = i;
		goto err_out_unmap_rx;
	}

	printk(KERN_INFO "%s: %s type %x at %p, %pM, IRQ %d.\n",
		   dev->name, chip_tbl[chip_id].name, readl(ioaddr + ChipRev),
		   ioaddr, dev->dev_addr, irq);
	i = readb(ioaddr + PCIClkMeas);
	printk(KERN_INFO "%s:  %d-bit %d Mhz PCI bus (%d), Virtual Jumpers "
		   "%2.2x, LPA %4.4x.\n",
		   dev->name, readw(ioaddr + MiscStatus) & 1 ? 64 : 32,
		   i ? 2000/(i&0x7f) : 0, i&0x7f, (int)readb(ioaddr + VirtualJumpers),
		   readw(ioaddr + ANLinkPartnerAbility));

	if (chip_tbl[hmp->chip_id].flags & CanHaveMII) {
		int phy, phy_idx = 0;
		for (phy = 0; phy < 32 && phy_idx < MII_CNT; phy++) {
			int mii_status = mdio_read(dev, phy, MII_BMSR);
			if (mii_status != 0xffff  &&
				mii_status != 0x0000) {
				hmp->phys[phy_idx++] = phy;
				hmp->mii_if.advertising = mdio_read(dev, phy, MII_ADVERTISE);
				printk(KERN_INFO "%s: MII PHY found at address %d, status "
					   "0x%4.4x advertising %4.4x.\n",
					   dev->name, phy, mii_status, hmp->mii_if.advertising);
			}
		}
		hmp->mii_cnt = phy_idx;
		if (hmp->mii_cnt > 0)
			hmp->mii_if.phy_id = hmp->phys[0];
		else
			memset(&hmp->mii_if, 0, sizeof(hmp->mii_if));
	}
	/* Configure gigabit autonegotiation. */
	writew(0x0400, ioaddr + ANXchngCtrl);	/* Enable legacy links. */
	writew(0x08e0, ioaddr + ANAdvertise);	/* Set our advertise word. */
	writew(0x1000, ioaddr + ANCtrl);			/* Enable negotiation */

	card_idx++;
	return 0;

err_out_unmap_rx:
	dma_free_coherent(&pdev->dev, RX_TOTAL_SIZE, hmp->rx_ring,
			  hmp->rx_ring_dma);
err_out_unmap_tx:
	dma_free_coherent(&pdev->dev, TX_TOTAL_SIZE, hmp->tx_ring,
			  hmp->tx_ring_dma);
err_out_cleardev:
	free_netdev (dev);
err_out_iounmap:
	iounmap(ioaddr);
err_out_release:
	pci_release_regions(pdev);
err_out:
	return ret;
}

static int read_eeprom(void __iomem *ioaddr, int location)
{
	int bogus_cnt = 1000;

	/* We should check busy first - per docs -KDU */
	while ((readb(ioaddr + EECmdStatus) & 0x40)  && --bogus_cnt > 0);
	writew(location, ioaddr + EEAddr);
	writeb(0x02, ioaddr + EECmdStatus);
	bogus_cnt = 1000;
	while ((readb(ioaddr + EECmdStatus) & 0x40)  && --bogus_cnt > 0);
	if (hamachi_debug > 5)
		printk("   EEPROM status is %2.2x after %d ticks.\n",
			   (int)readb(ioaddr + EECmdStatus), 1000- bogus_cnt);
	return readb(ioaddr + EEData);
}

/* MII Managemen Data I/O accesses.
   These routines assume the MDIO controller is idle, and do not exit until
   the command is finished. */

static int mdio_read(struct net_device *dev, int phy_id, int location)
{
	struct hamachi_private *hmp = netdev_priv(dev);
	void __iomem *ioaddr = hmp->base;
	int i;

	/* We should check busy first - per docs -KDU */
	for (i = 10000; i >= 0; i--)
		if ((readw(ioaddr + MII_Status) & 1) == 0)
			break;
	writew((phy_id<<8) + location, ioaddr + MII_Addr);
	writew(0x0001, ioaddr + MII_Cmd);
	for (i = 10000; i >= 0; i--)
		if ((readw(ioaddr + MII_Status) & 1) == 0)
			break;
	return readw(ioaddr + MII_Rd_Data);
}

static void mdio_write(struct net_device *dev, int phy_id, int location, int value)
{
	struct hamachi_private *hmp = netdev_priv(dev);
	void __iomem *ioaddr = hmp->base;
	int i;

	/* We should check busy first - per docs -KDU */
	for (i = 10000; i >= 0; i--)
		if ((readw(ioaddr + MII_Status) & 1) == 0)
			break;
	writew((phy_id<<8) + location, ioaddr + MII_Addr);
	writew(value, ioaddr + MII_Wr_Data);

	/* Wait for the command to finish. */
	for (i = 10000; i >= 0; i--)
		if ((readw(ioaddr + MII_Status) & 1) == 0)
			break;
}


static int hamachi_open(struct net_device *dev)
{
	struct hamachi_private *hmp = netdev_priv(dev);
	void __iomem *ioaddr = hmp->base;
	int i;
	u32 rx_int_var, tx_int_var;
	u16 fifo_info;

	i = request_irq(hmp->pci_dev->irq, hamachi_interrupt, IRQF_SHARED,
			dev->name, dev);
	if (i)
		return i;

	hamachi_init_ring(dev);

#if ADDRLEN == 64
	/* writellll anyone ? */
	writel(hmp->rx_ring_dma, ioaddr + RxPtr);
	writel(hmp->rx_ring_dma >> 32, ioaddr + RxPtr + 4);
	writel(hmp->tx_ring_dma, ioaddr + TxPtr);
	writel(hmp->tx_ring_dma >> 32, ioaddr + TxPtr + 4);
#else
	writel(hmp->rx_ring_dma, ioaddr + RxPtr);
	writel(hmp->tx_ring_dma, ioaddr + TxPtr);
#endif

	/* TODO:  It would make sense to organize this as words since the card
	 * documentation does. -KDU
	 */
	for (i = 0; i < 6; i++)
		writeb(dev->dev_addr[i], ioaddr + StationAddr + i);

	/* Initialize other registers: with so many this eventually this will
	   converted to an offset/value list. */

