aboutsummaryrefslogblamecommitdiffstats
path: root/drivers/net/acenic.c
blob: b8953de5664a4cdf091374af819525518414a6cd (plain) (tree)
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
1172
1173
1174
1175
1176
1177
1178
1179
1180
1181
1182
1183
1184
1185
1186
1187
1188
1189
1190
1191
1192
1193
1194
1195
1196
1197
1198
1199
1200
1201
1202
1203
1204
1205
1206
1207
1208
1209
1210
1211
1212
1213
1214
1215
1216
1217
1218
1219
1220
1221
1222
1223
1224
1225
1226
1227
1228
1229
1230
1231
1232
1233
1234
1235
1236
1237
1238
1239
1240
1241
1242
1243
1244
1245
1246
1247
1248
1249
1250
1251
1252
1253
1254
1255
1256
1257
1258
1259
1260
1261
1262
1263
1264
1265
1266
1267
1268
1269
1270
1271
1272
1273
1274
1275
1276
1277
1278
1279
1280
1281
1282
1283
1284
1285
1286
1287
1288
1289
1290
1291
1292
1293
1294
1295
1296
1297
1298
1299
1300
1301
1302
1303
1304
1305
1306
1307
1308
1309
1310
1311
1312
1313
1314
1315
1316
1317
1318
1319
1320
1321
1322
1323
1324
1325
1326
1327
1328
1329
1330
1331
1332
1333
1334
1335
1336
1337
1338
1339
1340
1341
1342
1343
1344
1345
1346
1347
1348
1349
1350
1351
1352
1353
1354
1355
1356
1357
1358
1359
1360
1361
1362
1363
1364
1365
1366
1367
1368
1369
1370
1371
1372
1373
1374
1375
1376
1377
1378
1379
1380
1381
1382
1383
1384
1385
1386
1387
1388
1389
1390
1391
1392
1393
1394
1395
1396
1397
1398
1399
1400
1401
1402
1403
1404
1405
1406
1407
1408
1409
1410
1411
1412
1413
1414
1415
1416
1417
1418
1419
1420
1421
1422
1423
1424
1425
1426
1427
1428
1429
1430
1431
1432
1433
1434
1435
1436
1437
1438
1439
1440
1441
1442
1443
1444
1445
1446
1447
1448
1449
1450
1451
1452
1453
1454
1455
1456
1457
1458
1459
1460
1461
1462
1463
1464
1465
1466
1467
1468
1469
1470
1471
1472
1473
1474
1475
1476
1477
1478
1479
1480
1481
1482
1483
1484
1485
1486
1487
1488
1489
1490
1491
1492
1493
1494
1495
1496
1497
1498
1499
1500
1501
1502
1503
1504
1505
1506
1507
1508
1509
1510
1511
1512
1513
1514
1515
1516
1517
1518
1519
1520
1521
1522
1523
1524
1525
1526
1527
1528
1529
1530
1531
1532
1533
1534
1535
1536
1537
1538
1539
1540
1541
1542
1543
1544
1545
1546
1547
1548
1549
1550
1551
1552
1553
1554
1555
1556
1557
1558
1559
1560
1561
1562
1563
1564
1565
1566
1567
1568
1569
1570
1571
1572
1573
1574
1575
1576
1577
1578
1579
1580
1581
1582
1583
1584
1585
1586
1587
1588
1589
1590
1591
1592
1593
1594
1595
1596
1597
1598
1599
1600
1601
1602
1603
1604
1605
1606
1607
1608
1609
1610
1611
1612
1613
1614
1615
1616
1617
1618
1619
1620
1621
1622
1623
1624
1625
1626
1627
1628
1629
1630
1631
1632
1633
1634
1635
1636
1637
1638
1639
1640
1641
1642
1643
1644
1645
1646
1647
1648
1649
1650
1651
1652
1653
1654
1655
1656
1657
1658
1659
1660
1661
1662
1663
1664
1665
1666
1667
1668
1669
1670
1671
1672
1673
1674
1675
1676
1677
1678
1679
1680
1681
1682
1683
1684
1685
1686
1687
1688
1689
1690
1691
1692
1693
1694
1695
1696
1697
1698
1699
1700
1701
1702
1703
1704
1705
1706
1707
1708
1709
1710
1711
1712
1713
1714
1715
1716
1717
1718
1719
1720
1721
1722
1723
1724
1725
1726
1727
1728
1729
1730
1731
1732
1733
1734
1735
1736
1737
1738
1739
1740
1741
1742
1743
1744
1745
1746
1747
1748
1749
1750
1751
1752
1753
1754
1755
1756
1757
1758
1759
1760
1761
1762
1763
1764
1765
1766
1767
1768
1769
1770
1771
1772
1773
1774
1775
1776
1777
1778
1779
1780
1781
1782
1783
1784
1785
1786
1787
1788
1789
1790
1791
1792
1793
1794
1795
1796
1797
1798
1799
1800
1801
1802
1803
1804
1805
1806
1807
1808
1809
1810
1811
1812
1813
1814
1815
1816
1817
1818
1819
1820
1821
1822
1823
1824
1825
1826
1827
1828
1829
1830
1831
1832
1833
1834
1835
1836
1837
1838
1839
1840
1841
1842
1843
1844
1845
1846
1847
1848
1849
1850
1851
1852
1853
1854
1855
1856
1857
1858
1859
1860
1861
1862
1863
1864
1865
1866
1867
1868
1869
1870
1871
1872
1873
1874
1875
1876
1877
1878
1879
1880
1881
1882
1883
1884
1885
1886
1887
1888
1889
1890
1891
1892
1893
1894
1895
1896
1897
1898
1899
1900
1901
1902
1903
1904
1905
1906
1907
1908
1909
1910
1911
1912
1913
1914
1915
1916
1917
1918
1919
1920
1921
1922
1923
1924
1925
1926
1927
1928
1929
1930
1931
1932
1933
1934
1935
1936
1937
1938
1939
1940
1941
1942
1943
1944
1945
1946
1947
1948
1949
1950
1951
1952
1953
1954
1955
1956
1957
1958
1959
1960
1961
1962
1963
1964
1965
1966
1967
1968
1969
1970
1971
1972
1973
1974
1975
1976
1977
1978
1979
1980
1981
1982
1983
1984
1985
1986
1987
1988
1989
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
2011
2012
2013
2014
2015
2016
2017
2018
2019
2020
2021
2022
2023
2024
2025
2026
2027
2028
2029
2030
2031
2032
2033
2034
2035
2036
2037
2038
2039
2040
2041
2042
2043
2044
2045
2046
2047
2048
2049
2050
2051
2052
2053
2054
2055
2056
2057
2058
2059
2060
2061
2062
2063
2064
2065
2066
2067
2068
2069
2070
2071
2072
2073
2074
2075
2076
2077
2078
2079
2080
2081
2082
2083
2084
2085
2086
2087
2088
2089
2090
2091
2092
2093
2094
2095
2096
2097
2098
2099
2100
2101
2102
2103
2104
2105
2106
2107
2108
2109
2110
2111
2112
2113
2114
2115
2116
2117
2118
2119
2120
2121
2122
2123
2124
2125
2126
2127
2128
2129
2130
2131
2132
2133
2134
2135
2136
2137
2138
2139
2140
2141
2142
2143
2144
2145
2146
2147
2148
2149
2150
2151
2152
2153
2154
2155
2156
2157
2158
2159
2160
2161
2162
2163
2164
2165
2166
2167
2168
2169
2170
2171
2172
2173
2174
2175
2176
2177
2178
2179
2180
2181
2182
2183
2184
2185
2186
2187
2188
2189
2190
2191
2192
2193
2194
2195
2196
2197
2198
2199
2200
2201
2202
2203
2204
2205
2206
2207
2208
2209
2210
2211
2212
2213
2214
2215
2216
2217
2218
2219
2220
2221
2222
2223
2224
2225
2226
2227
2228
2229
2230
2231
2232
2233
2234
2235
2236
2237
2238
2239
2240
2241
2242
2243
2244
2245
2246
2247
2248
2249
2250
2251
2252
2253
2254
2255
2256
2257
2258
2259
2260
2261
2262
2263
2264
2265
2266
2267
2268
2269
2270
2271
2272
2273
2274
2275
2276
2277
2278
2279
2280
2281
2282
2283
2284
2285
2286
2287
2288
2289
2290
2291
2292
2293
2294
2295
2296
2297
2298
2299
2300
2301
2302
2303
2304
2305
2306
2307
2308
2309
2310
2311
2312
2313
2314
2315
2316
2317
2318
2319
2320
2321
2322
2323
2324
2325
2326
2327
2328
2329
2330
2331
2332
2333
2334
2335
2336
2337
2338
2339
2340
2341
2342
2343
2344
2345
2346
2347
2348
2349
2350
2351
2352
2353
2354
2355
2356
2357
2358
2359
2360
2361
2362
2363
2364
2365
2366
2367
2368
2369
2370
2371
2372
2373
2374
2375
2376
2377
2378
2379
2380
2381
2382
2383
2384
2385
2386
2387
2388
2389
2390
2391
2392
2393
2394
2395
2396
2397
2398
2399
2400
2401
2402
2403
2404
2405
2406
2407
2408
2409
2410
2411
2412
2413
2414
2415
2416
2417
2418
2419
2420
2421
2422
2423
2424
2425
2426
2427
2428
2429
2430
2431
2432
2433
2434
2435
2436
2437
2438
2439
2440
2441
2442
2443
2444
2445
2446
2447
2448
2449
2450
2451
2452
2453
2454
2455
2456
2457
2458
2459
2460
2461
2462
2463
2464
2465
2466
2467
2468
2469
2470
2471
2472
2473
2474
2475
2476
2477
2478
2479
2480
2481
2482
2483
2484
2485
2486
2487
2488
2489
2490
2491
2492
2493
2494
2495
2496
2497
2498
2499
2500
2501
2502
2503
2504
2505
2506
2507
2508
2509
2510
2511
2512
2513
2514
2515
2516
2517
2518
2519
2520
2521
2522
2523
2524
2525
2526
2527
2528
2529
2530
2531
2532
2533
2534
2535
2536
2537
2538
2539
2540
2541
2542
2543
2544
2545
2546
2547
2548
2549
2550
2551
2552
2553
2554
2555
2556
2557
2558
2559
2560
2561
2562
2563
2564
2565
2566
2567
2568
2569
2570
2571
2572
2573
2574
2575
2576
2577
2578
2579
2580
2581
2582
2583
2584
2585
2586
2587
2588
2589
2590
2591
2592
2593
2594
2595
2596
2597
2598
2599
2600
2601
2602
2603
2604
2605
2606
2607
2608
2609
2610
2611
2612
2613
2614
2615
2616
2617
2618
2619
2620
2621
2622
2623
2624
2625
2626
2627
2628
2629
2630
2631
2632
2633
2634
2635
2636
2637
2638
2639
2640
2641
2642
2643
2644
2645
2646
2647
2648
2649
2650
2651
2652
2653
2654
2655
2656
2657
2658
2659
2660
2661
2662
2663
2664
2665
2666
2667
2668
2669
2670
2671
2672
2673
2674
2675
2676
2677
2678
2679
2680
2681
2682
2683
2684
2685
2686
2687
2688
2689
2690
2691
2692
2693
2694
2695
2696
2697
2698
2699
2700
2701
2702
2703
2704
2705
2706
2707
2708
2709
2710
2711
2712
2713
2714
2715
2716
2717
2718
2719
2720
2721
2722
2723
2724
2725
2726
2727
2728
2729
2730
2731
2732
2733
2734
2735
2736
2737
2738
2739
2740
2741
2742
2743
2744
2745
2746
2747
2748
2749
2750
2751
2752
2753
2754
2755
2756
2757
2758
2759
2760
2761
2762
2763
2764
2765
2766
2767
2768
2769
2770
2771
2772
2773
2774
2775
2776
2777
2778
2779
2780
2781
2782
2783
2784
2785
2786
2787
2788
2789
2790
2791
2792
2793
2794
2795
2796
2797
2798
2799
2800
2801
2802
2803
2804
2805
2806
2807
2808
2809
2810
2811
2812
2813
2814
2815
2816
2817
2818
2819
2820
2821
2822
2823
2824
2825
2826
2827
2828
2829
2830
2831
2832
2833
2834
2835
2836
2837
2838
2839
2840
2841
2842
2843
2844
2845
2846
2847
2848
2849
2850
2851
2852
2853
2854
2855
2856
2857
2858
2859
2860
2861
2862
2863
2864
2865
2866
2867
2868
2869
2870
2871
2872
2873
2874
2875
2876
2877
2878
2879
2880
2881
2882
2883
2884
2885
2886
2887
2888
2889
2890
2891
2892
2893
2894
2895
2896
2897
2898
2899
2900
2901
2902
2903
2904
2905
2906
2907
2908
2909
2910
2911
2912
2913
2914
2915
2916
2917
2918
2919
2920
2921
2922
2923
2924
2925
2926
2927
2928
2929
2930
2931
2932
2933
2934
2935
2936
2937
2938
2939
2940
2941
2942
2943
2944
2945
2946
2947
2948
2949
2950
2951
2952
2953
2954
2955
2956
2957
2958
2959
2960
2961
2962
2963
2964
2965
2966
2967
2968
2969
2970
2971
2972
2973
2974
2975
2976
2977
2978
2979
2980
2981
2982
2983
2984
2985
2986
2987
2988
2989
2990
2991
2992
2993
2994
2995
2996
2997
2998
2999
3000
3001
3002
3003
3004
3005
3006
3007
3008
3009
3010
3011
3012
3013
3014
3015
3016
3017
3018
3019
3020
3021
3022
3023
3024
3025
3026
3027
3028
3029
3030
3031
3032
3033
3034
3035
3036
3037
3038
3039
3040
3041
3042
3043
3044
3045
3046
3047
3048
3049
3050
3051
3052
3053
3054
3055
3056
3057
3058
3059
3060
3061
3062
3063
3064
3065
3066
3067
3068
3069
3070
3071
3072
3073
3074
3075
3076
3077
3078
3079
3080
3081
3082
3083
3084
3085
3086
3087
3088
3089
3090
3091
3092
3093
3094
3095
3096
3097
3098
3099
3100
3101
3102
3103
3104
3105
3106
3107
3108
3109
3110
3111
3112
3113
3114
3115
3116
3117
3118
3119
3120
3121
3122
3123
3124
3125
3126
3127
3128
3129
3130
3131
3132
3133
3134
3135
3136
3137
3138
3139
3140
3141
3142
3143
3144
3145
3146
3147
3148
3149
3150
3151
3152
3153
3154
3155
3156
3157
3158
3159
3160
3161
3162
3163
3164
3165
3166
3167
3168
3169
3170
3171
3172
3173
3174
3175
3176
3177
3178
3179
3180
3181
3182
3183
3184
3185
3186
3187
3188
3189
3190
3191
3192
3193
3194
3195
3196
3197
3198
3199
3200
3201
3202
3203
3204
3205
3206
3207
3208
3209
3210
3211
3212
3213
3214
3215
3216
3217
3218
3219
3220
3221
3222
3223
3224
3225
3226
3227
3228
3229
3230
3231
3232
3233
3234
3235
3236
3237
3238
3239
3240
3241
3242
3243
3244
3245
3246
3247
3248
3249
3250
3251
3252
3253
3254
3255
3256
3257
3258
3259
3260
3261
3262
3263
3264
3265
3266
3267
3268
3269
3270


























































                                                                             
                              










































































































































































































































































































































































































































































































































































































































































































































































































































                                                                                                              
                             



































































































































































































































































































                                                                                
                                                      
                                      
                                                             

















