	/* Configure the FIFO */
	fifo_info = (readw(ioaddr + GPIO) & 0x00C0) >> 6;
	switch (fifo_info){
		case 0 :
			/* No FIFO */
			writew(0x0000, ioaddr + FIFOcfg);
			break;
		case 1 :
			/* Configure the FIFO for 512K external, 16K used for Tx. */
			writew(0x0028, ioaddr + FIFOcfg);
			break;
		case 2 :
			/* Configure the FIFO for 1024 external, 32K used for Tx. */
			writew(0x004C, ioaddr + FIFOcfg);
			break;
		case 3 :
			/* Configure the FIFO for 2048 external, 32K used for Tx. */
			writew(0x006C, ioaddr + FIFOcfg);
			break;
		default :
			printk(KERN_WARNING "%s:  Unsupported external memory config!\n",
				dev->name);
			/* Default to no FIFO */
			writew(0x0000, ioaddr + FIFOcfg);
			break;
	}

	if (dev->if_port == 0)
		dev->if_port = hmp->default_port;


	/* Setting the Rx mode will start the Rx process. */
	/* If someone didn't choose a duplex, default to full-duplex */
	if (hmp->duplex_lock != 1)
		hmp->mii_if.full_duplex = 1;

	/* always 1, takes no more time to do it */
	writew(0x0001, ioaddr + RxChecksum);
	writew(0x0000, ioaddr + TxChecksum);
	writew(0x8000, ioaddr + MACCnfg); /* Soft reset the MAC */
	writew(0x215F, ioaddr + MACCnfg);
	writew(0x000C, ioaddr + FrameGap0);
	/* WHAT?!?!?  Why isn't this documented somewhere? -KDU */
	writew(0x1018, ioaddr + FrameGap1);
	/* Why do we enable receives/transmits here? -KDU */
	writew(0x0780, ioaddr + MACCnfg2); /* Upper 16 bits control LEDs. */
	/* Enable automatic generation of flow control frames, period 0xffff. */
	writel(0x0030FFFF, ioaddr + FlowCtrl);
	writew(MAX_FRAME_SIZE, ioaddr + MaxFrameSize); 	/* dev->mtu+14 ??? */

	/* Enable legacy links. */
	writew(0x0400, ioaddr + ANXchngCtrl);	/* Enable legacy links. */
	/* Initial Link LED to blinking red. */
	writeb(0x03, ioaddr + LEDCtrl);

	/* Configure interrupt mitigation.  This has a great effect on
	   performance, so systems tuning should start here!. */

	rx_int_var = hmp->rx_int_var;
	tx_int_var = hmp->tx_int_var;

	if (hamachi_debug > 1) {
		printk("max_tx_latency: %d, max_tx_gap: %d, min_tx_pkt: %d\n",
			tx_int_var & 0x00ff, (tx_int_var & 0x00ff00) >> 8,
			(tx_int_var & 0x00ff0000) >> 16);
		printk("max_rx_latency: %d, max_rx_gap: %d, min_rx_pkt: %d\n",
			rx_int_var & 0x00ff, (rx_int_var & 0x00ff00) >> 8,
			(rx_int_var & 0x00ff0000) >> 16);
		printk("rx_int_var: %x, tx_int_var: %x\n", rx_int_var, tx_int_var);
	}

	writel(tx_int_var, ioaddr + TxIntrCtrl);
	writel(rx_int_var, ioaddr + RxIntrCtrl);

	set_rx_mode(dev);

	netif_start_queue(dev);

	/* Enable interrupts by setting the interrupt mask. */
	writel(0x80878787, ioaddr + InterruptEnable);
	writew(0x0000, ioaddr + EventStatus);	/* Clear non-interrupting events */

	/* Configure and start the DMA channels. */
	/* Burst sizes are in the low three bits: size = 4<<(val&7) */
#if ADDRLEN == 64
	writew(0x005D, ioaddr + RxDMACtrl); 		/* 128 dword bursts */
	writew(0x005D, ioaddr + TxDMACtrl);
#else
	writew(0x001D, ioaddr + RxDMACtrl);
	writew(0x001D, ioaddr + TxDMACtrl);
#endif
	writew(0x0001, ioaddr + RxCmd);

	if (hamachi_debug > 2) {
		printk(KERN_DEBUG "%s: Done hamachi_open(), status: Rx %x Tx %x.\n",
			   dev->name, readw(ioaddr + RxStatus), readw(ioaddr + TxStatus));
	}
	/* Set the timer to check for link beat. */
	timer_setup(&hmp->timer, hamachi_timer, 0);
	hmp->timer.expires = RUN_AT((24*HZ)/10);			/* 2.4 sec. */
	add_timer(&hmp->timer);

	return 0;
}

static inline int hamachi_tx(struct net_device *dev)
{
	struct hamachi_private *hmp = netdev_priv(dev);

	/* Update the dirty pointer until we find an entry that is
		still owned by the card */
	for (; hmp->cur_tx - hmp->dirty_tx > 0; hmp->dirty_tx++) {
		int entry = hmp->dirty_tx % TX_RING_SIZE;
		struct sk_buff *skb;

		if (hmp->tx_ring[entry].status_n_length & cpu_to_le32(DescOwn))
			break;
		/* Free the original skb. */
		skb = hmp->tx_skbuff[entry];
		if (skb) {
			dma_unmap_single(&hmp->pci_dev->dev,
					 leXX_to_cpu(hmp->tx_ring[entry].addr),
					 skb->len, DMA_TO_DEVICE);
			dev_kfree_skb(skb);
			hmp->tx_skbuff[entry] = NULL;
		}
		hmp->tx_ring[entry].status_n_length = 0;
		if (entry >= TX_RING_SIZE-1)
			hmp->tx_ring[TX_RING_SIZE-1].status_n_length |=
				cpu_to_le32(DescEndRing);
		dev->stats.tx_packets++;
	}

	return 0;
}

static void hamachi_timer(struct timer_list *t)
{
	struct hamachi_private *hmp = from_timer(hmp, t, timer);
	struct net_device *dev = hmp->mii_if.dev;
	void __iomem *ioaddr = hmp->base;
	int next_tick = 10*HZ;

	if (hamachi_debug > 2) {
		printk(KERN_INFO "%s: Hamachi Autonegotiation status %4.4x, LPA "
			   "%4.4x.\n", dev->name, readw(ioaddr + ANStatus),
			   readw(ioaddr + ANLinkPartnerAbility));
		printk(KERN_INFO "%s: Autonegotiation regs %4.4x %4.4x %4.4x "
		       "%4.4x %4.4x %4.4x.\n", dev->name,
		       readw(ioaddr + 0x0e0),
		       readw(ioaddr + 0x0e2),
		       readw(ioaddr + 0x0e4),
		       readw(ioaddr + 0x0e6),
		       readw(ioaddr + 0x0e8),
		       readw(ioaddr + 0x0eA));
	}
	/* We could do something here... nah. */
	hmp->timer.expires = RUN_AT(next_tick);
	add_timer(&hmp->timer);
}

static void hamachi_tx_timeout(struct net_device *dev, unsigned int txqueue)
{
	int i;
	struct hamachi_private *hmp = netdev_priv(dev);
	void __iomem *ioaddr = hmp->base;

	printk(KERN_WARNING "%s: Hamachi transmit timed out, status %8.8x,"
		   " resetting...\n", dev->name, (int)readw(ioaddr + TxStatus));