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































                                                                                                                                                                                                                                          
/*
 * acenic.c: Linux driver for the Alteon AceNIC Gigabit Ethernet card
 *           and other Tigon based cards.
 *
 * Copyright 1998-2002 by Jes Sorensen, <jes@trained-monkey.org>.
 *
 * Thanks to Alteon and 3Com for providing hardware and documentation
 * enabling me to write this driver.
 *
 * A mailing list for discussing the use of this driver has been
 * setup, please subscribe to the lists if you have any questions
 * about the driver. Send mail to linux-acenic-help@sunsite.auc.dk to
 * see how to subscribe.
 *
 * This program is free software; you can redistribute it and/or modify
 * it under the terms of the GNU General Public License as published by
 * the Free Software Foundation; either version 2 of the License, or
 * (at your option) any later version.
 *
 * Additional credits:
 *   Pete Wyckoff <wyckoff@ca.sandia.gov>: Initial Linux/Alpha and trace
 *       dump support. The trace dump support has not been
 *       integrated yet however.
 *   Troy Benjegerdes: Big Endian (PPC) patches.
 *   Nate Stahl: Better out of memory handling and stats support.
 *   Aman Singla: Nasty race between interrupt handler and tx code dealing
 *                with 'testing the tx_ret_csm and setting tx_full'
 *   David S. Miller <davem@redhat.com>: conversion to new PCI dma mapping
 *                                       infrastructure and Sparc support
 *   Pierrick Pinasseau (CERN): For lending me an Ultra 5 to test the
 *                              driver under Linux/Sparc64
 *   Matt Domsch <Matt_Domsch@dell.com>: Detect Alteon 1000baseT cards
 *                                       ETHTOOL_GDRVINFO support
 *   Chip Salzenberg <chip@valinux.com>: Fix race condition between tx
 *                                       handler and close() cleanup.
 *   Ken Aaker <kdaaker@rchland.vnet.ibm.com>: Correct check for whether
 *                                       memory mapped IO is enabled to
 *                                       make the driver work on RS/6000.
 *   Takayoshi Kouchi <kouchi@hpc.bs1.fc.nec.co.jp>: Identifying problem
 *                                       where the driver would disable
 *                                       bus master mode if it had to disable
 *                                       write and invalidate.
 *   Stephen Hack <stephen_hack@hp.com>: Fixed ace_set_mac_addr for little
 *                                       endian systems.
 *   Val Henson <vhenson@esscom.com>:    Reset Jumbo skb producer and
 *                                       rx producer index when
 *                                       flushing the Jumbo ring.
 *   Hans Grobler <grobh@sun.ac.za>:     Memory leak fixes in the
 *                                       driver init path.
 *   Grant Grundler <grundler@cup.hp.com>: PCI write posting fixes.
 */

#include <linux/config.h>
#include <linux/module.h>
#include <linux/moduleparam.h>
#include <linux/version.h>
#include <linux/types.h>
#include <linux/errno.h>
#include <linux/ioport.h>
#include <linux/pci.h>
#include <linux/dma-mapping.h>
#include <linux/kernel.h>
#include <linux/netdevice.h>
#include <linux/etherdevice.h>
#include <linux/skbuff.h>
#include <linux/init.h>
#include <linux/delay.h>
#include <linux/mm.h>
#include <linux/highmem.h>
#include <linux/sockios.h>

#if defined(CONFIG_VLAN_8021Q) || defined(CONFIG_VLAN_8021Q_MODULE)
#include <linux/if_vlan.h>
#endif

#ifdef SIOCETHTOOL
#include <linux/ethtool.h>
#endif

#include <net/sock.h>
#include <net/ip.h>

#include <asm/system.h>
#include <asm/io.h>
#include <asm/irq.h>
#include <asm/byteorder.h>
#include <asm/uaccess.h>


#define DRV_NAME "acenic"

#undef INDEX_DEBUG

#ifdef CONFIG_ACENIC_OMIT_TIGON_I
#define ACE_IS_TIGON_I(ap)	0
#define ACE_TX_RING_ENTRIES(ap)	MAX_TX_RING_ENTRIES
#else
#define ACE_IS_TIGON_I(ap)	(ap->version == 1)
#define ACE_TX_RING_ENTRIES(ap)	ap->tx_ring_entries
#endif

#ifndef PCI_VENDOR_ID_ALTEON
#define PCI_VENDOR_ID_ALTEON		0x12ae	
#endif
#ifndef PCI_DEVICE_ID_ALTEON_ACENIC_FIBRE
#define PCI_DEVICE_ID_ALTEON_ACENIC_FIBRE  0x0001
#define PCI_DEVICE_ID_ALTEON_ACENIC_COPPER 0x0002
#endif
#ifndef PCI_DEVICE_ID_3COM_3C985
#define PCI_DEVICE_ID_3COM_3C985	0x0001
#endif
#ifndef PCI_VENDOR_ID_NETGEAR
#define PCI_VENDOR_ID_NETGEAR		0x1385
#define PCI_DEVICE_ID_NETGEAR_GA620	0x620a
#endif
#ifndef PCI_DEVICE_ID_NETGEAR_GA620T
#define PCI_DEVICE_ID_NETGEAR_GA620T	0x630a
#endif


/*
 * Farallon used the DEC vendor ID by mistake and they seem not
 * to care - stinky!
 */
#ifndef PCI_DEVICE_ID_FARALLON_PN9000SX
#define PCI_DEVICE_ID_FARALLON_PN9000SX	0x1a
#endif
#ifndef PCI_DEVICE_ID_FARALLON_PN9100T
#define PCI_DEVICE_ID_FARALLON_PN9100T  0xfa
#endif
#ifndef PCI_VENDOR_ID_SGI
#define PCI_VENDOR_ID_SGI		0x10a9
#endif
#ifndef PCI_DEVICE_ID_SGI_ACENIC
#define PCI_DEVICE_ID_SGI_ACENIC	0x0009
#endif

static struct pci_device_id acenic_pci_tbl[] = {
	{ PCI_VENDOR_ID_ALTEON, PCI_DEVICE_ID_ALTEON_ACENIC_FIBRE,
	  PCI_ANY_ID, PCI_ANY_ID, PCI_CLASS_NETWORK_ETHERNET << 8, 0xffff00, },
	{ PCI_VENDOR_ID_ALTEON, PCI_DEVICE_ID_ALTEON_ACENIC_COPPER,
	  PCI_ANY_ID, PCI_ANY_ID, PCI_CLASS_NETWORK_ETHERNET << 8, 0xffff00, },
	{ PCI_VENDOR_ID_3COM, PCI_DEVICE_ID_3COM_3C985,
	  PCI_ANY_ID, PCI_ANY_ID, PCI_CLASS_NETWORK_ETHERNET << 8, 0xffff00, },
	{ PCI_VENDOR_ID_NETGEAR, PCI_DEVICE_ID_NETGEAR_GA620,
	  PCI_ANY_ID, PCI_ANY_ID, PCI_CLASS_NETWORK_ETHERNET << 8, 0xffff00, },
	{ PCI_VENDOR_ID_NETGEAR, PCI_DEVICE_ID_NETGEAR_GA620T,
	  PCI_ANY_ID, PCI_ANY_ID, PCI_CLASS_NETWORK_ETHERNET << 8, 0xffff00, },
	/*
	 * Farallon used the DEC vendor ID on their cards incorrectly,
	 * then later Alteon's ID.
	 */
	{ PCI_VENDOR_ID_DEC, PCI_DEVICE_ID_FARALLON_PN9000SX,
	  PCI_ANY_ID, PCI_ANY_ID, PCI_CLASS_NETWORK_ETHERNET << 8, 0xffff00, },
	{ PCI_VENDOR_ID_ALTEON, PCI_DEVICE_ID_FARALLON_PN9100T,
	  PCI_ANY_ID, PCI_ANY_ID, PCI_CLASS_NETWORK_ETHERNET << 8, 0xffff00, },
	{ PCI_VENDOR_ID_SGI, PCI_DEVICE_ID_SGI_ACENIC,
	  PCI_ANY_ID, PCI_ANY_ID, PCI_CLASS_NETWORK_ETHERNET << 8, 0xffff00, },
	{ }
};
MODULE_DEVICE_TABLE(pci, acenic_pci_tbl);

#ifndef SET_NETDEV_DEV
#define SET_NETDEV_DEV(net, pdev)	do{} while(0)
#endif

#if LINUX_VERSION_CODE >= 0x2051c
#define ace_sync_irq(irq)	synchronize_irq(irq)
#else
#define ace_sync_irq(irq)	synchronize_irq()
#endif

#ifndef offset_in_page
#define offset_in_page(ptr)	((unsigned long)(ptr) & ~PAGE_MASK)
#endif

#define ACE_MAX_MOD_PARMS	8
#define BOARD_IDX_STATIC	0
#define BOARD_IDX_OVERFLOW	-1

#if (defined(CONFIG_VLAN_8021Q) || defined(CONFIG_VLAN_8021Q_MODULE)) && \
	defined(NETIF_F_HW_VLAN_RX)
#define ACENIC_DO_VLAN		1
#define ACE_RCB_VLAN_FLAG	RCB_FLG_VLAN_ASSIST
#else
#define ACENIC_DO_VLAN		0
#define ACE_RCB_VLAN_FLAG	0
#endif

#include "acenic.h"

/*
 * These must be defined before the firmware is included.
 */
#define MAX_TEXT_LEN	96*1024
#define MAX_RODATA_LEN	8*1024
#define MAX_DATA_LEN	2*1024

#include "acenic_firmware.h"

#ifndef tigon2FwReleaseLocal
#define tigon2FwReleaseLocal 0
#endif

/*
 * This driver currently supports Tigon I and Tigon II based cards
 * including the Alteon AceNIC, the 3Com 3C985[B] and NetGear
 * GA620. The driver should also work on the SGI, DEC and Farallon
 * versions of the card, however I have not been able to test that
 * myself.
 *
 * This card is really neat, it supports receive hardware checksumming
 * and jumbo frames (up to 9000 bytes) and does a lot of work in the
 * firmware. Also the programming interface is quite neat, except for
 * the parts dealing with the i2c eeprom on the card ;-)
 *
 * Using jumbo frames:
 *
 * To enable jumbo frames, simply specify an mtu between 1500 and 9000
 * bytes to ifconfig. Jumbo frames can be enabled or disabled at any time
 * by running `ifconfig eth<X> mtu <MTU>' with <X> being the Ethernet
 * interface number and <MTU> being the MTU value.
 *
 * Module parameters:
 *
 * When compiled as a loadable module, the driver allows for a number
 * of module parameters to be specified. The driver supports the
 * following module parameters:
 *
 *  trace=<val> - Firmware trace level. This requires special traced
 *                firmware to replace the firmware supplied with
 *                the driver - for debugging purposes only.
 *
 *  link=<val>  - Link state. Normally you want to use the default link
 *                parameters set by the driver. This can be used to
 *                override these in case your switch doesn't negotiate
 *                the link properly. Valid values are:
 *         0x0001 - Force half duplex link.
 *         0x0002 - Do not negotiate line speed with the other end.
 *         0x0010 - 10Mbit/sec link.
 *         0x0020 - 100Mbit/sec link.
 *         0x0040 - 1000Mbit/sec link.
 *         0x0100 - Do not negotiate flow control.
 *         0x0200 - Enable RX flow control Y
 *         0x0400 - Enable TX flow control Y (Tigon II NICs only).
 *                Default value is 0x0270, ie. enable link+flow
 *                control negotiation. Negotiating the highest
 *                possible link speed with RX flow control enabled.
 *
 *                When disabling link speed negotiation, only one link
 *                speed is allowed to be specified!
 *
 *  tx_coal_tick=<val> - number of coalescing clock ticks (us) allowed
 *                to wait for more packets to arive before
 *                interrupting the host, from the time the first
 *                packet arrives.
 *
 *  rx_coal_tick=<val> - number of coalescing clock ticks (us) allowed
 *                to wait for more packets to arive in the transmit ring,
 *                before interrupting the host, after transmitting the
 *                first packet in the ring.
 *
 *  max_tx_desc=<val> - maximum number of transmit descriptors
 *                (packets) transmitted before interrupting the host.
 *
 *  max_rx_desc=<val> - maximum number of receive descriptors
 *                (packets) received before interrupting the host.
 *
 *  tx_ratio=<val> - 7 bit value (0 - 63) specifying the split in 64th
 *                increments of the NIC's on board memory to be used for
 *                transmit and receive buffers. For the 1MB NIC app. 800KB
 *                is available, on the 1/2MB NIC app. 300KB is available.
 *                68KB will always be available as a minimum for both
 *                directions. The default value is a 50/50 split.
 *  dis_pci_mem_inval=<val> - disable PCI memory write and invalidate
 *                operations, default (1) is to always disable this as
 *                that is what Alteon does on NT. I have not been able
 *                to measure any real performance differences with
 *                this on my systems. Set <val>=0 if you want to
 *                enable these operations.
 *
 * If you use more than one NIC, specify the parameters for the
 * individual NICs with a comma, ie. trace=0,0x00001fff,0 you want to
 * run tracing on NIC #2 but not on NIC #1 and #3.
 *
 * TODO:
 *
 * - Proper multicast support.
 * - NIC dump support.
 * - More tuning parameters.
 *
 * The mini ring is not used under Linux and I am not sure it makes sense
 * to actually use it.
 *
 * New interrupt handler strategy:
 *
 * The old interrupt handler worked using the traditional method of
 * replacing an skbuff with a new one when a packet arrives. However
 * the rx rings do not need to contain a static number of buffer
 * descriptors, thus it makes sense to move the memory allocation out
 * of the main interrupt handler and do it in a bottom half handler
 * and only allocate new buffers when the number of buffers in the
 * ring is below a certain threshold. In order to avoid starving the
 * NIC under heavy load it is however necessary to force allocation
 * when hitting a minimum threshold. The strategy for alloction is as
 * follows:
 *
 *     RX_LOW_BUF_THRES    - allocate buffers in the bottom half
 *     RX_PANIC_LOW_THRES  - we are very low on buffers, allocate
 *                           the buffers in the interrupt handler
 *     RX_RING_THRES       - maximum number of buffers in the rx ring
 *     RX_MINI_THRES       - maximum number of buffers in the mini ring
 *     RX_JUMBO_THRES      - maximum number of buffers in the jumbo ring
 *
 * One advantagous side effect of this allocation approach is that the
 * entire rx processing can be done without holding any spin lock
 * since the rx rings and registers are totally independent of the tx
 * ring and its registers.  This of course includes the kmalloc's of
 * new skb's. Thus start_xmit can run in parallel with rx processing
 * and the memory allocation on SMP systems.
 *
 * Note that running the skb reallocation in a bottom half opens up
 * another can of races which needs to be handled properly. In
 * particular it can happen that the interrupt handler tries to run
 * the reallocation while the bottom half is either running on another
 * CPU or was interrupted on the same CPU. To get around this the
 * driver uses bitops to prevent the reallocation routines from being
 * reentered.
 *
 * TX handling can also be done without holding any spin lock, wheee
 * this is fun! since tx_ret_csm is only written to by the interrupt
 * handler. The case to be aware of is when shutting down the device
 * and cleaning up where it is necessary to make sure that
 * start_xmit() is not running while this is happening. Well DaveM
 * informs me that this case is already protected against ... bye bye
 * Mr. Spin Lock, it was nice to know you.
 *
 * TX interrupts are now partly disabled so the NIC will only generate
 * TX interrupts for the number of coal ticks, not for the number of
 * TX packets in the queue. This should reduce the number of TX only,
 * ie. when no RX processing is done, interrupts seen.
 */

/*
 * Threshold values for RX buffer allocation - the low water marks for
 * when to start refilling the rings are set to 75% of the ring
 * sizes. It seems to make sense to refill the rings entirely from the
 * intrrupt handler once it gets below the panic threshold, that way
 * we don't risk that the refilling is moved to another CPU when the
 * one running the interrupt handler just got the slab code hot in its
 * cache.
 */
#define RX_RING_SIZE		72
#define RX_MINI_SIZE		64
#define RX_JUMBO_SIZE		48

#define RX_PANIC_STD_THRES	16
#define RX_PANIC_STD_REFILL	(3*RX_PANIC_STD_THRES)/2
#define RX_LOW_STD_THRES	(3*RX_RING_SIZE)/4
#define RX_PANIC_MINI_THRES	12
#define RX_PANIC_MINI_REFILL	(3*RX_PANIC_MINI_THRES)/2
#define RX_LOW_MINI_THRES	(3*RX_MINI_SIZE)/4
#define RX_PANIC_JUMBO_THRES	6
#define RX_PANIC_JUMBO_REFILL	(3*RX_PANIC_JUMBO_THRES)/2
#define RX_LOW_JUMBO_THRES	(3*RX_JUMBO_SIZE)/4


/*
 * Size of the mini ring entries, basically these just should be big
 * enough to take TCP ACKs
 */
#define ACE_MINI_SIZE		100

#define ACE_MINI_BUFSIZE	ACE_MINI_SIZE
#define ACE_STD_BUFSIZE		(ACE_STD_MTU + ETH_HLEN + 4)
#define ACE_JUMBO_BUFSIZE	(ACE_JUMBO_MTU + ETH_HLEN + 4)

/*
 * There seems to be a magic difference in the effect between 995 and 996
 * but little difference between 900 and 995 ... no idea why.
 *
 * There is now a default set of tuning parameters which is set, depending
 * on whether or not the user enables Jumbo frames. It's assumed that if
 * Jumbo frames are enabled, the user wants optimal tuning for that case.
 */
#define DEF_TX_COAL		400 /* 996 */
#define DEF_TX_MAX_DESC		60  /* was 40 */
#define DEF_RX_COAL		120 /* 1000 */
#define DEF_RX_MAX_DESC		25
#define DEF_TX_RATIO		21 /* 24 */

#define DEF_JUMBO_TX_COAL	20
#define DEF_JUMBO_TX_MAX_DESC	60
#define DEF_JUMBO_RX_COAL	30
#define DEF_JUMBO_RX_MAX_DESC	6
#define DEF_JUMBO_TX_RATIO	21

#if tigon2FwReleaseLocal < 20001118
/*
 * Standard firmware and early modifications duplicate
 * IRQ load without this flag (coal timer is never reset).
 * Note that with this flag tx_coal should be less than
 * time to xmit full tx ring.
 * 400usec is not so bad for tx ring size of 128.
 */
#define TX_COAL_INTS_ONLY	1	/* worth it */
#else
/*
 * With modified firmware, this is not necessary, but still useful.
 */
#define TX_COAL_INTS_ONLY	1
#endif

#define DEF_TRACE		0
#define DEF_STAT		(2 * TICKS_PER_SEC)


static int link[ACE_MAX_MOD_PARMS];
static int trace[ACE_MAX_MOD_PARMS];
static int tx_coal_tick[ACE_MAX_MOD_PARMS];
static int rx_coal_tick[ACE_MAX_MOD_PARMS];
static int max_tx_desc[ACE_MAX_MOD_PARMS];
static int max_rx_desc[ACE_MAX_MOD_PARMS];
static int tx_ratio[ACE_MAX_MOD_PARMS];
static int dis_pci_mem_inval[ACE_MAX_MOD_PARMS] = {1, 1, 1, 1, 1, 1, 1, 1};

MODULE_AUTHOR("Jes Sorensen <jes@trained-monkey.org>");
MODULE_LICENSE("GPL");
MODULE_DESCRIPTION("AceNIC/3C985/GA620 Gigabit Ethernet driver");

module_param_array(link, int, NULL, 0);
module_param_array(trace, int, NULL, 0);
module_param_array(tx_coal_tick, int, NULL, 0);
module_param_array(max_tx_desc, int, NULL, 0);
module_param_array(rx_coal_tick, int, NULL, 0);
module_param_array(max_rx_desc, int, NULL, 0);
module_param_array(tx_ratio, int, NULL, 0);
MODULE_PARM_DESC(link, "AceNIC/3C985/NetGear link state");
MODULE_PARM_DESC(trace, "AceNIC/3C985/NetGear firmware trace level");
MODULE_PARM_DESC(tx_coal_tick, "AceNIC/3C985/GA620 max clock ticks to wait from first tx descriptor arrives");
MODULE_PARM_DESC(max_tx_desc, "AceNIC/3C985/GA620 max number of transmit descriptors to wait");
MODULE_PARM_DESC(rx_coal_tick, "AceNIC/3C985/GA620 max clock ticks to wait from first rx descriptor arrives");
MODULE_PARM_DESC(max_rx_desc, "AceNIC/3C985/GA620 max number of receive descriptors to wait");
MODULE_PARM_DESC(tx_ratio, "AceNIC/3C985/GA620 ratio of NIC memory used for TX/RX descriptors (range 0-63)");


static char version[] __devinitdata = 
  "acenic.c: v0.92 08/05/2002  Jes Sorensen, linux-acenic@SunSITE.dk\n"
  "                            http://home.cern.ch/~jes/gige/acenic.html\n";

static int ace_get_settings(struct net_device *, struct ethtool_cmd *);
static int ace_set_settings(struct net_device *, struct ethtool_cmd *);
static void ace_get_drvinfo(struct net_device *, struct ethtool_drvinfo *);

static struct ethtool_ops ace_ethtool_ops = {
	.get_settings = ace_get_settings,
	.set_settings = ace_set_settings,
	.get_drvinfo = ace_get_drvinfo,
};

static void ace_watchdog(struct net_device *dev);

static int __devinit acenic_probe_one(struct pci_dev *pdev,
		const struct pci_device_id *id)
{
	struct net_device *dev;
	struct ace_private *ap;
	static int boards_found;

	dev = alloc_etherdev(sizeof(struct ace_private));
	if (dev == NULL) {
		printk(KERN_ERR "acenic: Unable to allocate "
		       "net_device structure!\n");
		return -ENOMEM;
	}

	SET_MODULE_OWNER(dev);
	SET_NETDEV_DEV(dev, &pdev->dev);

	ap = dev->priv;
	ap->pdev = pdev;
	ap->name = pci_name(pdev);

	dev->features |= NETIF_F_SG | NETIF_F_IP_CSUM;
#if ACENIC_DO_VLAN
	dev->features |= NETIF_F_HW_VLAN_TX | NETIF_F_HW_VLAN_RX;
	dev->vlan_rx_register = ace_vlan_rx_register;
	dev->vlan_rx_kill_vid = ace_vlan_rx_kill_vid;
#endif
	if (1) {
		dev->tx_timeout = &ace_watchdog;
		dev->watchdog_timeo = 5*HZ;
	}

	dev->open = &ace_open;
	dev->stop = &ace_close;
	dev->hard_start_xmit = &ace_start_xmit;
	dev->get_stats = &ace_get_stats;
	dev->set_multicast_list = &ace_set_multicast_list;
	SET_ETHTOOL_OPS(dev, &ace_ethtool_ops);
	dev->set_mac_address = &ace_set_mac_addr;
	dev->change_mtu = &ace_change_mtu;

	/* we only display this string ONCE */
	if (!boards_found)
		printk(version);

	if (pci_enable_device(pdev))
		goto fail_free_netdev;

	/*
	 * Enable master mode before we start playing with the
	 * pci_command word since pci_set_master() will modify
	 * it.
	 */
	pci_set_master(pdev);

	pci_read_config_word(pdev, PCI_COMMAND, &ap->pci_command);

	/* OpenFirmware on Mac's does not set this - DOH.. */ 
	if (!(ap->pci_command & PCI_COMMAND_MEMORY)) {
		printk(KERN_INFO "%s: Enabling PCI Memory Mapped "
		       "access - was not enabled by BIOS/Firmware\n",
		       ap->name);
		ap->pci_command = ap->pci_command | PCI_COMMAND_MEMORY;
		pci_write_config_word(ap->pdev, PCI_COMMAND,
				      ap->pci_command);
		wmb();
	}

	pci_read_config_byte(pdev, PCI_LATENCY_TIMER, &ap->pci_latency);
	if (ap->pci_latency <= 0x40) {
		ap->pci_latency = 0x40;
		pci_write_config_byte(pdev, PCI_LATENCY_TIMER, ap->pci_latency);
	}

	/*
	 * Remap the regs into kernel space - this is abuse of
	 * dev->base_addr since it was means for I/O port
	 * addresses but who gives a damn.
	 */
	dev->base_addr = pci_resource_start(pdev, 0);
	ap->regs = ioremap(dev->base_addr, 0x4000);
	if (!ap->regs) {
		printk(KERN_ERR "%s:  Unable to map I/O register, "
		       "AceNIC %i will be disabled.\n",
		       ap->name, boards_found);
		goto fail_free_netdev;
	}

	switch(pdev->vendor) {
	case PCI_VENDOR_ID_ALTEON:
		if (pdev->device == PCI_DEVICE_ID_FARALLON_PN9100T) {
			printk(KERN_INFO "%s: Farallon PN9100-T ",
			       ap->name);
		} else {
			printk(KERN_INFO "%s: Alteon AceNIC ",
			       ap->name);
		}
		break;
	case PCI_VENDOR_ID_3COM:
		printk(KERN_INFO "%s: 3Com 3C985 ", ap->name);
		break;
	case PCI_VENDOR_ID_NETGEAR:
		printk(KERN_INFO "%s: NetGear GA620 ", ap->name);
		break;
	case PCI_VENDOR_ID_DEC:
		if (pdev->device == PCI_DEVICE_ID_FARALLON_PN9000SX) {
			printk(KERN_INFO "%s: Farallon PN9000-SX ",
			       ap->name);
			break;
		}
	case PCI_VENDOR_ID_SGI:
		printk(KERN_INFO "%s: SGI AceNIC ", ap->name);
		break;
	default:
		printk(KERN_INFO "%s: Unknown AceNIC ", ap->name);
		break;
	}

	printk("Gigabit Ethernet at 0x%08lx, ", dev->base_addr);
#ifdef __sparc__
	printk("irq %s\n", __irq_itoa(pdev->irq));
#else
	printk("irq %i\n", pdev->irq);
#endif

#ifdef CONFIG_ACENIC_OMIT_TIGON_I
	if ((readl(&ap->regs->HostCtrl) >> 28) == 4) {
		printk(KERN_ERR "%s: Driver compiled without Tigon I"
		       " support - NIC disabled\n", dev->name);
		goto fail_uninit;
	}
#endif

	if (ace_allocate_descriptors(dev))
		goto fail_free_netdev;

#ifdef MODULE
	if (boards_found >= ACE_MAX_MOD_PARMS)
		ap->board_idx = BOARD_IDX_OVERFLOW;
	else
		ap->board_idx = boards_found;
#else
	ap->board_idx = BOARD_IDX_STATIC;
#endif

	if (ace_init(dev))
		goto fail_free_netdev;

	if (register_netdev(dev)) {
		printk(KERN_ERR "acenic: device registration failed\n");
		goto fail_uninit;
	}
	ap->name = dev->name;

	if (ap->pci_using_dac)
		dev->features |= NETIF_F_HIGHDMA;

	pci_set_drvdata(pdev, dev);

	boards_found++;
	return 0;

 fail_uninit:
	ace_init_cleanup(dev);
 fail_free_netdev:
	free_netdev(dev);
	return -ENODEV;
}

static void __devexit acenic_remove_one(struct pci_dev *pdev)
{
	struct net_device *dev = pci_get_drvdata(pdev);
	struct ace_private *ap = netdev_priv(dev);
	struct ace_regs __iomem *regs = ap->regs;
	short i;

	unregister_netdev(dev);

	writel(readl(&regs->CpuCtrl) | CPU_HALT, &regs->CpuCtrl);
	if (ap->version >= 2)
		writel(readl(&regs->CpuBCtrl) | CPU_HALT, &regs->CpuBCtrl);
	
	/*
	 * This clears any pending interrupts
	 */
	writel(1, &regs->Mb0Lo);
	readl(&regs->CpuCtrl);	/* flush */

	/*
	 * Make sure no other CPUs are processing interrupts
	 * on the card before the buffers are being released.
	 * Otherwise one might experience some `interesting'
	 * effects.
	 *
	 * Then release the RX buffers - jumbo buffers were
	 * already released in ace_close().
	 */
	ace_sync_irq(dev->irq);

	for (i = 0; i < RX_STD_RING_ENTRIES; i++) {
		struct sk_buff *skb = ap->skb->rx_std_skbuff[i].skb;

		if (skb) {
			struct ring_info *ringp;
			dma_addr_t mapping;

			ringp = &ap->skb->rx_std_skbuff[i];
			mapping = pci_unmap_addr(ringp, mapping);
			pci_unmap_page(ap->pdev, mapping,
				       ACE_STD_BUFSIZE,
				       PCI_DMA_FROMDEVICE);

			ap->rx_std_ring[i].size = 0;
			ap->skb->rx_std_skbuff[i].skb = NULL;
			dev_kfree_skb(skb);
		}
	}

	if (ap->version >= 2) {
		for (i = 0; i < RX_MINI_RING_ENTRIES; i++) {
			struct sk_buff *skb = ap->skb->rx_mini_skbuff[i].skb;

			if (skb) {
				struct ring_info *ringp;
				dma_addr_t mapping;

				ringp = &ap->skb->rx_mini_skbuff[i];
				mapping = pci_unmap_addr(ringp,mapping);
				pci_unmap_page(ap->pdev, mapping,
					       ACE_MINI_BUFSIZE,
					       PCI_DMA_FROMDEVICE);

				ap->rx_mini_ring[i].size = 0;
				ap->skb->rx_mini_skbuff[i].skb = NULL;
				dev_kfree_skb(skb);
			}
		}
	}

	for (i = 0; i < RX_JUMBO_RING_ENTRIES; i++) {
		struct sk_buff *skb = ap->skb->rx_jumbo_skbuff[i].skb;
		if (skb) {
			struct ring_info *ringp;
			dma_addr_t mapping;

			ringp = &ap->skb->rx_jumbo_skbuff[i];
			mapping = pci_unmap_addr(ringp, mapping);
			pci_unmap_page(ap->pdev, mapping,
				       ACE_JUMBO_BUFSIZE,
				       PCI_DMA_FROMDEVICE);

			ap->rx_jumbo_ring[i].size = 0;
			ap->skb->rx_jumbo_skbuff[i].skb = NULL;
			dev_kfree_skb(skb);
		}
	}

	ace_init_cleanup(dev);
	free_netdev(dev);
}

static struct pci_driver acenic_pci_driver = {
	.name		= "acenic",
	.id_table	= acenic_pci_tbl,
	.probe		= acenic_probe_one,
	.remove		= __devexit_p(acenic_remove_one),
};

static int __init acenic_init(void)
{
	return pci_module_init(&acenic_pci_driver);
}

static void __exit acenic_exit(void)
{
	pci_unregister_driver(&acenic_pci_driver);
}

module_init(acenic_init);
module_exit(acenic_exit);

static void ace_free_descriptors(struct net_device *dev)
{
	struct ace_private *ap = netdev_priv(dev);
	int size;

	if (ap->rx_std_ring != NULL) {
		size = (sizeof(struct rx_desc) *
			(RX_STD_RING_ENTRIES +
			 RX_JUMBO_RING_ENTRIES +
			 RX_MINI_RING_ENTRIES +
			 RX_RETURN_RING_ENTRIES));
		pci_free_consistent(ap->pdev, size, ap->rx_std_ring,
				    ap->rx_ring_base_dma);
		ap->rx_std_ring = NULL;
		ap->rx_jumbo_ring = NULL;
		ap->rx_mini_ring = NULL;
		ap->rx_return_ring = NULL;
	}
	if (ap->evt_ring != NULL) {
		size = (sizeof(struct event) * EVT_RING_ENTRIES);
		pci_free_consistent(ap->pdev, size, ap->evt_ring,
				    ap->evt_ring_dma);
		ap->evt_ring = NULL;
	}
	if (ap->tx_ring != NULL && !ACE_IS_TIGON_I(ap)) {
		size = (sizeof(struct tx_desc) * MAX_TX_RING_ENTRIES);
		pci_free_consistent(ap->pdev, size, ap->tx_ring,
				    ap->tx_ring_dma);
	}
	ap->tx_ring = NULL;

	if (ap->evt_prd != NULL) {
		pci_free_consistent(ap->pdev, sizeof(u32),
				    (void *)ap->evt_prd, ap->evt_prd_dma);
		ap->evt_prd = NULL;
	}
	if (ap->rx_ret_prd != NULL) {
		pci_free_consistent(ap->pdev, sizeof(u32),
				    (void *)ap->rx_ret_prd,
				    ap->rx_ret_prd_dma);
		ap->rx_ret_prd = NULL;
	}
	if (ap->tx_csm != NULL) {
		pci_free_consistent(ap->pdev, sizeof(u32),
				    (void *)ap->tx_csm, ap->tx_csm_dma);
		ap->tx_csm = NULL;
	}
}


static int ace_allocate_descriptors(struct net_device *dev)
{
	struct ace_private *ap = netdev_priv(dev);
	int size;

	size = (sizeof(struct rx_desc) *
		(RX_STD_RING_ENTRIES +
		 RX_JUMBO_RING_ENTRIES +
		 RX_MINI_RING_ENTRIES +
		 RX_RETURN_RING_ENTRIES));

	ap->rx_std_ring = pci_alloc_consistent(ap->pdev, size,
					       &ap->rx_ring_base_dma);
	if (ap->rx_std_ring == NULL)
		goto fail;

	ap->rx_jumbo_ring = ap->rx_std_ring + RX_STD_RING_ENTRIES;
	ap->rx_mini_ring = ap->rx_jumbo_ring + RX_JUMBO_RING_ENTRIES;
	ap->rx_return_ring = ap->rx_mini_ring + RX_MINI_RING_ENTRIES;

	size = (sizeof(struct event) * EVT_RING_ENTRIES);

	ap->evt_ring = pci_alloc_consistent(ap->pdev, size, &ap->evt_ring_dma);

	if (ap->evt_ring == NULL)
		goto fail;