	{
		printk(KERN_DEBUG "  Rx ring %p: ", hmp->rx_ring);
		for (i = 0; i < RX_RING_SIZE; i++)
			printk(KERN_CONT " %8.8x",
			       le32_to_cpu(hmp->rx_ring[i].status_n_length));
		printk(KERN_CONT "\n");
		printk(KERN_DEBUG"  Tx ring %p: ", hmp->tx_ring);
		for (i = 0; i < TX_RING_SIZE; i++)
			printk(KERN_CONT " %4.4x",
			       le32_to_cpu(hmp->tx_ring[i].status_n_length));
		printk(KERN_CONT "\n");
	}

	/* Reinit the hardware and make sure the Rx and Tx processes
		are up and running.
	 */
	dev->if_port = 0;
	/* The right way to do Reset. -KDU
	 *		-Clear OWN bit in all Rx/Tx descriptors
	 *		-Wait 50 uS for channels to go idle
	 *		-Turn off MAC receiver
	 *		-Issue Reset
	 */

	for (i = 0; i < RX_RING_SIZE; i++)
		hmp->rx_ring[i].status_n_length &= cpu_to_le32(~DescOwn);

	/* Presume that all packets in the Tx queue are gone if we have to
	 * re-init the hardware.
	 */
	for (i = 0; i < TX_RING_SIZE; i++){
		struct sk_buff *skb;

		if (i >= TX_RING_SIZE - 1)
			hmp->tx_ring[i].status_n_length =
				cpu_to_le32(DescEndRing) |
				(hmp->tx_ring[i].status_n_length &
				 cpu_to_le32(0x0000ffff));
		else
			hmp->tx_ring[i].status_n_length &= cpu_to_le32(0x0000ffff);
		skb = hmp->tx_skbuff[i];
		if (skb){
			dma_unmap_single(&hmp->pci_dev->dev,
					 leXX_to_cpu(hmp->tx_ring[i].addr),
					 skb->len, DMA_TO_DEVICE);
			dev_kfree_skb(skb);
			hmp->tx_skbuff[i] = NULL;
		}
	}

	udelay(60); /* Sleep 60 us just for safety sake */
	writew(0x0002, ioaddr + RxCmd); /* STOP Rx */

	writeb(0x01, ioaddr + ChipReset);  /* Reinit the hardware */

	hmp->tx_full = 0;
	hmp->cur_rx = hmp->cur_tx = 0;
	hmp->dirty_rx = hmp->dirty_tx = 0;
	/* Rx packets are also presumed lost; however, we need to make sure a
	 * ring of buffers is in tact. -KDU
	 */
	for (i = 0; i < RX_RING_SIZE; i++){
		struct sk_buff *skb = hmp->rx_skbuff[i];

		if (skb){
			dma_unmap_single(&hmp->pci_dev->dev,
					 leXX_to_cpu(hmp->rx_ring[i].addr),
					 hmp->rx_buf_sz, DMA_FROM_DEVICE);
			dev_kfree_skb(skb);
			hmp->rx_skbuff[i] = NULL;
		}
	}
	/* Fill in the Rx buffers.  Handle allocation failure gracefully. */
	for (i = 0; i < RX_RING_SIZE; i++) {
		struct sk_buff *skb;

		skb = netdev_alloc_skb_ip_align(dev, hmp->rx_buf_sz);
		hmp->rx_skbuff[i] = skb;
		if (skb == NULL)
			break;

		hmp->rx_ring[i].addr = cpu_to_leXX(dma_map_single(&hmp->pci_dev->dev,
								  skb->data,
								  hmp->rx_buf_sz,
								  DMA_FROM_DEVICE));
		hmp->rx_ring[i].status_n_length = cpu_to_le32(DescOwn |
			DescEndPacket | DescIntr | (hmp->rx_buf_sz - 2));
	}
	hmp->dirty_rx = (unsigned int)(i - RX_RING_SIZE);
	/* Mark the last entry as wrapping the ring. */
	hmp->rx_ring[RX_RING_SIZE-1].status_n_length |= cpu_to_le32(DescEndRing);

	/* Trigger an immediate transmit demand. */
	netif_trans_update(dev); /* prevent tx timeout */
	dev->stats.tx_errors++;

	/* Restart the chip's Tx/Rx processes . */
	writew(0x0002, ioaddr + TxCmd); /* STOP Tx */
	writew(0x0001, ioaddr + TxCmd); /* START Tx */
	writew(0x0001, ioaddr + RxCmd); /* START Rx */

	netif_wake_queue(dev);
}


/* Initialize the Rx and Tx rings, along with various 'dev' bits. */
static void hamachi_init_ring(struct net_device *dev)
{
	struct hamachi_private *hmp = netdev_priv(dev);
	int i;

	hmp->tx_full = 0;
	hmp->cur_rx = hmp->cur_tx = 0;
	hmp->dirty_rx = hmp->dirty_tx = 0;

	/* +26 gets the maximum ethernet encapsulation, +7 & ~7 because the
	 * card needs room to do 8 byte alignment, +2 so we can reserve
	 * the first 2 bytes, and +16 gets room for the status word from the
	 * card.  -KDU
	 */
	hmp->rx_buf_sz = (dev->mtu <= 1492 ? PKT_BUF_SZ :
		(((dev->mtu+26+7) & ~7) + 16));