	/*
	 * Only allocate a host TX ring for the Tigon II, the Tigon I
	 * has to use PCI registers for this ;-(
	 */
	if (!ACE_IS_TIGON_I(ap)) {
		size = (sizeof(struct tx_desc) * MAX_TX_RING_ENTRIES);

		ap->tx_ring = pci_alloc_consistent(ap->pdev, size,
						   &ap->tx_ring_dma);

		if (ap->tx_ring == NULL)
			goto fail;
	}

	ap->evt_prd = pci_alloc_consistent(ap->pdev, sizeof(u32),
					   &ap->evt_prd_dma);
	if (ap->evt_prd == NULL)
		goto fail;

	ap->rx_ret_prd = pci_alloc_consistent(ap->pdev, sizeof(u32),
					      &ap->rx_ret_prd_dma);
	if (ap->rx_ret_prd == NULL)
		goto fail;

	ap->tx_csm = pci_alloc_consistent(ap->pdev, sizeof(u32),
					  &ap->tx_csm_dma);
	if (ap->tx_csm == NULL)
		goto fail;

	return 0;

fail:
	/* Clean up. */
	ace_init_cleanup(dev);
	return 1;
}


/*
 * Generic cleanup handling data allocated during init. Used when the
 * module is unloaded or if an error occurs during initialization
 */
static void ace_init_cleanup(struct net_device *dev)
{
	struct ace_private *ap;

	ap = netdev_priv(dev);

	ace_free_descriptors(dev);

	if (ap->info)
		pci_free_consistent(ap->pdev, sizeof(struct ace_info),
				    ap->info, ap->info_dma);
	kfree(ap->skb);
	kfree(ap->trace_buf);

	if (dev->irq)
		free_irq(dev->irq, dev);

	iounmap(ap->regs);
}


/*
 * Commands are considered to be slow.
 */
static inline void ace_issue_cmd(struct ace_regs __iomem *regs, struct cmd *cmd)
{
	u32 idx;

	idx = readl(&regs->CmdPrd);

	writel(*(u32 *)(cmd), &regs->CmdRng[idx]);
	idx = (idx + 1) % CMD_RING_ENTRIES;

	writel(idx, &regs->CmdPrd);
}


static int __devinit ace_init(struct net_device *dev)
{
	struct ace_private *ap;
	struct ace_regs __iomem *regs;
	struct ace_info *info = NULL;
	struct pci_dev *pdev;
	unsigned long myjif;
	u64 tmp_ptr;
	u32 tig_ver, mac1, mac2, tmp, pci_state;
	int board_idx, ecode = 0;
	short i;
	unsigned char cache_size;

	ap = netdev_priv(dev);
	regs = ap->regs;

	board_idx = ap->board_idx;

	/*
	 * aman@sgi.com - its useful to do a NIC reset here to
	 * address the `Firmware not running' problem subsequent
	 * to any crashes involving the NIC
	 */
	writel(HW_RESET | (HW_RESET << 24), &regs->HostCtrl);
	readl(&regs->HostCtrl);		/* PCI write posting */
	udelay(5);

	/*
	 * Don't access any other registers before this point!
	 */
#ifdef __BIG_ENDIAN
	/*
	 * This will most likely need BYTE_SWAP once we switch
	 * to using __raw_writel()
	 */
	writel((WORD_SWAP | CLR_INT | ((WORD_SWAP | CLR_INT) << 24)),
	       &regs->HostCtrl);
#else
	writel((CLR_INT | WORD_SWAP | ((CLR_INT | WORD_SWAP) << 24)),
	       &regs->HostCtrl);
#endif
	readl(&regs->HostCtrl);		/* PCI write posting */

	/*
	 * Stop the NIC CPU and clear pending interrupts
	 */
	writel(readl(&regs->CpuCtrl) | CPU_HALT, &regs->CpuCtrl);
	readl(&regs->CpuCtrl);		/* PCI write posting */
	writel(0, &regs->Mb0Lo);

	tig_ver = readl(&regs->HostCtrl) >> 28;

	switch(tig_ver){
#ifndef CONFIG_ACENIC_OMIT_TIGON_I
	case 4:
	case 5:
		printk(KERN_INFO "  Tigon I  (Rev. %i), Firmware: %i.%i.%i, ",
		       tig_ver, tigonFwReleaseMajor, tigonFwReleaseMinor,
		       tigonFwReleaseFix);
		writel(0, &regs->LocalCtrl);
		ap->version = 1;
		ap->tx_ring_entries = TIGON_I_TX_RING_ENTRIES;
		break;
#endif
	case 6:
		printk(KERN_INFO "  Tigon II (Rev. %i), Firmware: %i.%i.%i, ",
		       tig_ver, tigon2FwReleaseMajor, tigon2FwReleaseMinor,
		       tigon2FwReleaseFix);
		writel(readl(&regs->CpuBCtrl) | CPU_HALT, &regs->CpuBCtrl);
		readl(&regs->CpuBCtrl);		/* PCI write posting */
		/*
		 * The SRAM bank size does _not_ indicate the amount
		 * of memory on the card, it controls the _bank_ size!
		 * Ie. a 1MB AceNIC will have two banks of 512KB.
		 */
		writel(SRAM_BANK_512K, &regs->LocalCtrl);
		writel(SYNC_SRAM_TIMING, &regs->MiscCfg);
		ap->version = 2;
		ap->tx_ring_entries = MAX_TX_RING_ENTRIES;
		break;
	default:
		printk(KERN_WARNING "  Unsupported Tigon version detected "
		       "(%i)\n", tig_ver);
		ecode = -ENODEV;
		goto init_error;
	}

	/*
	 * ModeStat _must_ be set after the SRAM settings as this change
	 * seems to corrupt the ModeStat and possible other registers.
	 * The SRAM settings survive resets and setting it to the same
	 * value a second time works as well. This is what caused the
	 * `Firmware not running' problem on the Tigon II.
	 */
#ifdef __BIG_ENDIAN
	writel(ACE_BYTE_SWAP_DMA | ACE_WARN | ACE_FATAL | ACE_BYTE_SWAP_BD |
	       ACE_WORD_SWAP_BD | ACE_NO_JUMBO_FRAG, &regs->ModeStat);
#else
	writel(ACE_BYTE_SWAP_DMA | ACE_WARN | ACE_FATAL |
	       ACE_WORD_SWAP_BD | ACE_NO_JUMBO_FRAG, &regs->ModeStat);
#endif
	readl(&regs->ModeStat);		/* PCI write posting */

	mac1 = 0;
	for(i = 0; i < 4; i++) {
		mac1 = mac1 << 8;
		tmp = read_eeprom_byte(dev, 0x8c+i);
		if (tmp < 0) {
			ecode = -EIO;
			goto init_error;
		} else
			mac1 |= (tmp & 0xff);
	}
	mac2 = 0;
	for(i = 4; i < 8; i++) {
		mac2 = mac2 << 8;
		tmp = read_eeprom_byte(dev, 0x8c+i);
		if (tmp < 0) {
			ecode = -EIO;
			goto init_error;
		} else
			mac2 |= (tmp & 0xff);
	}

	writel(mac1, &regs->MacAddrHi);
	writel(mac2, &regs->MacAddrLo);

	printk("MAC: %02x:%02x:%02x:%02x:%02x:%02x\n",
	       (mac1 >> 8) & 0xff, mac1 & 0xff, (mac2 >> 24) &0xff,
	       (mac2 >> 16) & 0xff, (mac2 >> 8) & 0xff, mac2 & 0xff);

	dev->dev_addr[0] = (mac1 >> 8) & 0xff;
	dev->dev_addr[1] = mac1 & 0xff;
	dev->dev_addr[2] = (mac2 >> 24) & 0xff;
	dev->dev_addr[3] = (mac2 >> 16) & 0xff;
	dev->dev_addr[4] = (mac2 >> 8) & 0xff;
	dev->dev_addr[5] = mac2 & 0xff;

	/*
	 * Looks like this is necessary to deal with on all architectures,
	 * even this %$#%$# N440BX Intel based thing doesn't get it right.
	 * Ie. having two NICs in the machine, one will have the cache
	 * line set at boot time, the other will not.
	 */
	pdev = ap->pdev;
	pci_read_config_byte(pdev, PCI_CACHE_LINE_SIZE, &cache_size);
	cache_size <<= 2;
	if (cache_size != SMP_CACHE_BYTES) {
		printk(KERN_INFO "  PCI cache line size set incorrectly "
		       "(%i bytes) by BIOS/FW, ", cache_size);
		if (cache_size > SMP_CACHE_BYTES)
			printk("expecting %i\n", SMP_CACHE_BYTES);
		else {
			printk("correcting to %i\n", SMP_CACHE_BYTES);
			pci_write_config_byte(pdev, PCI_CACHE_LINE_SIZE,
					      SMP_CACHE_BYTES >> 2);
		}
	}

	pci_state = readl(&regs->PciState);
	printk(KERN_INFO "  PCI bus width: %i bits, speed: %iMHz, "
	       "latency: %i clks\n",
	       	(pci_state & PCI_32BIT) ? 32 : 64,
		(pci_state & PCI_66MHZ) ? 66 : 33, 
		ap->pci_latency);

	/*
	 * Set the max DMA transfer size. Seems that for most systems
	 * the performance is better when no MAX parameter is
	 * set. However for systems enabling PCI write and invalidate,
	 * DMA writes must be set to the L1 cache line size to get
	 * optimal performance.
	 *
	 * The default is now to turn the PCI write and invalidate off
	 * - that is what Alteon does for NT.
	 */
	tmp = READ_CMD_MEM | WRITE_CMD_MEM;
	if (ap->version >= 2) {
		tmp |= (MEM_READ_MULTIPLE | (pci_state & PCI_66MHZ));
		/*
		 * Tuning parameters only supported for 8 cards
		 */
		if (board_idx == BOARD_IDX_OVERFLOW ||
		    dis_pci_mem_inval[board_idx]) {
			if (ap->pci_command & PCI_COMMAND_INVALIDATE) {
				ap->pci_command &= ~PCI_COMMAND_INVALIDATE;
				pci_write_config_word(pdev, PCI_COMMAND,
						      ap->pci_command);
				printk(KERN_INFO "  Disabling PCI memory "
				       "write and invalidate\n");
			}
		} else if (ap->pci_command & PCI_COMMAND_INVALIDATE) {
			printk(KERN_INFO "  PCI memory write & invalidate "
			       "enabled by BIOS, enabling counter measures\n");

			switch(SMP_CACHE_BYTES) {
			case 16:
				tmp |= DMA_WRITE_MAX_16;
				break;
			case 32:
				tmp |= DMA_WRITE_MAX_32;
				break;
			case 64:
				tmp |= DMA_WRITE_MAX_64;
				break;
			case 128:
				tmp |= DMA_WRITE_MAX_128;
				break;
			default:
				printk(KERN_INFO "  Cache line size %i not "
				       "supported, PCI write and invalidate "
				       "disabled\n", SMP_CACHE_BYTES);
				ap->pci_command &= ~PCI_COMMAND_INVALIDATE;
				pci_write_config_word(pdev, PCI_COMMAND,
						      ap->pci_command);
			}
		}
	}

#ifdef __sparc__
	/*
	 * On this platform, we know what the best dma settings
	 * are.  We use 64-byte maximum bursts, because if we
	 * burst larger than the cache line size (or even cross
	 * a 64byte boundary in a single burst) the UltraSparc
	 * PCI controller will disconnect at 64-byte multiples.
	 *
	 * Read-multiple will be properly enabled above, and when
	 * set will give the PCI controller proper hints about
	 * prefetching.
	 */
	tmp &= ~DMA_READ_WRITE_MASK;
	tmp |= DMA_READ_MAX_64;
	tmp |= DMA_WRITE_MAX_64;
#endif
#ifdef __alpha__
	tmp &= ~DMA_READ_WRITE_MASK;
	tmp |= DMA_READ_MAX_128;
	/*
	 * All the docs say MUST NOT. Well, I did.
	 * Nothing terrible happens, if we load wrong size.
	 * Bit w&i still works better!
	 */
	tmp |= DMA_WRITE_MAX_128;
#endif
	writel(tmp, &regs->PciState);

#if 0
	/*
	 * The Host PCI bus controller driver has to set FBB.
	 * If all devices on that PCI bus support FBB, then the controller
	 * can enable FBB support in the Host PCI Bus controller (or on
	 * the PCI-PCI bridge if that applies).
	 * -ggg
	 */
	/*
	 * I have received reports from people having problems when this
	 * bit is enabled.
	 */
	if (!(ap->pci_command & PCI_COMMAND_FAST_BACK)) {
		printk(KERN_INFO "  Enabling PCI Fast Back to Back\n");
		ap->pci_command |= PCI_COMMAND_FAST_BACK;
		pci_write_config_word(pdev, PCI_COMMAND, ap->pci_command);
	}
#endif
		
	/*
	 * Configure DMA attributes.
	 */
	if (!pci_set_dma_mask(pdev, DMA_64BIT_MASK)) {
		ap->pci_using_dac = 1;
	} else if (!pci_set_dma_mask(pdev, DMA_32BIT_MASK)) {
		ap->pci_using_dac = 0;
	} else {
		ecode = -ENODEV;
		goto init_error;
	}

	/*
	 * Initialize the generic info block and the command+event rings
	 * and the control blocks for the transmit and receive rings
	 * as they need to be setup once and for all.
	 */
	if (!(info = pci_alloc_consistent(ap->pdev, sizeof(struct ace_info),
					  &ap->info_dma))) {
		ecode = -EAGAIN;
		goto init_error;
	}
	ap->info = info;

	/*
	 * Get the memory for the skb rings.
	 */
	if (!(ap->skb = kmalloc(sizeof(struct ace_skb), GFP_KERNEL))) {
		ecode = -EAGAIN;
		goto init_error;
	}

	ecode = request_irq(pdev->irq, ace_interrupt, SA_SHIRQ,
			    DRV_NAME, dev);
	if (ecode) {
		printk(KERN_WARNING "%s: Requested IRQ %d is busy\n",
		       DRV_NAME, pdev->irq);
		goto init_error;
	} else
		dev->irq = pdev->irq;

#ifdef INDEX_DEBUG
	spin_lock_init(&ap->debug_lock);
	ap->last_tx = ACE_TX_RING_ENTRIES(ap) - 1;
	ap->last_std_rx = 0;
	ap->last_mini_rx = 0;
#endif

	memset(ap->info, 0, sizeof(struct ace_info));
	memset(ap->skb, 0, sizeof(struct ace_skb));

	ace_load_firmware(dev);
	ap->fw_running = 0;

	tmp_ptr = ap->info_dma;
	writel(tmp_ptr >> 32, &regs->InfoPtrHi);
	writel(tmp_ptr & 0xffffffff, &regs->InfoPtrLo);

	memset(ap->evt_ring, 0, EVT_RING_ENTRIES * sizeof(struct event));

	set_aceaddr(&info->evt_ctrl.rngptr, ap->evt_ring_dma);
	info->evt_ctrl.flags = 0;