	/* Initialize all Rx descriptors. */
	for (i = 0; i < RX_RING_SIZE; i++) {
		hmp->rx_ring[i].status_n_length = 0;
		hmp->rx_skbuff[i] = NULL;
	}
	/* Fill in the Rx buffers.  Handle allocation failure gracefully. */
	for (i = 0; i < RX_RING_SIZE; i++) {
		struct sk_buff *skb = netdev_alloc_skb(dev, hmp->rx_buf_sz + 2);
		hmp->rx_skbuff[i] = skb;
		if (skb == NULL)
			break;
		skb_reserve(skb, 2); /* 16 byte align the IP header. */
		hmp->rx_ring[i].addr = cpu_to_leXX(dma_map_single(&hmp->pci_dev->dev,
								  skb->data,
								  hmp->rx_buf_sz,
								  DMA_FROM_DEVICE));
		/* -2 because it doesn't REALLY have that first 2 bytes -KDU */
		hmp->rx_ring[i].status_n_length = cpu_to_le32(DescOwn |
			DescEndPacket | DescIntr | (hmp->rx_buf_sz -2));
	}
	hmp->dirty_rx = (unsigned int)(i - RX_RING_SIZE);
	hmp->rx_ring[RX_RING_SIZE-1].status_n_length |= cpu_to_le32(DescEndRing);

	for (i = 0; i < TX_RING_SIZE; i++) {
		hmp->tx_skbuff[i] = NULL;
		hmp->tx_ring[i].status_n_length = 0;
	}
	/* Mark the last entry of the ring */
	hmp->tx_ring[TX_RING_SIZE-1].status_n_length |= cpu_to_le32(DescEndRing);
}


static netdev_tx_t hamachi_start_xmit(struct sk_buff *skb,
				      struct net_device *dev)
{
	struct hamachi_private *hmp = netdev_priv(dev);
	unsigned entry;
	u16 status;

	/* Ok, now make sure that the queue has space before trying to
		add another skbuff.  if we return non-zero the scheduler
		should interpret this as a queue full and requeue the buffer
		for later.
	 */
	if (hmp->tx_full) {
		/* We should NEVER reach this point -KDU */
		printk(KERN_WARNING "%s: Hamachi transmit queue full at slot %d.\n",dev->name, hmp->cur_tx);

		/* Wake the potentially-idle transmit channel. */
		/* If we don't need to read status, DON'T -KDU */
		status=readw(hmp->base + TxStatus);
		if( !(status & 0x0001) || (status & 0x0002))
			writew(0x0001, hmp->base + TxCmd);
		return NETDEV_TX_BUSY;
	}

	/* Caution: the write order is important here, set the field
	   with the "ownership" bits last. */

	/* Calculate the next Tx descriptor entry. */
	entry = hmp->cur_tx % TX_RING_SIZE;

	hmp->tx_skbuff[entry] = skb;

	hmp->tx_ring[entry].addr = cpu_to_leXX(dma_map_single(&hmp->pci_dev->dev,
							      skb->data,
							      skb->len,
							      DMA_TO_DEVICE));

	/* Hmmmm, could probably put a DescIntr on these, but the way
		the driver is currently coded makes Tx interrupts unnecessary
		since the clearing of the Tx ring is handled by the start_xmit
		routine.  This organization helps mitigate the interrupts a
		bit and probably renders the max_tx_latency param useless.

		Update: Putting a DescIntr bit on all of the descriptors and
		mitigating interrupt frequency with the tx_min_pkt parameter. -KDU
	*/
	if (entry >= TX_RING_SIZE-1)		 /* Wrap ring */
		hmp->tx_ring[entry].status_n_length = cpu_to_le32(DescOwn |
			DescEndPacket | DescEndRing | DescIntr | skb->len);
	else
		hmp->tx_ring[entry].status_n_length = cpu_to_le32(DescOwn |
			DescEndPacket | DescIntr | skb->len);
	hmp->cur_tx++;

	/* Non-x86 Todo: explicitly flush cache lines here. */

	/* Wake the potentially-idle transmit channel. */
	/* If we don't need to read status, DON'T -KDU */
	status=readw(hmp->base + TxStatus);
	if( !(status & 0x0001) || (status & 0x0002))
		writew(0x0001, hmp->base + TxCmd);

	/* Immediately before returning, let's clear as many entries as we can. */
	hamachi_tx(dev);

	/* We should kick the bottom half here, since we are not accepting
	 * interrupts with every packet.  i.e. realize that Gigabit ethernet
	 * can transmit faster than ordinary machines can load packets;
	 * hence, any packet that got put off because we were in the transmit
	 * routine should IMMEDIATELY get a chance to be re-queued. -KDU
	 */
	if ((hmp->cur_tx - hmp->dirty_tx) < (TX_RING_SIZE - 4))
		netif_wake_queue(dev);  /* Typical path */
	else {
		hmp->tx_full = 1;
		netif_stop_queue(dev);
	}

	if (hamachi_debug > 4) {
		printk(KERN_DEBUG "%s: Hamachi transmit frame #%d queued in slot %d.\n",
			   dev->name, hmp->cur_tx, entry);
	}
	return NETDEV_TX_OK;
}

/* The interrupt handler does all of the Rx thread work and cleans up
   after the Tx thread. */
static irqreturn_t hamachi_interrupt(int irq, void *dev_instance)
{
	struct net_device *dev = dev_instance;
	struct hamachi_private *hmp = netdev_priv(dev);
	void __iomem *ioaddr = hmp->base;
	long boguscnt = max_interrupt_work;
	int handled = 0;

#ifndef final_version			/* Can never occur. */
	if (dev == NULL) {
		printk (KERN_ERR "hamachi_interrupt(): irq %d for unknown device.\n", irq);
		return IRQ_NONE;
	}
#endif

	spin_lock(&hmp->lock);

	do {
		u32 intr_status = readl(ioaddr + InterruptClear);

		if (hamachi_debug > 4)
			printk(KERN_DEBUG "%s: Hamachi interrupt, status %4.4x.\n",
				   dev->name, intr_status);

		if (intr_status == 0)
			break;

		handled = 1;

		if (intr_status & IntrRxDone)
			hamachi_rx(dev);

		if (intr_status & IntrTxDone){
			/* This code should RARELY need to execute. After all, this is
			 * a gigabit link, it should consume packets as fast as we put
			 * them in AND we clear the Tx ring in hamachi_start_xmit().
			 */
			if (hmp->tx_full){
				for (; hmp->cur_tx - hmp->dirty_tx > 0; hmp->dirty_tx++){
					int entry = hmp->dirty_tx % TX_RING_SIZE;
					struct sk_buff *skb;

					if (hmp->tx_ring[entry].status_n_length & cpu_to_le32(DescOwn))
						break;
					skb = hmp->tx_skbuff[entry];
					/* Free the original skb. */
					if (skb){
						dma_unmap_single(&hmp->pci_dev->dev,
								 leXX_to_cpu(hmp->tx_ring[entry].addr),
								 skb->len,
								 DMA_TO_DEVICE);
						dev_consume_skb_irq(skb);
						hmp->tx_skbuff[entry] = NULL;
					}
					hmp->tx_ring[entry].status_n_length = 0;
					if (entry >= TX_RING_SIZE-1)
						hmp->tx_ring[TX_RING_SIZE-1].status_n_length |=
							cpu_to_le32(DescEndRing);
					dev->stats.tx_packets++;
				}
				if (hmp->cur_tx - hmp->dirty_tx < TX_RING_SIZE - 4){
					/* The ring is no longer full */
					hmp->tx_full = 0;
					netif_wake_queue(dev);
				}
			} else {
				netif_wake_queue(dev);
			}
		}


		/* Abnormal error summary/uncommon events handlers. */
		if (intr_status &
			(IntrTxPCIFault | IntrTxPCIErr | IntrRxPCIFault | IntrRxPCIErr |
			 LinkChange | NegotiationChange | StatsMax))
			hamachi_error(dev, intr_status);

		if (--boguscnt < 0) {
			printk(KERN_WARNING "%s: Too much work at interrupt, status=0x%4.4x.\n",
				   dev->name, intr_status);
			break;
		}
	} while (1);

	if (hamachi_debug > 3)
		printk(KERN_DEBUG "%s: exiting interrupt, status=%#4.4x.\n",
			   dev->name, readl(ioaddr + IntrStatus));

#ifndef final_version
	/* Code that should never be run!  Perhaps remove after testing.. */
	{
		static int stopit = 10;
		if (dev->start == 0  &&  --stopit < 0) {
			printk(KERN_ERR "%s: Emergency stop, looping startup interrupt.\n",
				   dev->name);
			free_irq(irq, dev);
		}
	}
#endif

	spin_unlock(&hmp->lock);
	return IRQ_RETVAL(handled);
}

/* This routine is logically part of the interrupt handler, but separated
   for clarity and better register allocation. */
static int hamachi_rx(struct net_device *dev)
{
	struct hamachi_private *hmp = netdev_priv(dev);
	int entry = hmp->cur_rx % RX_RING_SIZE;
	int boguscnt = (hmp->dirty_rx + RX_RING_SIZE) - hmp->cur_rx;

	if (hamachi_debug > 4) {
		printk(KERN_DEBUG " In hamachi_rx(), entry %d status %4.4x.\n",
			   entry, hmp->rx_ring[entry].status_n_length);
	}

	/* If EOP is set on the next entry, it's a new packet. Send it up. */
	while (1) {
		struct hamachi_desc *desc = &(hmp->rx_ring[entry]);
		u32 desc_status = le32_to_cpu(desc->status_n_length);
		u16 data_size = desc_status;	/* Implicit truncate */
		u8 *buf_addr;
		s32 frame_status;

		if (desc_status & DescOwn)
			break;
		dma_sync_single_for_cpu(&hmp->pci_dev->dev,
					leXX_to_cpu(desc->addr),
					hmp->rx_buf_sz, DMA_FROM_DEVICE);
		buf_addr = (u8 *) hmp->rx_skbuff[entry]->data;
		frame_status = get_unaligned_le32(&(buf_addr[data_size - 12]));
		if (hamachi_debug > 4)
			printk(KERN_DEBUG "  hamachi_rx() status was %8.8x.\n",
				frame_status);
		if (--boguscnt < 0)
			break;
		if ( ! (desc_status & DescEndPacket)) {
			printk(KERN_WARNING "%s: Oversized Ethernet frame spanned "
				   "multiple buffers, entry %#x length %d status %4.4x!\n",
				   dev->name, hmp->cur_rx, data_size, desc_status);
			printk(KERN_WARNING "%s: Oversized Ethernet frame %p vs %p.\n",
				   dev->name, desc, &hmp->rx_ring[hmp->cur_rx % RX_RING_SIZE]);
			printk(KERN_WARNING "%s: Oversized Ethernet frame -- next status %x/%x last status %x.\n",
				   dev->name,
				   le32_to_cpu(hmp->rx_ring[(hmp->cur_rx+1) % RX_RING_SIZE].status_n_length) & 0xffff0000,
				   le32_to_cpu(hmp->rx_ring[(hmp->cur_rx+1) % RX_RING_SIZE].status_n_length) & 0x0000ffff,
				   le32_to_cpu(hmp->rx_ring[(hmp->cur_rx-1) % RX_RING_SIZE].status_n_length));
			dev->stats.rx_length_errors++;
		} /* else  Omit for prototype errata??? */
		if (frame_status & 0x00380000) {
			/* There was an error. */
			if (hamachi_debug > 2)
				printk(KERN_DEBUG "  hamachi_rx() Rx error was %8.8x.\n",
					   frame_status);
			dev->stats.rx_errors++;
			if (frame_status & 0x00600000)
				dev->stats.rx_length_errors++;
			if (frame_status & 0x00080000)
				dev->stats.rx_frame_errors++;
			if (frame_status & 0x00100000)
				dev->stats.rx_crc_errors++;
			if (frame_status < 0)
				dev->stats.rx_dropped++;
		} else {
			struct sk_buff *skb;
			/* Omit CRC */
			u16 pkt_len = (frame_status & 0x07ff) - 4;
#ifdef RX_CHECKSUM
			u32 pfck = *(u32 *) &buf_addr[data_size - 8];
#endif