	*(ap->evt_prd) = 0;
	wmb();
	set_aceaddr(&info->evt_prd_ptr, ap->evt_prd_dma);
	writel(0, &regs->EvtCsm);

	set_aceaddr(&info->cmd_ctrl.rngptr, 0x100);
	info->cmd_ctrl.flags = 0;
	info->cmd_ctrl.max_len = 0;

	for (i = 0; i < CMD_RING_ENTRIES; i++)
		writel(0, &regs->CmdRng[i]);

	writel(0, &regs->CmdPrd);
	writel(0, &regs->CmdCsm);

	tmp_ptr = ap->info_dma;
	tmp_ptr += (unsigned long) &(((struct ace_info *)0)->s.stats);
	set_aceaddr(&info->stats2_ptr, (dma_addr_t) tmp_ptr);

	set_aceaddr(&info->rx_std_ctrl.rngptr, ap->rx_ring_base_dma);
	info->rx_std_ctrl.max_len = ACE_STD_BUFSIZE;
	info->rx_std_ctrl.flags =
	  RCB_FLG_TCP_UDP_SUM | RCB_FLG_NO_PSEUDO_HDR | ACE_RCB_VLAN_FLAG;

	memset(ap->rx_std_ring, 0,
	       RX_STD_RING_ENTRIES * sizeof(struct rx_desc));

	for (i = 0; i < RX_STD_RING_ENTRIES; i++)
		ap->rx_std_ring[i].flags = BD_FLG_TCP_UDP_SUM;

	ap->rx_std_skbprd = 0;
	atomic_set(&ap->cur_rx_bufs, 0);

	set_aceaddr(&info->rx_jumbo_ctrl.rngptr,
		    (ap->rx_ring_base_dma +
		     (sizeof(struct rx_desc) * RX_STD_RING_ENTRIES)));
	info->rx_jumbo_ctrl.max_len = 0;
	info->rx_jumbo_ctrl.flags =
	  RCB_FLG_TCP_UDP_SUM | RCB_FLG_NO_PSEUDO_HDR | ACE_RCB_VLAN_FLAG;

	memset(ap->rx_jumbo_ring, 0,
	       RX_JUMBO_RING_ENTRIES * sizeof(struct rx_desc));

	for (i = 0; i < RX_JUMBO_RING_ENTRIES; i++)
		ap->rx_jumbo_ring[i].flags = BD_FLG_TCP_UDP_SUM | BD_FLG_JUMBO;

	ap->rx_jumbo_skbprd = 0;
	atomic_set(&ap->cur_jumbo_bufs, 0);

	memset(ap->rx_mini_ring, 0,
	       RX_MINI_RING_ENTRIES * sizeof(struct rx_desc));

	if (ap->version >= 2) {
		set_aceaddr(&info->rx_mini_ctrl.rngptr,
			    (ap->rx_ring_base_dma +
			     (sizeof(struct rx_desc) *
			      (RX_STD_RING_ENTRIES +
			       RX_JUMBO_RING_ENTRIES))));
		info->rx_mini_ctrl.max_len = ACE_MINI_SIZE;
		info->rx_mini_ctrl.flags = 
		  RCB_FLG_TCP_UDP_SUM|RCB_FLG_NO_PSEUDO_HDR|ACE_RCB_VLAN_FLAG;

		for (i = 0; i < RX_MINI_RING_ENTRIES; i++)
			ap->rx_mini_ring[i].flags =
				BD_FLG_TCP_UDP_SUM | BD_FLG_MINI;
	} else {
		set_aceaddr(&info->rx_mini_ctrl.rngptr, 0);
		info->rx_mini_ctrl.flags = RCB_FLG_RNG_DISABLE;
		info->rx_mini_ctrl.max_len = 0;
	}

	ap->rx_mini_skbprd = 0;
	atomic_set(&ap->cur_mini_bufs, 0);

	set_aceaddr(&info->rx_return_ctrl.rngptr,
		    (ap->rx_ring_base_dma +
		     (sizeof(struct rx_desc) *
		      (RX_STD_RING_ENTRIES +
		       RX_JUMBO_RING_ENTRIES +
		       RX_MINI_RING_ENTRIES))));
	info->rx_return_ctrl.flags = 0;
	info->rx_return_ctrl.max_len = RX_RETURN_RING_ENTRIES;

	memset(ap->rx_return_ring, 0,
	       RX_RETURN_RING_ENTRIES * sizeof(struct rx_desc));

	set_aceaddr(&info->rx_ret_prd_ptr, ap->rx_ret_prd_dma);
	*(ap->rx_ret_prd) = 0;

	writel(TX_RING_BASE, &regs->WinBase);

	if (ACE_IS_TIGON_I(ap)) {
		ap->tx_ring = (struct tx_desc *) regs->Window;
		for (i = 0; i < (TIGON_I_TX_RING_ENTRIES 
				 * sizeof(struct tx_desc)) / sizeof(u32); i++)
			writel(0, (void __iomem *)ap->tx_ring  + i * 4);

		set_aceaddr(&info->tx_ctrl.rngptr, TX_RING_BASE);
	} else {
		memset(ap->tx_ring, 0,
		       MAX_TX_RING_ENTRIES * sizeof(struct tx_desc));

		set_aceaddr(&info->tx_ctrl.rngptr, ap->tx_ring_dma);
	}

	info->tx_ctrl.max_len = ACE_TX_RING_ENTRIES(ap);
	tmp = RCB_FLG_TCP_UDP_SUM | RCB_FLG_NO_PSEUDO_HDR | ACE_RCB_VLAN_FLAG;

	/*
	 * The Tigon I does not like having the TX ring in host memory ;-(
	 */
	if (!ACE_IS_TIGON_I(ap))
		tmp |= RCB_FLG_TX_HOST_RING;
#if TX_COAL_INTS_ONLY
	tmp |= RCB_FLG_COAL_INT_ONLY;
#endif
	info->tx_ctrl.flags = tmp;

	set_aceaddr(&info->tx_csm_ptr, ap->tx_csm_dma);

	/*
	 * Potential item for tuning parameter
	 */
#if 0 /* NO */
	writel(DMA_THRESH_16W, &regs->DmaReadCfg);
	writel(DMA_THRESH_16W, &regs->DmaWriteCfg);
#else
	writel(DMA_THRESH_8W, &regs->DmaReadCfg);
	writel(DMA_THRESH_8W, &regs->DmaWriteCfg);
#endif

	writel(0, &regs->MaskInt);
	writel(1, &regs->IfIdx);
#if 0
	/*
	 * McKinley boxes do not like us fiddling with AssistState
	 * this early
	 */
	writel(1, &regs->AssistState);
#endif

	writel(DEF_STAT, &regs->TuneStatTicks);
	writel(DEF_TRACE, &regs->TuneTrace);

	ace_set_rxtx_parms(dev, 0);

	if (board_idx == BOARD_IDX_OVERFLOW) {
		printk(KERN_WARNING "%s: more than %i NICs detected, "
		       "ignoring module parameters!\n",
		       ap->name, ACE_MAX_MOD_PARMS);
	} else if (board_idx >= 0) {
		if (tx_coal_tick[board_idx])
			writel(tx_coal_tick[board_idx],
			       &regs->TuneTxCoalTicks);
		if (max_tx_desc[board_idx])
			writel(max_tx_desc[board_idx], &regs->TuneMaxTxDesc);

		if (rx_coal_tick[board_idx])
			writel(rx_coal_tick[board_idx],
			       &regs->TuneRxCoalTicks);
		if (max_rx_desc[board_idx])
			writel(max_rx_desc[board_idx], &regs->TuneMaxRxDesc);

		if (trace[board_idx])
			writel(trace[board_idx], &regs->TuneTrace);

		if ((tx_ratio[board_idx] > 0) && (tx_ratio[board_idx] < 64))
			writel(tx_ratio[board_idx], &regs->TxBufRat);
	}

	/*
	 * Default link parameters
	 */
	tmp = LNK_ENABLE | LNK_FULL_DUPLEX | LNK_1000MB | LNK_100MB |
		LNK_10MB | LNK_RX_FLOW_CTL_Y | LNK_NEG_FCTL | LNK_NEGOTIATE;
	if(ap->version >= 2)
		tmp |= LNK_TX_FLOW_CTL_Y;

	/*
	 * Override link default parameters
	 */
	if ((board_idx >= 0) && link[board_idx]) {
		int option = link[board_idx];

		tmp = LNK_ENABLE;

		if (option & 0x01) {
			printk(KERN_INFO "%s: Setting half duplex link\n",
			       ap->name);
			tmp &= ~LNK_FULL_DUPLEX;
		}
		if (option & 0x02)
			tmp &= ~LNK_NEGOTIATE;
		if (option & 0x10)
			tmp |= LNK_10MB;
		if (option & 0x20)
			tmp |= LNK_100MB;
		if (option & 0x40)
			tmp |= LNK_1000MB;
		if ((option & 0x70) == 0) {
			printk(KERN_WARNING "%s: No media speed specified, "
			       "forcing auto negotiation\n", ap->name);
			tmp |= LNK_NEGOTIATE | LNK_1000MB |
				LNK_100MB | LNK_10MB;
		}
		if ((option & 0x100) == 0)
			tmp |= LNK_NEG_FCTL;
		else
			printk(KERN_INFO "%s: Disabling flow control "
			       "negotiation\n", ap->name);
		if (option & 0x200)
			tmp |= LNK_RX_FLOW_CTL_Y;
		if ((option & 0x400) && (ap->version >= 2)) {
			printk(KERN_INFO "%s: Enabling TX flow control\n",
			       ap->name);
			tmp |= LNK_TX_FLOW_CTL_Y;
		}
	}

	ap->link = tmp;
	writel(tmp, &regs->TuneLink);
	if (ap->version >= 2)
		writel(tmp, &regs->TuneFastLink);

	if (ACE_IS_TIGON_I(ap))
		writel(tigonFwStartAddr, &regs->Pc);
	if (ap->version == 2)
		writel(tigon2FwStartAddr, &regs->Pc);

	writel(0, &regs->Mb0Lo);

	/*
	 * Set tx_csm before we start receiving interrupts, otherwise
	 * the interrupt handler might think it is supposed to process
	 * tx ints before we are up and running, which may cause a null
	 * pointer access in the int handler.
	 */
	ap->cur_rx = 0;
	ap->tx_prd = *(ap->tx_csm) = ap->tx_ret_csm = 0;

	wmb();
	ace_set_txprd(regs, ap, 0);
	writel(0, &regs->RxRetCsm);

	/*
	 * Zero the stats before starting the interface
	 */
	memset(&ap->stats, 0, sizeof(ap->stats));

       /*
	* Enable DMA engine now.
	* If we do this sooner, Mckinley box pukes.
	* I assume it's because Tigon II DMA engine wants to check
	* *something* even before the CPU is started.
	*/
       writel(1, &regs->AssistState);  /* enable DMA */

	/*
	 * Start the NIC CPU
	 */
	writel(readl(&regs->CpuCtrl) & ~(CPU_HALT|CPU_TRACE), &regs->CpuCtrl);
	readl(&regs->CpuCtrl);

	/*
	 * Wait for the firmware to spin up - max 3 seconds.
	 */
	myjif = jiffies + 3 * HZ;
	while (time_before(jiffies, myjif) && !ap->fw_running)
		cpu_relax();

	if (!ap->fw_running) {
		printk(KERN_ERR "%s: Firmware NOT running!\n", ap->name);

		ace_dump_trace(ap);
		writel(readl(&regs->CpuCtrl) | CPU_HALT, &regs->CpuCtrl);
		readl(&regs->CpuCtrl);

		/* aman@sgi.com - account for badly behaving firmware/NIC:
		 * - have observed that the NIC may continue to generate
		 *   interrupts for some reason; attempt to stop it - halt
		 *   second CPU for Tigon II cards, and also clear Mb0
		 * - if we're a module, we'll fail to load if this was
		 *   the only GbE card in the system => if the kernel does
		 *   see an interrupt from the NIC, code to handle it is
		 *   gone and OOps! - so free_irq also
		 */
		if (ap->version >= 2)
			writel(readl(&regs->CpuBCtrl) | CPU_HALT,
			       &regs->CpuBCtrl);
		writel(0, &regs->Mb0Lo);
		readl(&regs->Mb0Lo);

		ecode = -EBUSY;
		goto init_error;
	}

	/*
	 * We load the ring here as there seem to be no way to tell the
	 * firmware to wipe the ring without re-initializing it.
	 */
	if (!test_and_set_bit(0, &ap->std_refill_busy))
		ace_load_std_rx_ring(ap, RX_RING_SIZE);
	else
		printk(KERN_ERR "%s: Someone is busy refilling the RX ring\n",
		       ap->name);
	if (ap->version >= 2) {
		if (!test_and_set_bit(0, &ap->mini_refill_busy))
			ace_load_mini_rx_ring(ap, RX_MINI_SIZE);
		else
			printk(KERN_ERR "%s: Someone is busy refilling "
			       "the RX mini ring\n", ap->name);
	}
	return 0;

 init_error:
	ace_init_cleanup(dev);
	return ecode;
}


static void ace_set_rxtx_parms(struct net_device *dev, int jumbo)
{
	struct ace_private *ap = netdev_priv(dev);
	struct ace_regs __iomem *regs = ap->regs;
	int board_idx = ap->board_idx;

	if (board_idx >= 0) {
		if (!jumbo) {
			if (!tx_coal_tick[board_idx])
				writel(DEF_TX_COAL, &regs->TuneTxCoalTicks);
			if (!max_tx_desc[board_idx])
				writel(DEF_TX_MAX_DESC, &regs->TuneMaxTxDesc);
			if (!rx_coal_tick[board_idx])
				writel(DEF_RX_COAL, &regs->TuneRxCoalTicks);
			if (!max_rx_desc[board_idx])
				writel(DEF_RX_MAX_DESC, &regs->TuneMaxRxDesc);
			if (!tx_ratio[board_idx])
				writel(DEF_TX_RATIO, &regs->TxBufRat);
		} else {
			if (!tx_coal_tick[board_idx])
				writel(DEF_JUMBO_TX_COAL,
				       &regs->TuneTxCoalTicks);
			if (!max_tx_desc[board_idx])
				writel(DEF_JUMBO_TX_MAX_DESC,
				       &regs->TuneMaxTxDesc);
			if (!rx_coal_tick[board_idx])
				writel(DEF_JUMBO_RX_COAL,
				       &regs->TuneRxCoalTicks);
			if (!max_rx_desc[board_idx])
				writel(DEF_JUMBO_RX_MAX_DESC,
				       &regs->TuneMaxRxDesc);
			if (!tx_ratio[board_idx])
				writel(DEF_JUMBO_TX_RATIO, &regs->TxBufRat);
		}
	}
}


static void ace_watchdog(struct net_device *data)
{
	struct net_device *dev = data;
	struct ace_private *ap = netdev_priv(dev);
	struct ace_regs __iomem *regs = ap->regs;

	/*
	 * We haven't received a stats update event for more than 2.5
	 * seconds and there is data in the transmit queue, thus we
	 * asume the card is stuck.
	 */
	if (*ap->tx_csm != ap->tx_ret_csm) {
		printk(KERN_WARNING "%s: Transmitter is stuck, %08x\n",
		       dev->name, (unsigned int)readl(&regs->HostCtrl));
		/* This can happen due to ieee flow control. */
	} else {
		printk(KERN_DEBUG "%s: BUG... transmitter died. Kicking it.\n",
		       dev->name);
#if 0
		netif_wake_queue(dev);
#endif
	}
}


static void ace_tasklet(unsigned long dev)
{
	struct ace_private *ap = netdev_priv((struct net_device *)dev);
	int cur_size;

	cur_size = atomic_read(&ap->cur_rx_bufs);
	if ((cur_size < RX_LOW_STD_THRES) &&
	    !test_and_set_bit(0, &ap->std_refill_busy)) {
#ifdef DEBUG
		printk("refilling buffers (current %i)\n", cur_size);
#endif
		ace_load_std_rx_ring(ap, RX_RING_SIZE - cur_size);
	}

	if (ap->version >= 2) {
		cur_size = atomic_read(&ap->cur_mini_bufs);
		if ((cur_size < RX_LOW_MINI_THRES) &&
		    !test_and_set_bit(0, &ap->mini_refill_busy)) {
#ifdef DEBUG
			printk("refilling mini buffers (current %i)\n",
			       cur_size);
#endif
			ace_load_mini_rx_ring(ap, RX_MINI_SIZE - cur_size);
		}
	}

	cur_size = atomic_read(&ap->cur_jumbo_bufs);
	if (ap->jumbo && (cur_size < RX_LOW_JUMBO_THRES) &&
	    !test_and_set_bit(0, &ap->jumbo_refill_busy)) {
#ifdef DEBUG
		printk("refilling jumbo buffers (current %i)\n", cur_size);
#endif
		ace_load_jumbo_rx_ring(ap, RX_JUMBO_SIZE - cur_size);
	}
	ap->tasklet_pending = 0;
}


/*
 * Copy the contents of the NIC's trace buffer to kernel memory.
 */
static void ace_dump_trace(struct ace_private *ap)
{
#if 0
	if (!ap->trace_buf)
		if (!(ap->trace_buf = kmalloc(ACE_TRACE_SIZE, GFP_KERNEL)))
		    return;
#endif
}