#ifndef final_version
			if (hamachi_debug > 4)
				printk(KERN_DEBUG "  hamachi_rx() normal Rx pkt length %d"
					   " of %d, bogus_cnt %d.\n",
					   pkt_len, data_size, boguscnt);
			if (hamachi_debug > 5)
				printk(KERN_DEBUG"%s:  rx status %8.8x %8.8x %8.8x %8.8x %8.8x.\n",
					   dev->name,
					   *(s32*)&(buf_addr[data_size - 20]),
					   *(s32*)&(buf_addr[data_size - 16]),
					   *(s32*)&(buf_addr[data_size - 12]),
					   *(s32*)&(buf_addr[data_size - 8]),
					   *(s32*)&(buf_addr[data_size - 4]));
#endif
			/* Check if the packet is long enough to accept without copying
			   to a minimally-sized skbuff. */
			if (pkt_len < rx_copybreak &&
			    (skb = netdev_alloc_skb(dev, pkt_len + 2)) != NULL) {
#ifdef RX_CHECKSUM
				printk(KERN_ERR "%s: rx_copybreak non-zero "
				  "not good with RX_CHECKSUM\n", dev->name);
#endif
				skb_reserve(skb, 2);	/* 16 byte align the IP header */
				dma_sync_single_for_cpu(&hmp->pci_dev->dev,
							leXX_to_cpu(hmp->rx_ring[entry].addr),
							hmp->rx_buf_sz,
							DMA_FROM_DEVICE);
				/* Call copy + cksum if available. */
#if 1 || USE_IP_COPYSUM
				skb_copy_to_linear_data(skb,
					hmp->rx_skbuff[entry]->data, pkt_len);
				skb_put(skb, pkt_len);
#else
				skb_put_data(skb, hmp->rx_ring_dma
					     + entry*sizeof(*desc), pkt_len);
#endif
				dma_sync_single_for_device(&hmp->pci_dev->dev,
							   leXX_to_cpu(hmp->rx_ring[entry].addr),
							   hmp->rx_buf_sz,
							   DMA_FROM_DEVICE);
			} else {
				dma_unmap_single(&hmp->pci_dev->dev,
						 leXX_to_cpu(hmp->rx_ring[entry].addr),
						 hmp->rx_buf_sz,
						 DMA_FROM_DEVICE);
				skb_put(skb = hmp->rx_skbuff[entry], pkt_len);
				hmp->rx_skbuff[entry] = NULL;
			}
			skb->protocol = eth_type_trans(skb, dev);


#ifdef RX_CHECKSUM
			/* TCP or UDP on ipv4, DIX encoding */
			if (pfck>>24 == 0x91 || pfck>>24 == 0x51) {
				struct iphdr *ih = (struct iphdr *) skb->data;
				/* Check that IP packet is at least 46 bytes, otherwise,
				 * there may be pad bytes included in the hardware checksum.
				 * This wouldn't happen if everyone padded with 0.
				 */
				if (ntohs(ih->tot_len) >= 46){
					/* don't worry about frags */
					if (!(ih->frag_off & cpu_to_be16(IP_MF|IP_OFFSET))) {
						u32 inv = *(u32 *) &buf_addr[data_size - 16];
						u32 *p = (u32 *) &buf_addr[data_size - 20];
						register u32 crc, p_r, p_r1;

						if (inv & 4) {
							inv &= ~4;
							--p;
						}
						p_r = *p;
						p_r1 = *(p-1);
						switch (inv) {
							case 0:
								crc = (p_r & 0xffff) + (p_r >> 16);
								break;
							case 1:
								crc = (p_r >> 16) + (p_r & 0xffff)
									+ (p_r1 >> 16 & 0xff00);
								break;
							case 2:
								crc = p_r + (p_r1 >> 16);
								break;
							case 3:
								crc = p_r + (p_r1 & 0xff00) + (p_r1 >> 16);
								break;
							default:	/*NOTREACHED*/ crc = 0;
						}
						if (crc & 0xffff0000) {
							crc &= 0xffff;
							++crc;
						}
						/* tcp/udp will add in pseudo */
						skb->csum = ntohs(pfck & 0xffff);
						if (skb->csum > crc)
							skb->csum -= crc;
						else
							skb->csum += (~crc & 0xffff);
						/*
						* could do the pseudo myself and return
						* CHECKSUM_UNNECESSARY
						*/
						skb->ip_summed = CHECKSUM_COMPLETE;
					}
				}
			}
#endif  /* RX_CHECKSUM */

			netif_rx(skb);
			dev->stats.rx_packets++;
		}
		entry = (++hmp->cur_rx) % RX_RING_SIZE;
	}

	/* Refill the Rx ring buffers. */
	for (; hmp->cur_rx - hmp->dirty_rx > 0; hmp->dirty_rx++) {
		struct hamachi_desc *desc;

		entry = hmp->dirty_rx % RX_RING_SIZE;
		desc = &(hmp->rx_ring[entry]);
		if (hmp->rx_skbuff[entry] == NULL) {
			struct sk_buff *skb = netdev_alloc_skb(dev, hmp->rx_buf_sz + 2);

			hmp->rx_skbuff[entry] = skb;
			if (skb == NULL)
				break;		/* Better luck next round. */
			skb_reserve(skb, 2);	/* Align IP on 16 byte boundaries */
			desc->addr = cpu_to_leXX(dma_map_single(&hmp->pci_dev->dev,
								skb->data,
								hmp->rx_buf_sz,
								DMA_FROM_DEVICE));
		}
		desc->status_n_length = cpu_to_le32(hmp->rx_buf_sz);
		if (entry >= RX_RING_SIZE-1)
			desc->status_n_length |= cpu_to_le32(DescOwn |
				DescEndPacket | DescEndRing | DescIntr);
		else
			desc->status_n_length |= cpu_to_le32(DescOwn |
				DescEndPacket | DescIntr);
	}

	/* Restart Rx engine if stopped. */
	/* If we don't need to check status, don't. -KDU */
	if (readw(hmp->base + RxStatus) & 0x0002)
		writew(0x0001, hmp->base + RxCmd);

	return 0;
}

/* This is more properly named "uncommon interrupt events", as it covers more
   than just errors. */
static void hamachi_error(struct net_device *dev, int intr_status)
{
	struct hamachi_private *hmp = netdev_priv(dev);
	void __iomem *ioaddr = hmp->base;

	if (intr_status & (LinkChange|NegotiationChange)) {
		if (hamachi_debug > 1)
			printk(KERN_INFO "%s: Link changed: AutoNegotiation Ctrl"
				   " %4.4x, Status %4.4x %4.4x Intr status %4.4x.\n",
				   dev->name, readw(ioaddr + 0x0E0), readw(ioaddr + 0x0E2),
				   readw(ioaddr + ANLinkPartnerAbility),
				   readl(ioaddr + IntrStatus));
		if (readw(ioaddr + ANStatus) & 0x20)
			writeb(0x01, ioaddr + LEDCtrl);
		else
			writeb(0x03, ioaddr + LEDCtrl);
	}
	if (intr_status & StatsMax) {
		hamachi_get_stats(dev);
		/* Read the overflow bits to clear. */
		readl(ioaddr + 0x370);
		readl(ioaddr + 0x3F0);
	}
	if ((intr_status & ~(LinkChange|StatsMax|NegotiationChange|IntrRxDone|IntrTxDone)) &&
	    hamachi_debug)
		printk(KERN_ERR "%s: Something Wicked happened! %4.4x.\n",
		       dev->name, intr_status);
	/* Hmmmmm, it's not clear how to recover from PCI faults. */
	if (intr_status & (IntrTxPCIErr | IntrTxPCIFault))
		dev->stats.tx_fifo_errors++;
	if (intr_status & (IntrRxPCIErr | IntrRxPCIFault))
		dev->stats.rx_fifo_errors++;
}

static int hamachi_close(struct net_device *dev)
{
	struct hamachi_private *hmp = netdev_priv(dev);
	void __iomem *ioaddr = hmp->base;
	struct sk_buff *skb;
	int i;

	netif_stop_queue(dev);

	if (hamachi_debug > 1) {
		printk(KERN_DEBUG "%s: Shutting down ethercard, status was Tx %4.4x Rx %4.4x Int %2.2x.\n",
			   dev->name, readw(ioaddr + TxStatus),
			   readw(ioaddr + RxStatus), readl(ioaddr + IntrStatus));
		printk(KERN_DEBUG "%s: Queue pointers were Tx %d / %d,  Rx %d / %d.\n",
			   dev->name, hmp->cur_tx, hmp->dirty_tx, hmp->cur_rx, hmp->dirty_rx);
	}

	/* Disable interrupts by clearing the interrupt mask. */
	writel(0x0000, ioaddr + InterruptEnable);

	/* Stop the chip's Tx and Rx processes. */
	writel(2, ioaddr + RxCmd);
	writew(2, ioaddr + TxCmd);

#ifdef __i386__
	if (hamachi_debug > 2) {
		printk(KERN_DEBUG "  Tx ring at %8.8x:\n",
			   (int)hmp->tx_ring_dma);
		for (i = 0; i < TX_RING_SIZE; i++)
			printk(KERN_DEBUG " %c #%d desc. %8.8x %8.8x.\n",
				   readl(ioaddr + TxCurPtr) == (long)&hmp->tx_ring[i] ? '>' : ' ',
				   i, hmp->tx_ring[i].status_n_length, hmp->tx_ring[i].addr);
		printk(KERN_DEBUG "  Rx ring %8.8x:\n",
			   (int)hmp->rx_ring_dma);
		for (i = 0; i < RX_RING_SIZE; i++) {
			printk(KERN_DEBUG " %c #%d desc. %4.4x %8.8x\n",
				   readl(ioaddr + RxCurPtr) == (long)&hmp->rx_ring[i] ? '>' : ' ',
				   i, hmp->rx_ring[i].status_n_length, hmp->rx_ring[i].addr);
			if (hamachi_debug > 6) {
				if (*(u8*)hmp->rx_skbuff[i]->data != 0x69) {
					u16 *addr = (u16 *)
						hmp->rx_skbuff[i]->data;
					int j;
					printk(KERN_DEBUG "Addr: ");
					for (j = 0; j < 0x50; j++)
						printk(" %4.4x", addr[j]);
					printk("\n");
				}
			}
		}
	}
#endif /* __i386__ debugging only */

	free_irq(hmp->pci_dev->irq, dev);

	del_timer_sync(&hmp->timer);

	/* Free all the skbuffs in the Rx queue. */
	for (i = 0; i < RX_RING_SIZE; i++) {
		skb = hmp->rx_skbuff[i];
		hmp->rx_ring[i].status_n_length = 0;
		if (skb) {
			dma_unmap_single(&hmp->pci_dev->dev,
					 leXX_to_cpu(hmp->rx_ring[i].addr),
					 hmp->rx_buf_sz, DMA_FROM_DEVICE);
			dev_kfree_skb(skb);
			hmp->rx_skbuff[i] = NULL;
		}
		hmp->rx_ring[i].addr = cpu_to_leXX(0xBADF00D0); /* An invalid address. */
	}
	for (i = 0; i < TX_RING_SIZE; i++) {
		skb = hmp->tx_skbuff[i];
		if (skb) {
			dma_unmap_single(&hmp->pci_dev->dev,
					 leXX_to_cpu(hmp->tx_ring[i].addr),
					 skb->len, DMA_TO_DEVICE);
			dev_kfree_skb(skb);
			hmp->tx_skbuff[i] = NULL;
		}
	}

	writeb(0x00, ioaddr + LEDCtrl);

	return 0;
}

static struct net_device_stats *hamachi_get_stats(struct net_device *dev)
{
	struct hamachi_private *hmp = netdev_priv(dev);
	void __iomem *ioaddr = hmp->base;

	/* We should lock this segment of code for SMP eventually, although
	   the vulnerability window is very small and statistics are
	   non-critical. */
        /* Ok, what goes here?  This appears to be stuck at 21 packets
           according to ifconfig.  It does get incremented in hamachi_tx(),
           so I think I'll comment it out here and see if better things
           happen.
        */
	/* dev->stats.tx_packets	= readl(ioaddr + 0x000); */

	/* Total Uni+Brd+Multi */
	dev->stats.rx_bytes = readl(ioaddr + 0x330);
	/* Total Uni+Brd+Multi */
	dev->stats.tx_bytes = readl(ioaddr + 0x3B0);
	/* Multicast Rx */
	dev->stats.multicast = readl(ioaddr + 0x320);