/*
 * Load the standard rx ring.
 *
 * Loading rings is safe without holding the spin lock since this is
 * done only before the device is enabled, thus no interrupts are
 * generated and by the interrupt handler/tasklet handler.
 */
static void ace_load_std_rx_ring(struct ace_private *ap, int nr_bufs)
{
	struct ace_regs __iomem *regs = ap->regs;
	short i, idx;
	

	prefetchw(&ap->cur_rx_bufs);

	idx = ap->rx_std_skbprd;

	for (i = 0; i < nr_bufs; i++) {
		struct sk_buff *skb;
		struct rx_desc *rd;
		dma_addr_t mapping;

		skb = alloc_skb(ACE_STD_BUFSIZE + NET_IP_ALIGN, GFP_ATOMIC);
		if (!skb)
			break;

		skb_reserve(skb, NET_IP_ALIGN);
		mapping = pci_map_page(ap->pdev, virt_to_page(skb->data),
				       offset_in_page(skb->data),
				       ACE_STD_BUFSIZE,
				       PCI_DMA_FROMDEVICE);
		ap->skb->rx_std_skbuff[idx].skb = skb;
		pci_unmap_addr_set(&ap->skb->rx_std_skbuff[idx],
				   mapping, mapping);

		rd = &ap->rx_std_ring[idx];
		set_aceaddr(&rd->addr, mapping);
		rd->size = ACE_STD_BUFSIZE;
		rd->idx = idx;
		idx = (idx + 1) % RX_STD_RING_ENTRIES;
	}

	if (!i)
		goto error_out;

	atomic_add(i, &ap->cur_rx_bufs);
	ap->rx_std_skbprd = idx;

	if (ACE_IS_TIGON_I(ap)) {
		struct cmd cmd;
		cmd.evt = C_SET_RX_PRD_IDX;
		cmd.code = 0;
		cmd.idx = ap->rx_std_skbprd;
		ace_issue_cmd(regs, &cmd);
	} else {
		writel(idx, &regs->RxStdPrd);
		wmb();
	}

 out:
	clear_bit(0, &ap->std_refill_busy);
	return;

 error_out:
	printk(KERN_INFO "Out of memory when allocating "
	       "standard receive buffers\n");
	goto out;
}


static void ace_load_mini_rx_ring(struct ace_private *ap, int nr_bufs)
{
	struct ace_regs __iomem *regs = ap->regs;
	short i, idx;

	prefetchw(&ap->cur_mini_bufs);

	idx = ap->rx_mini_skbprd;
	for (i = 0; i < nr_bufs; i++) {
		struct sk_buff *skb;
		struct rx_desc *rd;
		dma_addr_t mapping;

		skb = alloc_skb(ACE_MINI_BUFSIZE + NET_IP_ALIGN, GFP_ATOMIC);
		if (!skb)
			break;

		skb_reserve(skb, NET_IP_ALIGN);
		mapping = pci_map_page(ap->pdev, virt_to_page(skb->data),
				       offset_in_page(skb->data),
				       ACE_MINI_BUFSIZE,
				       PCI_DMA_FROMDEVICE);
		ap->skb->rx_mini_skbuff[idx].skb = skb;
		pci_unmap_addr_set(&ap->skb->rx_mini_skbuff[idx],
				   mapping, mapping);

		rd = &ap->rx_mini_ring[idx];
		set_aceaddr(&rd->addr, mapping);
		rd->size = ACE_MINI_BUFSIZE;
		rd->idx = idx;
		idx = (idx + 1) % RX_MINI_RING_ENTRIES;
	}

	if (!i)
		goto error_out;

	atomic_add(i, &ap->cur_mini_bufs);

	ap->rx_mini_skbprd = idx;

	writel(idx, &regs->RxMiniPrd);
	wmb();

 out:
	clear_bit(0, &ap->mini_refill_busy);
	return;
 error_out:
	printk(KERN_INFO "Out of memory when allocating "
	       "mini receive buffers\n");
	goto out;
}


/*
 * Load the jumbo rx ring, this may happen at any time if the MTU
 * is changed to a value > 1500.
 */
static void ace_load_jumbo_rx_ring(struct ace_private *ap, int nr_bufs)
{
	struct ace_regs __iomem *regs = ap->regs;
	short i, idx;

	idx = ap->rx_jumbo_skbprd;

	for (i = 0; i < nr_bufs; i++) {
		struct sk_buff *skb;
		struct rx_desc *rd;
		dma_addr_t mapping;

		skb = alloc_skb(ACE_JUMBO_BUFSIZE + NET_IP_ALIGN, GFP_ATOMIC);
		if (!skb)
			break;

		skb_reserve(skb, NET_IP_ALIGN);
		mapping = pci_map_page(ap->pdev, virt_to_page(skb->data),
				       offset_in_page(skb->data),
				       ACE_JUMBO_BUFSIZE,
				       PCI_DMA_FROMDEVICE);
		ap->skb->rx_jumbo_skbuff[idx].skb = skb;
		pci_unmap_addr_set(&ap->skb->rx_jumbo_skbuff[idx],
				   mapping, mapping);

		rd = &ap->rx_jumbo_ring[idx];
		set_aceaddr(&rd->addr, mapping);
		rd->size = ACE_JUMBO_BUFSIZE;
		rd->idx = idx;
		idx = (idx + 1) % RX_JUMBO_RING_ENTRIES;
	}

	if (!i)
		goto error_out;

	atomic_add(i, &ap->cur_jumbo_bufs);
	ap->rx_jumbo_skbprd = idx;

	if (ACE_IS_TIGON_I(ap)) {
		struct cmd cmd;
		cmd.evt = C_SET_RX_JUMBO_PRD_IDX;
		cmd.code = 0;
		cmd.idx = ap->rx_jumbo_skbprd;
		ace_issue_cmd(regs, &cmd);
	} else {
		writel(idx, &regs->RxJumboPrd);
		wmb();
	}

 out:
	clear_bit(0, &ap->jumbo_refill_busy);
	return;
 error_out:
	if (net_ratelimit())
		printk(KERN_INFO "Out of memory when allocating "
		       "jumbo receive buffers\n");
	goto out;
}


/*
 * All events are considered to be slow (RX/TX ints do not generate
 * events) and are handled here, outside the main interrupt handler,
 * to reduce the size of the handler.
 */
static u32 ace_handle_event(struct net_device *dev, u32 evtcsm, u32 evtprd)
{
	struct ace_private *ap;

	ap = netdev_priv(dev);

	while (evtcsm != evtprd) {
		switch (ap->evt_ring[evtcsm].evt) {
		case E_FW_RUNNING:
			printk(KERN_INFO "%s: Firmware up and running\n",
			       ap->name);
			ap->fw_running = 1;
			wmb();
			break;
		case E_STATS_UPDATED:
			break;
		case E_LNK_STATE:
		{
			u16 code = ap->evt_ring[evtcsm].code;
			switch (code) {
			case E_C_LINK_UP:
			{
				u32 state = readl(&ap->regs->GigLnkState);
				printk(KERN_WARNING "%s: Optical link UP "
				       "(%s Duplex, Flow Control: %s%s)\n",
				       ap->name,
				       state & LNK_FULL_DUPLEX ? "Full":"Half",
				       state & LNK_TX_FLOW_CTL_Y ? "TX " : "",
				       state & LNK_RX_FLOW_CTL_Y ? "RX" : "");
				break;
			}
			case E_C_LINK_DOWN:
				printk(KERN_WARNING "%s: Optical link DOWN\n",
				       ap->name);
				break;
			case E_C_LINK_10_100:
				printk(KERN_WARNING "%s: 10/100BaseT link "
				       "UP\n", ap->name);
				break;
			default:
				printk(KERN_ERR "%s: Unknown optical link "
				       "state %02x\n", ap->name, code);
			}
			break;
		}
		case E_ERROR:
			switch(ap->evt_ring[evtcsm].code) {
			case E_C_ERR_INVAL_CMD:
				printk(KERN_ERR "%s: invalid command error\n",
				       ap->name);
				break;
			case E_C_ERR_UNIMP_CMD:
				printk(KERN_ERR "%s: unimplemented command "
				       "error\n", ap->name);
				break;
			case E_C_ERR_BAD_CFG:
				printk(KERN_ERR "%s: bad config error\n",
				       ap->name);
				break;
			default:
				printk(KERN_ERR "%s: unknown error %02x\n",
				       ap->name, ap->evt_ring[evtcsm].code);
			}
			break;
		case E_RESET_JUMBO_RNG:
		{
			int i;
			for (i = 0; i < RX_JUMBO_RING_ENTRIES; i++) {
				if (ap->skb->rx_jumbo_skbuff[i].skb) {
					ap->rx_jumbo_ring[i].size = 0;
					set_aceaddr(&ap->rx_jumbo_ring[i].addr, 0);
					dev_kfree_skb(ap->skb->rx_jumbo_skbuff[i].skb);
					ap->skb->rx_jumbo_skbuff[i].skb = NULL;
				}
			}

 			if (ACE_IS_TIGON_I(ap)) {
 				struct cmd cmd;
 				cmd.evt = C_SET_RX_JUMBO_PRD_IDX;
 				cmd.code = 0;
 				cmd.idx = 0;
 				ace_issue_cmd(ap->regs, &cmd);
 			} else {
 				writel(0, &((ap->regs)->RxJumboPrd));
 				wmb();
 			}

			ap->jumbo = 0;
			ap->rx_jumbo_skbprd = 0;
			printk(KERN_INFO "%s: Jumbo ring flushed\n",
			       ap->name);
			clear_bit(0, &ap->jumbo_refill_busy);
			break;
		}
		default:
			printk(KERN_ERR "%s: Unhandled event 0x%02x\n",
			       ap->name, ap->evt_ring[evtcsm].evt);
		}
		evtcsm = (evtcsm + 1) % EVT_RING_ENTRIES;
	}

	return evtcsm;
}


static void ace_rx_int(struct net_device *dev, u32 rxretprd, u32 rxretcsm)
{
	struct ace_private *ap = netdev_priv(dev);
	u32 idx;
	int mini_count = 0, std_count = 0;

	idx = rxretcsm;

	prefetchw(&ap->cur_rx_bufs);
	prefetchw(&ap->cur_mini_bufs);
	
	while (idx != rxretprd) {
		struct ring_info *rip;
		struct sk_buff *skb;
		struct rx_desc *rxdesc, *retdesc;
		u32 skbidx;
		int bd_flags, desc_type, mapsize;
		u16 csum;


		/* make sure the rx descriptor isn't read before rxretprd */
		if (idx == rxretcsm) 
			rmb();

		retdesc = &ap->rx_return_ring[idx];
		skbidx = retdesc->idx;
		bd_flags = retdesc->flags;
		desc_type = bd_flags & (BD_FLG_JUMBO | BD_FLG_MINI);

		switch(desc_type) {
			/*
			 * Normal frames do not have any flags set
			 *
			 * Mini and normal frames arrive frequently,
			 * so use a local counter to avoid doing
			 * atomic operations for each packet arriving.
			 */
		case 0:
			rip = &ap->skb->rx_std_skbuff[skbidx];
			mapsize = ACE_STD_BUFSIZE;
			rxdesc = &ap->rx_std_ring[skbidx];
			std_count++;
			break;
		case BD_FLG_JUMBO:
			rip = &ap->skb->rx_jumbo_skbuff[skbidx];
			mapsize = ACE_JUMBO_BUFSIZE;
			rxdesc = &ap->rx_jumbo_ring[skbidx];
			atomic_dec(&ap->cur_jumbo_bufs);
			break;
		case BD_FLG_MINI:
			rip = &ap->skb->rx_mini_skbuff[skbidx];
			mapsize = ACE_MINI_BUFSIZE;
			rxdesc = &ap->rx_mini_ring[skbidx];
			mini_count++; 
			break;
		default:
			printk(KERN_INFO "%s: unknown frame type (0x%02x) "
			       "returned by NIC\n", dev->name,
			       retdesc->flags);
			goto error;
		}

		skb = rip->skb;
		rip->skb = NULL;
		pci_unmap_page(ap->pdev,
			       pci_unmap_addr(rip, mapping),
			       mapsize,
			       PCI_DMA_FROMDEVICE);
		skb_put(skb, retdesc->size);

		/*
		 * Fly baby, fly!
		 */
		csum = retdesc->tcp_udp_csum;

		skb->dev = dev;
		skb->protocol = eth_type_trans(skb, dev);

		/*
		 * Instead of forcing the poor tigon mips cpu to calculate
		 * pseudo hdr checksum, we do this ourselves.
		 */
		if (bd_flags & BD_FLG_TCP_UDP_SUM) {
			skb->csum = htons(csum);
			skb->ip_summed = CHECKSUM_HW;
		} else {
			skb->ip_summed = CHECKSUM_NONE;
		}

		/* send it up */
#if ACENIC_DO_VLAN
		if (ap->vlgrp && (bd_flags & BD_FLG_VLAN_TAG)) {
			vlan_hwaccel_rx(skb, ap->vlgrp, retdesc->vlan);
		} else
#endif
			netif_rx(skb);

		dev->last_rx = jiffies;
		ap->stats.rx_packets++;
		ap->stats.rx_bytes += retdesc->size;

		idx = (idx + 1) % RX_RETURN_RING_ENTRIES;
	}

	atomic_sub(std_count, &ap->cur_rx_bufs);
	if (!ACE_IS_TIGON_I(ap))
		atomic_sub(mini_count, &ap->cur_mini_bufs);

 out:
	/*
	 * According to the documentation RxRetCsm is obsolete with
	 * the 12.3.x Firmware - my Tigon I NICs seem to disagree!
	 */
	if (ACE_IS_TIGON_I(ap)) {
		writel(idx, &ap->regs->RxRetCsm);
	}
	ap->cur_rx = idx;

	return;
 error:
	idx = rxretprd;
	goto out;
}


static inline void ace_tx_int(struct net_device *dev,
			      u32 txcsm, u32 idx)
{
	struct ace_private *ap = netdev_priv(dev);

	do {
		struct sk_buff *skb;
		dma_addr_t mapping;
		struct tx_ring_info *info;

		info = ap->skb->tx_skbuff + idx;
		skb = info->skb;
		mapping = pci_unmap_addr(info, mapping);

		if (mapping) {
			pci_unmap_page(ap->pdev, mapping,
				       pci_unmap_len(info, maplen),
				       PCI_DMA_TODEVICE);
			pci_unmap_addr_set(info, mapping, 0);
		}

		if (skb) {
			ap->stats.tx_packets++;
			ap->stats.tx_bytes += skb->len;
			dev_kfree_skb_irq(skb);
			info->skb = NULL;
		}

		idx = (idx + 1) % ACE_TX_RING_ENTRIES(ap);
	} while (idx != txcsm);

	if (netif_queue_stopped(dev))
		netif_wake_queue(dev);

	wmb();
	ap->tx_ret_csm = txcsm;

	/* So... tx_ret_csm is advanced _after_ check for device wakeup.
	 *
	 * We could try to make it before. In this case we would get
	 * the following race condition: hard_start_xmit on other cpu
	 * enters after we advanced tx_ret_csm and fills space,
	 * which we have just freed, so that we make illegal device wakeup.
	 * There is no good way to workaround this (at entry
	 * to ace_start_xmit detects this condition and prevents
	 * ring corruption, but it is not a good workaround.)
	 *
	 * When tx_ret_csm is advanced after, we wake up device _only_
	 * if we really have some space in ring (though the core doing
	 * hard_start_xmit can see full ring for some period and has to
	 * synchronize.) Superb.
	 * BUT! We get another subtle race condition. hard_start_xmit
	 * may think that ring is full between wakeup and advancing
	 * tx_ret_csm and will stop device instantly! It is not so bad.
	 * We are guaranteed that there is something in ring, so that
	 * the next irq will resume transmission. To speedup this we could
	 * mark descriptor, which closes ring with BD_FLG_COAL_NOW
	 * (see ace_start_xmit).
	 *
	 * Well, this dilemma exists in all lock-free devices.
	 * We, following scheme used in drivers by Donald Becker,
	 * select the least dangerous.
	 *							--ANK
	 */
}


static irqreturn_t ace_interrupt(int irq, void *dev_id, struct pt_regs *ptregs)
{
	struct net_device *dev = (struct net_device *)dev_id;
	struct ace_private *ap = netdev_priv(dev);
	struct ace_regs __iomem *regs = ap->regs;
	u32 idx;
	u32 txcsm, rxretcsm, rxretprd;
	u32 evtcsm, evtprd;

	/*
	 * In case of PCI shared interrupts or spurious interrupts,
	 * we want to make sure it is actually our interrupt before
	 * spending any time in here.
	 */
	if (!(readl(&regs->HostCtrl) & IN_INT))
		return IRQ_NONE;

	/*
	 * ACK intr now. Otherwise we will lose updates to rx_ret_prd,
	 * which happened _after_ rxretprd = *ap->rx_ret_prd; but before
	 * writel(0, &regs->Mb0Lo).
	 *
	 * "IRQ avoidance" recommended in docs applies to IRQs served
	 * threads and it is wrong even for that case.
	 */
	writel(0, &regs->Mb0Lo);
	readl(&regs->Mb0Lo);

	/*
	 * There is no conflict between transmit handling in
	 * start_xmit and receive processing, thus there is no reason
	 * to take a spin lock for RX handling. Wait until we start
	 * working on the other stuff - hey we don't need a spin lock
	 * anymore.
	 */
	rxretprd = *ap->rx_ret_prd;
	rxretcsm = ap->cur_rx;

	if (rxretprd != rxretcsm)
		ace_rx_int(dev, rxretprd, rxretcsm);

	txcsm = *ap->tx_csm;
	idx = ap->tx_ret_csm;

	if (txcsm != idx) {
		/*
		 * If each skb takes only one descriptor this check degenerates
		 * to identity, because new space has just been opened.
		 * But if skbs are fragmented we must check that this index
		 * update releases enough of space, otherwise we just
		 * wait for device to make more work.
		 */
		if (!tx_ring_full(ap, txcsm, ap->tx_prd))
			ace_tx_int(dev, txcsm, idx);
	}

	evtcsm = readl(&regs->EvtCsm);
	evtprd = *ap->evt_prd;

	if (evtcsm != evtprd) {
		evtcsm = ace_handle_event(dev, evtcsm, evtprd);
		writel(evtcsm, &regs->EvtCsm);
	}