	/* Over+Undersized */
	dev->stats.rx_length_errors = readl(ioaddr + 0x368);
	/* Jabber */
	dev->stats.rx_over_errors = readl(ioaddr + 0x35C);
	/* Jabber */
	dev->stats.rx_crc_errors = readl(ioaddr + 0x360);
	/* Symbol Errs */
	dev->stats.rx_frame_errors = readl(ioaddr + 0x364);
	/* Dropped */
	dev->stats.rx_missed_errors = readl(ioaddr + 0x36C);

	return &dev->stats;
}

static void set_rx_mode(struct net_device *dev)
{
	struct hamachi_private *hmp = netdev_priv(dev);
	void __iomem *ioaddr = hmp->base;

	if (dev->flags & IFF_PROMISC) {			/* Set promiscuous. */
		writew(0x000F, ioaddr + AddrMode);
	} else if ((netdev_mc_count(dev) > 63) || (dev->flags & IFF_ALLMULTI)) {
		/* Too many to match, or accept all multicasts. */
		writew(0x000B, ioaddr + AddrMode);
	} else if (!netdev_mc_empty(dev)) { /* Must use the CAM filter. */
		struct netdev_hw_addr *ha;
		int i = 0;

		netdev_for_each_mc_addr(ha, dev) {
			writel(*(u32 *)(ha->addr), ioaddr + 0x100 + i*8);
			writel(0x20000 | (*(u16 *)&ha->addr[4]),
				   ioaddr + 0x104 + i*8);
			i++;
		}
		/* Clear remaining entries. */
		for (; i < 64; i++)
			writel(0, ioaddr + 0x104 + i*8);
		writew(0x0003, ioaddr + AddrMode);
	} else {					/* Normal, unicast/broadcast-only mode. */
		writew(0x0001, ioaddr + AddrMode);
	}
}

static int check_if_running(struct net_device *dev)
{
	if (!netif_running(dev))
		return -EINVAL;
	return 0;
}

static void hamachi_get_drvinfo(struct net_device *dev, struct ethtool_drvinfo *info)
{
	struct hamachi_private *np = netdev_priv(dev);

	strlcpy(info->driver, DRV_NAME, sizeof(info->driver));
	strlcpy(info->version, DRV_VERSION, sizeof(info->version));
	strlcpy(info->bus_info, pci_name(np->pci_dev), sizeof(info->bus_info));
}

static int hamachi_get_link_ksettings(struct net_device *dev,
				      struct ethtool_link_ksettings *cmd)
{
	struct hamachi_private *np = netdev_priv(dev);
	spin_lock_irq(&np->lock);
	mii_ethtool_get_link_ksettings(&np->mii_if, cmd);
	spin_unlock_irq(&np->lock);
	return 0;
}

static int hamachi_set_link_ksettings(struct net_device *dev,
				      const struct ethtool_link_ksettings *cmd)
{
	struct hamachi_private *np = netdev_priv(dev);
	int res;
	spin_lock_irq(&np->lock);
	res = mii_ethtool_set_link_ksettings(&np->mii_if, cmd);
	spin_unlock_irq(&np->lock);
	return res;
}

static int hamachi_nway_reset(struct net_device *dev)
{
	struct hamachi_private *np = netdev_priv(dev);
	return mii_nway_restart(&np->mii_if);
}

static u32 hamachi_get_link(struct net_device *dev)
{
	struct hamachi_private *np = netdev_priv(dev);
	return mii_link_ok(&np->mii_if);
}

static const struct ethtool_ops ethtool_ops = {
	.begin = check_if_running,
	.get_drvinfo = hamachi_get_drvinfo,
	.nway_reset = hamachi_nway_reset,
	.get_link = hamachi_get_link,
	.get_link_ksettings = hamachi_get_link_ksettings,
	.set_link_ksettings = hamachi_set_link_ksettings,
};

static const struct ethtool_ops ethtool_ops_no_mii = {
	.begin = check_if_running,
	.get_drvinfo = hamachi_get_drvinfo,
};

/* private ioctl: set rx,tx intr params */
static int hamachi_siocdevprivate(struct net_device *dev, struct ifreq *rq,
				  void __user *data, int cmd)
{
	struct hamachi_private *np = netdev_priv(dev);
	u32 *d = (u32 *)&rq->ifr_ifru;

	if (!netif_running(dev))
		return -EINVAL;

	if (cmd != SIOCDEVPRIVATE + 3)
		return -EOPNOTSUPP;

	/* Should add this check here or an ordinary user can do nasty
	 * things. -KDU
	 *
	 * TODO: Shut down the Rx and Tx engines while doing this.
	 */
	if (!capable(CAP_NET_ADMIN))
		return -EPERM;
	writel(d[0], np->base + TxIntrCtrl);
	writel(d[1], np->base + RxIntrCtrl);
	printk(KERN_NOTICE "%s: tx %08x, rx %08x intr\n", dev->name,
	       (u32)readl(np->base + TxIntrCtrl),
	       (u32)readl(np->base + RxIntrCtrl));

	return 0;
}

static int hamachi_ioctl(struct net_device *dev, struct ifreq *rq, int cmd)
{
	struct hamachi_private *np = netdev_priv(dev);
	struct mii_ioctl_data *data = if_mii(rq);
	int rc;

	if (!netif_running(dev))
		return -EINVAL;

	spin_lock_irq(&np->lock);
	rc = generic_mii_ioctl(&np->mii_if, data, cmd, NULL);
	spin_unlock_irq(&np->lock);

	return rc;
}


static void hamachi_remove_one(struct pci_dev *pdev)
{
	struct net_device *dev = pci_get_drvdata(pdev);

	if (dev) {
		struct hamachi_private *hmp = netdev_priv(dev);

		dma_free_coherent(&pdev->dev, RX_TOTAL_SIZE, hmp->rx_ring,
				  hmp->rx_ring_dma);
		dma_free_coherent(&pdev->dev, TX_TOTAL_SIZE, hmp->tx_ring,
				  hmp->tx_ring_dma);
		unregister_netdev(dev);
		iounmap(hmp->base);
		free_netdev(dev);
		pci_release_regions(pdev);
	}
}

static const struct pci_device_id hamachi_pci_tbl[] = {
	{ 0x1318, 0x0911, PCI_ANY_ID, PCI_ANY_ID, },
	{ 0, }
};
MODULE_DEVICE_TABLE(pci, hamachi_pci_tbl);

static struct pci_driver hamachi_driver = {
	.name		= DRV_NAME,
	.id_table	= hamachi_pci_tbl,
	.probe		= hamachi_init_one,
	.remove		= hamachi_remove_one,
};

static int __init hamachi_init (void)
{
/* when a module, this is printed whether or not devices are found in probe */
#ifdef MODULE
	printk(version);
#endif
	return pci_register_driver(&hamachi_driver);
}

static void __exit hamachi_exit (void)
{
	pci_unregister_driver(&hamachi_driver);
}


module_init(hamachi_init);
module_exit(hamachi_exit);