	/*
	 * This has to go last in the interrupt handler and run with
	 * the spin lock released ... what lock?
	 */
	if (netif_running(dev)) {
		int cur_size;
		int run_tasklet = 0;

		cur_size = atomic_read(&ap->cur_rx_bufs);
		if (cur_size < RX_LOW_STD_THRES) {
			if ((cur_size < RX_PANIC_STD_THRES) &&
			    !test_and_set_bit(0, &ap->std_refill_busy)) {
#ifdef DEBUG
				printk("low on std buffers %i\n", cur_size);
#endif
				ace_load_std_rx_ring(ap,
						     RX_RING_SIZE - cur_size);
			} else
				run_tasklet = 1;
		}

		if (!ACE_IS_TIGON_I(ap)) {
			cur_size = atomic_read(&ap->cur_mini_bufs);
			if (cur_size < RX_LOW_MINI_THRES) {
				if ((cur_size < RX_PANIC_MINI_THRES) &&
				    !test_and_set_bit(0,
						      &ap->mini_refill_busy)) {
#ifdef DEBUG
					printk("low on mini buffers %i\n",
					       cur_size);
#endif
					ace_load_mini_rx_ring(ap, RX_MINI_SIZE - cur_size);
				} else
					run_tasklet = 1;
			}
		}

		if (ap->jumbo) {
			cur_size = atomic_read(&ap->cur_jumbo_bufs);
			if (cur_size < RX_LOW_JUMBO_THRES) {
				if ((cur_size < RX_PANIC_JUMBO_THRES) &&
				    !test_and_set_bit(0,
						      &ap->jumbo_refill_busy)){
#ifdef DEBUG
					printk("low on jumbo buffers %i\n",
					       cur_size);
#endif
					ace_load_jumbo_rx_ring(ap, RX_JUMBO_SIZE - cur_size);
				} else
					run_tasklet = 1;
			}
		}
		if (run_tasklet && !ap->tasklet_pending) {
			ap->tasklet_pending = 1;
			tasklet_schedule(&ap->ace_tasklet);
		}
	}

	return IRQ_HANDLED;
}


#if ACENIC_DO_VLAN
static void ace_vlan_rx_register(struct net_device *dev, struct vlan_group *grp)
{
	struct ace_private *ap = netdev_priv(dev);
	unsigned long flags;

	local_irq_save(flags);
	ace_mask_irq(dev);

	ap->vlgrp = grp;

	ace_unmask_irq(dev);
	local_irq_restore(flags);
}


static void ace_vlan_rx_kill_vid(struct net_device *dev, unsigned short vid)
{
	struct ace_private *ap = netdev_priv(dev);
	unsigned long flags;

	local_irq_save(flags);
	ace_mask_irq(dev);

	if (ap->vlgrp)
		ap->vlgrp->vlan_devices[vid] = NULL;

	ace_unmask_irq(dev);
	local_irq_restore(flags);
}
#endif /* ACENIC_DO_VLAN */


static int ace_open(struct net_device *dev)
{
	struct ace_private *ap = netdev_priv(dev);
	struct ace_regs __iomem *regs = ap->regs;
	struct cmd cmd;

	if (!(ap->fw_running)) {
		printk(KERN_WARNING "%s: Firmware not running!\n", dev->name);
		return -EBUSY;
	}

	writel(dev->mtu + ETH_HLEN + 4, &regs->IfMtu);

	cmd.evt = C_CLEAR_STATS;
	cmd.code = 0;
	cmd.idx = 0;
	ace_issue_cmd(regs, &cmd);

	cmd.evt = C_HOST_STATE;
	cmd.code = C_C_STACK_UP;
	cmd.idx = 0;
	ace_issue_cmd(regs, &cmd);

	if (ap->jumbo &&
	    !test_and_set_bit(0, &ap->jumbo_refill_busy))
		ace_load_jumbo_rx_ring(ap, RX_JUMBO_SIZE);

	if (dev->flags & IFF_PROMISC) {
		cmd.evt = C_SET_PROMISC_MODE;
		cmd.code = C_C_PROMISC_ENABLE;
		cmd.idx = 0;
		ace_issue_cmd(regs, &cmd);

		ap->promisc = 1;
	}else
		ap->promisc = 0;
	ap->mcast_all = 0;

#if 0
	cmd.evt = C_LNK_NEGOTIATION;
	cmd.code = 0;
	cmd.idx = 0;
	ace_issue_cmd(regs, &cmd);
#endif

	netif_start_queue(dev);

	/*
	 * Setup the bottom half rx ring refill handler
	 */
	tasklet_init(&ap->ace_tasklet, ace_tasklet, (unsigned long)dev);
	return 0;
}


static int ace_close(struct net_device *dev)
{
	struct ace_private *ap = netdev_priv(dev);
	struct ace_regs __iomem *regs = ap->regs;
	struct cmd cmd;
	unsigned long flags;
	short i;

	/*
	 * Without (or before) releasing irq and stopping hardware, this
	 * is an absolute non-sense, by the way. It will be reset instantly
	 * by the first irq.
	 */
	netif_stop_queue(dev);

	
	if (ap->promisc) {
		cmd.evt = C_SET_PROMISC_MODE;
		cmd.code = C_C_PROMISC_DISABLE;
		cmd.idx = 0;
		ace_issue_cmd(regs, &cmd);
		ap->promisc = 0;
	}

	cmd.evt = C_HOST_STATE;
	cmd.code = C_C_STACK_DOWN;
	cmd.idx = 0;
	ace_issue_cmd(regs, &cmd);

	tasklet_kill(&ap->ace_tasklet);

	/*
	 * Make sure one CPU is not processing packets while
	 * buffers are being released by another.
	 */

	local_irq_save(flags);
	ace_mask_irq(dev);

	for (i = 0; i < ACE_TX_RING_ENTRIES(ap); i++) {
		struct sk_buff *skb;
		dma_addr_t mapping;
		struct tx_ring_info *info;

		info = ap->skb->tx_skbuff + i;
		skb = info->skb;
		mapping = pci_unmap_addr(info, mapping);

		if (mapping) {
			if (ACE_IS_TIGON_I(ap)) {
				struct tx_desc __iomem *tx 
					= (struct tx_desc __iomem *) &ap->tx_ring[i];
				writel(0, &tx->addr.addrhi);
				writel(0, &tx->addr.addrlo);
				writel(0, &tx->flagsize);
			} else
				memset(ap->tx_ring + i, 0,
				       sizeof(struct tx_desc));
			pci_unmap_page(ap->pdev, mapping,
				       pci_unmap_len(info, maplen),
				       PCI_DMA_TODEVICE);
			pci_unmap_addr_set(info, mapping, 0);
		}
		if (skb) {
			dev_kfree_skb(skb);
			info->skb = NULL;
		}
	}

	if (ap->jumbo) {
		cmd.evt = C_RESET_JUMBO_RNG;
		cmd.code = 0;
		cmd.idx = 0;
		ace_issue_cmd(regs, &cmd);
	}

	ace_unmask_irq(dev);
	local_irq_restore(flags);

	return 0;
}


static inline dma_addr_t
ace_map_tx_skb(struct ace_private *ap, struct sk_buff *skb,
	       struct sk_buff *tail, u32 idx)
{
	dma_addr_t mapping;
	struct tx_ring_info *info;

	mapping = pci_map_page(ap->pdev, virt_to_page(skb->data),
			       offset_in_page(skb->data),
			       skb->len, PCI_DMA_TODEVICE);

	info = ap->skb->tx_skbuff + idx;
	info->skb = tail;
	pci_unmap_addr_set(info, mapping, mapping);
	pci_unmap_len_set(info, maplen, skb->len);
	return mapping;
}


static inline void
ace_load_tx_bd(struct ace_private *ap, struct tx_desc *desc, u64 addr,
	       u32 flagsize, u32 vlan_tag)
{
#if !USE_TX_COAL_NOW
	flagsize &= ~BD_FLG_COAL_NOW;
#endif

	if (ACE_IS_TIGON_I(ap)) {
		struct tx_desc __iomem *io = (struct tx_desc __iomem *) desc;
		writel(addr >> 32, &io->addr.addrhi);
		writel(addr & 0xffffffff, &io->addr.addrlo);
		writel(flagsize, &io->flagsize);
#if ACENIC_DO_VLAN
		writel(vlan_tag, &io->vlanres);
#endif
	} else {
		desc->addr.addrhi = addr >> 32;
		desc->addr.addrlo = addr;
		desc->flagsize = flagsize;
#if ACENIC_DO_VLAN
		desc->vlanres = vlan_tag;
#endif
	}
}


static int ace_start_xmit(struct sk_buff *skb, struct net_device *dev)
{
	struct ace_private *ap = netdev_priv(dev);
	struct ace_regs __iomem *regs = ap->regs;
	struct tx_desc *desc;
	u32 idx, flagsize;
	unsigned long maxjiff = jiffies + 3*HZ;

restart:
	idx = ap->tx_prd;

	if (tx_ring_full(ap, ap->tx_ret_csm, idx))
		goto overflow;

	if (!skb_shinfo(skb)->nr_frags)	{
		dma_addr_t mapping;
		u32 vlan_tag = 0;

		mapping = ace_map_tx_skb(ap, skb, skb, idx);
		flagsize = (skb->len << 16) | (BD_FLG_END);
		if (skb->ip_summed == CHECKSUM_HW)
			flagsize |= BD_FLG_TCP_UDP_SUM;
#if ACENIC_DO_VLAN
		if (vlan_tx_tag_present(skb)) {
			flagsize |= BD_FLG_VLAN_TAG;
			vlan_tag = vlan_tx_tag_get(skb);
		}
#endif
		desc = ap->tx_ring + idx;
		idx = (idx + 1) % ACE_TX_RING_ENTRIES(ap);

		/* Look at ace_tx_int for explanations. */
		if (tx_ring_full(ap, ap->tx_ret_csm, idx))
			flagsize |= BD_FLG_COAL_NOW;

		ace_load_tx_bd(ap, desc, mapping, flagsize, vlan_tag);
	} else {
		dma_addr_t mapping;
		u32 vlan_tag = 0;
		int i, len = 0;

		mapping = ace_map_tx_skb(ap, skb, NULL, idx);
		flagsize = (skb_headlen(skb) << 16);
		if (skb->ip_summed == CHECKSUM_HW)
			flagsize |= BD_FLG_TCP_UDP_SUM;
#if ACENIC_DO_VLAN
		if (vlan_tx_tag_present(skb)) {
			flagsize |= BD_FLG_VLAN_TAG;
			vlan_tag = vlan_tx_tag_get(skb);
		}
#endif

		ace_load_tx_bd(ap, ap->tx_ring + idx, mapping, flagsize, vlan_tag);

		idx = (idx + 1) % ACE_TX_RING_ENTRIES(ap);

		for (i = 0; i < skb_shinfo(skb)->nr_frags; i++) {
			skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
			struct tx_ring_info *info;

			len += frag->size;
			info = ap->skb->tx_skbuff + idx;
			desc = ap->tx_ring + idx;

			mapping = pci_map_page(ap->pdev, frag->page,
					       frag->page_offset, frag->size,
					       PCI_DMA_TODEVICE);

			flagsize = (frag->size << 16);
			if (skb->ip_summed == CHECKSUM_HW)
				flagsize |= BD_FLG_TCP_UDP_SUM;
			idx = (idx + 1) % ACE_TX_RING_ENTRIES(ap);

			if (i == skb_shinfo(skb)->nr_frags - 1) {
				flagsize |= BD_FLG_END;
				if (tx_ring_full(ap, ap->tx_ret_csm, idx))
					flagsize |= BD_FLG_COAL_NOW;

				/*
				 * Only the last fragment frees
				 * the skb!
				 */
				info->skb = skb;
			} else {
				info->skb = NULL;
			}
			pci_unmap_addr_set(info, mapping, mapping);
			pci_unmap_len_set(info, maplen, frag->size);
			ace_load_tx_bd(ap, desc, mapping, flagsize, vlan_tag);
		}
	}

 	wmb();
 	ap->tx_prd = idx;
 	ace_set_txprd(regs, ap, idx);

	if (flagsize & BD_FLG_COAL_NOW) {
		netif_stop_queue(dev);

		/*
		 * A TX-descriptor producer (an IRQ) might have gotten
		 * inbetween, making the ring free again. Since xmit is
		 * serialized, this is the only situation we have to
		 * re-test.
		 */
		if (!tx_ring_full(ap, ap->tx_ret_csm, idx))
			netif_wake_queue(dev);
	}

	dev->trans_start = jiffies;
	return NETDEV_TX_OK;

overflow:
	/*
	 * This race condition is unavoidable with lock-free drivers.
	 * We wake up the queue _before_ tx_prd is advanced, so that we can
	 * enter hard_start_xmit too early, while tx ring still looks closed.
	 * This happens ~1-4 times per 100000 packets, so that we can allow
	 * to loop syncing to other CPU. Probably, we need an additional
	 * wmb() in ace_tx_intr as well.
	 *
	 * Note that this race is relieved by reserving one more entry
	 * in tx ring than it is necessary (see original non-SG driver).
	 * However, with SG we need to reserve 2*MAX_SKB_FRAGS+1, which
	 * is already overkill.
	 *
	 * Alternative is to return with 1 not throttling queue. In this
	 * case loop becomes longer, no more useful effects.
	 */
	if (time_before(jiffies, maxjiff)) {
		barrier();
		cpu_relax();
		goto restart;
	}
	
	/* The ring is stuck full. */
	printk(KERN_WARNING "%s: Transmit ring stuck full\n", dev->name);
	return NETDEV_TX_BUSY;
}


static int ace_change_mtu(struct net_device *dev, int new_mtu)
{
	struct ace_private *ap = netdev_priv(dev);
	struct ace_regs __iomem *regs = ap->regs;

	if (new_mtu > ACE_JUMBO_MTU)
		return -EINVAL;

	writel(new_mtu + ETH_HLEN + 4, &regs->IfMtu);
	dev->mtu = new_mtu;

	if (new_mtu > ACE_STD_MTU) {
		if (!(ap->jumbo)) {
			printk(KERN_INFO "%s: Enabling Jumbo frame "
			       "support\n", dev->name);
			ap->jumbo = 1;
			if (!test_and_set_bit(0, &ap->jumbo_refill_busy))
				ace_load_jumbo_rx_ring(ap, RX_JUMBO_SIZE);
			ace_set_rxtx_parms(dev, 1);
		}
	} else {
		while (test_and_set_bit(0, &ap->jumbo_refill_busy));
		ace_sync_irq(dev->irq);
		ace_set_rxtx_parms(dev, 0);
		if (ap->jumbo) {
			struct cmd cmd;

			cmd.evt = C_RESET_JUMBO_RNG;
			cmd.code = 0;
			cmd.idx = 0;
			ace_issue_cmd(regs, &cmd);
		}
	}

	return 0;
}

static int ace_get_settings(struct net_device *dev, struct ethtool_cmd *ecmd)
{
	struct ace_private *ap = netdev_priv(dev);
	struct ace_regs __iomem *regs = ap->regs;
	u32 link;

	memset(ecmd, 0, sizeof(struct ethtool_cmd));
	ecmd->supported =
		(SUPPORTED_10baseT_Half | SUPPORTED_10baseT_Full |
		 SUPPORTED_100baseT_Half | SUPPORTED_100baseT_Full |
		 SUPPORTED_1000baseT_Half | SUPPORTED_1000baseT_Full |
		 SUPPORTED_Autoneg | SUPPORTED_FIBRE);

	ecmd->port = PORT_FIBRE;
	ecmd->transceiver = XCVR_INTERNAL;

	link = readl(&regs->GigLnkState);
	if (link & LNK_1000MB)
		ecmd->speed = SPEED_1000;
	else {
		link = readl(&regs->FastLnkState);
		if (link & LNK_100MB)
			ecmd->speed = SPEED_100;
		else if (link & LNK_10MB)
			ecmd->speed = SPEED_10;
		else
			ecmd->speed = 0;
	}
	if (link & LNK_FULL_DUPLEX)
		ecmd->duplex = DUPLEX_FULL;
	else
		ecmd->duplex = DUPLEX_HALF;

	if (link & LNK_NEGOTIATE)
		ecmd->autoneg = AUTONEG_ENABLE;
	else
		ecmd->autoneg = AUTONEG_DISABLE;

#if 0
	/*
	 * Current struct ethtool_cmd is insufficient
	 */
	ecmd->trace = readl(&regs->TuneTrace);

	ecmd->txcoal = readl(&regs->TuneTxCoalTicks);
	ecmd->rxcoal = readl(&regs->TuneRxCoalTicks);
#endif
	ecmd->maxtxpkt = readl(&regs->TuneMaxTxDesc);
	ecmd->maxrxpkt = readl(&regs->TuneMaxRxDesc);

	return 0;
}

static int ace_set_settings(struct net_device *dev, struct ethtool_cmd *ecmd)
{
	struct ace_private *ap = netdev_priv(dev);
	struct ace_regs __iomem *regs = ap->regs;
	u32 link, speed;

	link = readl(&regs->GigLnkState);
	if (link & LNK_1000MB)
		speed = SPEED_1000;
	else {
		link = readl(&regs->FastLnkState);
		if (link & LNK_100MB)
			speed = SPEED_100;
		else if (link & LNK_10MB)
			speed = SPEED_10;
		else
			speed = SPEED_100;
	}

	link = LNK_ENABLE | LNK_1000MB | LNK_100MB | LNK_10MB |
		LNK_RX_FLOW_CTL_Y | LNK_NEG_FCTL;
	if (!ACE_IS_TIGON_I(ap))
		link |= LNK_TX_FLOW_CTL_Y;
	if (ecmd->autoneg == AUTONEG_ENABLE)
		link |= LNK_NEGOTIATE;
	if (ecmd->speed != speed) {
		link &= ~(LNK_1000MB | LNK_100MB | LNK_10MB);
		switch (speed) {
		case SPEED_1000:
			link |= LNK_1000MB;
			break;
		case SPEED_100:
			link |= LNK_100MB;
			break;
		case SPEED_10:
			link |= LNK_10MB;
			break;
		}
	}

	if (ecmd->duplex == DUPLEX_FULL)
		link |= LNK_FULL_DUPLEX;

	if (link != ap->link) {
		struct cmd cmd;
		printk(KERN_INFO "%s: Renegotiating link state\n",
		       dev->name);

		ap->link = link;
		writel(link, &regs->TuneLink);
		if (!ACE_IS_TIGON_I(ap))
			writel(link, &regs->TuneFastLink);
		wmb();

		cmd.evt = C_LNK_NEGOTIATION;
		cmd.code = 0;
		cmd.idx = 0;
		ace_issue_cmd(regs, &cmd);
	}
	return 0;
}

static void ace_get_drvinfo(struct net_device *dev, 
			    struct ethtool_drvinfo *info)
{
	struct ace_private *ap = netdev_priv(dev);

	strlcpy(info->driver, "acenic", sizeof(info->driver));
	snprintf(info->version, sizeof(info->version), "%i.%i.%i", 
		tigonFwReleaseMajor, tigonFwReleaseMinor,
		tigonFwReleaseFix);

	if (ap->pdev)
		strlcpy(info->bus_info, pci_name(ap->pdev), 
			sizeof(info->bus_info));

}

/*
 * Set the hardware MAC address.
 */
static int ace_set_mac_addr(struct net_device *dev, void *p)
{
	struct ace_private *ap = netdev_priv(dev);
	struct ace_regs __iomem *regs = ap->regs;
	struct sockaddr *addr=p;
	u8 *da;
	struct cmd cmd;

	if(netif_running(dev))
		return -EBUSY;

	memcpy(dev->dev_addr, addr->sa_data,dev->addr_len);

	da = (u8 *)dev->dev_addr;

	writel(da[0] << 8 | da[1], &regs->MacAddrHi);
	writel((da[2] << 24) | (da[3] << 16) | (da[4] << 8) | da[5],
	       &regs->MacAddrLo);

	cmd.evt = C_SET_MAC_ADDR;
	cmd.code = 0;
	cmd.idx = 0;
	ace_issue_cmd(regs, &cmd);

	return 0;
}


static void ace_set_multicast_list(struct net_device *dev)
{
	struct ace_private *ap = netdev_priv(dev);
	struct ace_regs __iomem *regs = ap->regs;
	struct cmd cmd;

	if ((dev->flags & IFF_ALLMULTI) && !(ap->mcast_all)) {
		cmd.evt = C_SET_MULTICAST_MODE;
		cmd.code = C_C_MCAST_ENABLE;
		cmd.idx = 0;
		ace_issue_cmd(regs, &cmd);
		ap->mcast_all = 1;
	} else if (ap->mcast_all) {
		cmd.evt = C_SET_MULTICAST_MODE;
		cmd.code = C_C_MCAST_DISABLE;
		cmd.idx = 0;
		ace_issue_cmd(regs, &cmd);
		ap->mcast_all = 0;
	}

	if ((dev->flags & IFF_PROMISC) && !(ap->promisc)) {
		cmd.evt = C_SET_PROMISC_MODE;
		cmd.code = C_C_PROMISC_ENABLE;
		cmd.idx = 0;
		ace_issue_cmd(regs, &cmd);
		ap->promisc = 1;
	}else if (!(dev->flags & IFF_PROMISC) && (ap->promisc)) {
		cmd.evt = C_SET_PROMISC_MODE;
		cmd.code = C_C_PROMISC_DISABLE;
		cmd.idx = 0;
		ace_issue_cmd(regs, &cmd);
		ap->promisc = 0;
	}

	/*
	 * For the time being multicast relies on the upper layers
	 * filtering it properly. The Firmware does not allow one to
	 * set the entire multicast list at a time and keeping track of
	 * it here is going to be messy.
	 */
	if ((dev->mc_count) && !(ap->mcast_all)) {
		cmd.evt = C_SET_MULTICAST_MODE;
		cmd.code = C_C_MCAST_ENABLE;
		cmd.idx = 0;
		ace_issue_cmd(regs, &cmd);
	}else if (!ap->mcast_all) {
		cmd.evt = C_SET_MULTICAST_MODE;
		cmd.code = C_C_MCAST_DISABLE;
		cmd.idx = 0;
		ace_issue_cmd(regs, &cmd);
	}
}


static struct net_device_stats *ace_get_stats(struct net_device *dev)
{
	struct ace_private *ap = netdev_priv(dev);
	struct ace_mac_stats __iomem *mac_stats =
		(struct ace_mac_stats __iomem *)ap->regs->Stats;

	ap->stats.rx_missed_errors = readl(&mac_stats->drop_space);
	ap->stats.multicast = readl(&mac_stats->kept_mc);
	ap->stats.collisions = readl(&mac_stats->coll);

	return &ap->stats;
}


static void __devinit ace_copy(struct ace_regs __iomem *regs, void *src,
			    u32 dest, int size)
{
	void __iomem *tdest;
	u32 *wsrc;
	short tsize, i;

	if (size <= 0)
		return;

	while (size > 0) {
		tsize = min_t(u32, ((~dest & (ACE_WINDOW_SIZE - 1)) + 1),
			    min_t(u32, size, ACE_WINDOW_SIZE));
		tdest = (void __iomem *) &regs->Window + 
			(dest & (ACE_WINDOW_SIZE - 1));
		writel(dest & ~(ACE_WINDOW_SIZE - 1), &regs->WinBase);
		/*
		 * This requires byte swapping on big endian, however
		 * writel does that for us
		 */
		wsrc = src;
		for (i = 0; i < (tsize / 4); i++) {
			writel(wsrc[i], tdest + i*4);
		}
		dest += tsize;
		src += tsize;
		size -= tsize;
	}

	return;
}


static void __devinit ace_clear(struct ace_regs __iomem *regs, u32 dest, int size)
{
	void __iomem *tdest;
	short tsize = 0, i;

	if (size <= 0)
		return;

	while (size > 0) {
		tsize = min_t(u32, ((~dest & (ACE_WINDOW_SIZE - 1)) + 1),
				min_t(u32, size, ACE_WINDOW_SIZE));
		tdest = (void __iomem *) &regs->Window + 
			(dest & (ACE_WINDOW_SIZE - 1));
		writel(dest & ~(ACE_WINDOW_SIZE - 1), &regs->WinBase);

		for (i = 0; i < (tsize / 4); i++) {
			writel(0, tdest + i*4);
		}

		dest += tsize;
		size -= tsize;
	}

	return;
}


/*
 * Download the firmware into the SRAM on the NIC
 *
 * This operation requires the NIC to be halted and is performed with
 * interrupts disabled and with the spinlock hold.
 */
int __devinit ace_load_firmware(struct net_device *dev)
{
	struct ace_private *ap = netdev_priv(dev);
	struct ace_regs __iomem *regs = ap->regs;

	if (!(readl(&regs->CpuCtrl) & CPU_HALTED)) {
		printk(KERN_ERR "%s: trying to download firmware while the "
		       "CPU is running!\n", ap->name);
		return -EFAULT;
	}

	/*
	 * Do not try to clear more than 512KB or we end up seeing
	 * funny things on NICs with only 512KB SRAM
	 */
	ace_clear(regs, 0x2000, 0x80000-0x2000);
	if (ACE_IS_TIGON_I(ap)) {
		ace_copy(regs, tigonFwText, tigonFwTextAddr, tigonFwTextLen);
		ace_copy(regs, tigonFwData, tigonFwDataAddr, tigonFwDataLen);
		ace_copy(regs, tigonFwRodata, tigonFwRodataAddr,
			 tigonFwRodataLen);
		ace_clear(regs, tigonFwBssAddr, tigonFwBssLen);
		ace_clear(regs, tigonFwSbssAddr, tigonFwSbssLen);
	}else if (ap->version == 2) {
		ace_clear(regs, tigon2FwBssAddr, tigon2FwBssLen);
		ace_clear(regs, tigon2FwSbssAddr, tigon2FwSbssLen);
		ace_copy(regs, tigon2FwText, tigon2FwTextAddr,tigon2FwTextLen);
		ace_copy(regs, tigon2FwRodata, tigon2FwRodataAddr,
			 tigon2FwRodataLen);
		ace_copy(regs, tigon2FwData, tigon2FwDataAddr,tigon2FwDataLen);
	}

	return 0;
}


/*
 * The eeprom on the AceNIC is an Atmel i2c EEPROM.
 *
 * Accessing the EEPROM is `interesting' to say the least - don't read
 * this code right after dinner.
 *
 * This is all about black magic and bit-banging the device .... I
 * wonder in what hospital they have put the guy who designed the i2c
 * specs.
 *
 * Oh yes, this is only the beginning!
 *
 * Thanks to Stevarino Webinski for helping tracking down the bugs in the
 * code i2c readout code by beta testing all my hacks.
 */
static void __devinit eeprom_start(struct ace_regs __iomem *regs)
{
	u32 local;

	readl(&regs->LocalCtrl);
	udelay(ACE_SHORT_DELAY);
	local = readl(&regs->LocalCtrl);
	local |= EEPROM_DATA_OUT | EEPROM_WRITE_ENABLE;
	writel(local, &regs->LocalCtrl);
	readl(&regs->LocalCtrl);
	mb();
	udelay(ACE_SHORT_DELAY);
	local |= EEPROM_CLK_OUT;
	writel(local, &regs->LocalCtrl);
	readl(&regs->LocalCtrl);
	mb();
	udelay(ACE_SHORT_DELAY);
	local &= ~EEPROM_DATA_OUT;
	writel(local, &regs->LocalCtrl);
	readl(&regs->LocalCtrl);
	mb();
	udelay(ACE_SHORT_DELAY);
	local &= ~EEPROM_CLK_OUT;
	writel(local, &regs->LocalCtrl);
	readl(&regs->LocalCtrl);
	mb();
}


static void __devinit eeprom_prep(struct ace_regs __iomem *regs, u8 magic)
{
	short i;
	u32 local;

	udelay(ACE_SHORT_DELAY);
	local = readl(&regs->LocalCtrl);
	local &= ~EEPROM_DATA_OUT;
	local |= EEPROM_WRITE_ENABLE;
	writel(local, &regs->LocalCtrl);
	readl(&regs->LocalCtrl);
	mb();

	for (i = 0; i < 8; i++, magic <<= 1) {
		udelay(ACE_SHORT_DELAY);
		if (magic & 0x80) 
			local |= EEPROM_DATA_OUT;
		else
			local &= ~EEPROM_DATA_OUT;
		writel(local, &regs->LocalCtrl);
		readl(&regs->LocalCtrl);
		mb();

		udelay(ACE_SHORT_DELAY);
		local |= EEPROM_CLK_OUT;
		writel(local, &regs->LocalCtrl);
		readl(&regs->LocalCtrl);
		mb();
		udelay(ACE_SHORT_DELAY);
		local &= ~(EEPROM_CLK_OUT | EEPROM_DATA_OUT);
		writel(local, &regs->LocalCtrl);
		readl(&regs->LocalCtrl);
		mb();
	}
}


static int __devinit eeprom_check_ack(struct ace_regs __iomem *regs)
{
	int state;
	u32 local;

	local = readl(&regs->LocalCtrl);
	local &= ~EEPROM_WRITE_ENABLE;
	writel(local, &regs->LocalCtrl);
	readl(&regs->LocalCtrl);
	mb();
	udelay(ACE_LONG_DELAY);
	local |= EEPROM_CLK_OUT;
	writel(local, &regs->LocalCtrl);
	readl(&regs->LocalCtrl);
	mb();
	udelay(ACE_SHORT_DELAY);
	/* sample data in middle of high clk */
	state = (readl(&regs->LocalCtrl) & EEPROM_DATA_IN) != 0;
	udelay(ACE_SHORT_DELAY);
	mb();
	writel(readl(&regs->LocalCtrl) & ~EEPROM_CLK_OUT, &regs->LocalCtrl);
	readl(&regs->LocalCtrl);
	mb();

	return state;
}


static void __devinit eeprom_stop(struct ace_regs __iomem *regs)
{
	u32 local;

	udelay(ACE_SHORT_DELAY);
	local = readl(&regs->LocalCtrl);
	local |= EEPROM_WRITE_ENABLE;
	writel(local, &regs->LocalCtrl);
	readl(&regs->LocalCtrl);
	mb();
	udelay(ACE_SHORT_DELAY);
	local &= ~EEPROM_DATA_OUT;
	writel(local, &regs->LocalCtrl);
	readl(&regs->LocalCtrl);
	mb();
	udelay(ACE_SHORT_DELAY);
	local |= EEPROM_CLK_OUT;
	writel(local, &regs->LocalCtrl);
	readl(&regs->LocalCtrl);
	mb();
	udelay(ACE_SHORT_DELAY);
	local |= EEPROM_DATA_OUT;
	writel(local, &regs->LocalCtrl);
	readl(&regs->LocalCtrl);
	mb();
	udelay(ACE_LONG_DELAY);
	local &= ~EEPROM_CLK_OUT;
	writel(local, &regs->LocalCtrl);
	mb();
}


/*
 * Read a whole byte from the EEPROM.
 */
static int __devinit read_eeprom_byte(struct net_device *dev,
				   unsigned long offset)
{
	struct ace_private *ap = netdev_priv(dev);
	struct ace_regs __iomem *regs = ap->regs;
	unsigned long flags;
	u32 local;
	int result = 0;
	short i;

	if (!dev) {
		printk(KERN_ERR "No device!\n");
		result = -ENODEV;
		goto out;
	}

	/*
	 * Don't take interrupts on this CPU will bit banging
	 * the %#%#@$ I2C device
	 */
	local_irq_save(flags);

	eeprom_start(regs);

	eeprom_prep(regs, EEPROM_WRITE_SELECT);
	if (eeprom_check_ack(regs)) {
		local_irq_restore(flags);
		printk(KERN_ERR "%s: Unable to sync eeprom\n", ap->name);
		result = -EIO;
		goto eeprom_read_error;
	}

	eeprom_prep(regs, (offset >> 8) & 0xff);
	if (eeprom_check_ack(regs)) {
		local_irq_restore(flags);
		printk(KERN_ERR "%s: Unable to set address byte 0\n",
		       ap->name);
		result = -EIO;
		goto eeprom_read_error;
	}

	eeprom_prep(regs, offset & 0xff);
	if (eeprom_check_ack(regs)) {
		local_irq_restore(flags);
		printk(KERN_ERR "%s: Unable to set address byte 1\n",
		       ap->name);
		result = -EIO;
		goto eeprom_read_error;
	}

	eeprom_start(regs);
	eeprom_prep(regs, EEPROM_READ_SELECT);
	if (eeprom_check_ack(regs)) {
		local_irq_restore(flags);
		printk(KERN_ERR "%s: Unable to set READ_SELECT\n",
		       ap->name);
		result = -EIO;
		goto eeprom_read_error;
	}

	for (i = 0; i < 8; i++) {
		local = readl(&regs->LocalCtrl);
		local &= ~EEPROM_WRITE_ENABLE;
		writel(local, &regs->LocalCtrl);
		readl(&regs->LocalCtrl);
		udelay(ACE_LONG_DELAY);
		mb();
		local |= EEPROM_CLK_OUT;
		writel(local, &regs->LocalCtrl);
		readl(&regs->LocalCtrl);
		mb();
		udelay(ACE_SHORT_DELAY);
		/* sample data mid high clk */
		result = (result << 1) |
			((readl(&regs->LocalCtrl) & EEPROM_DATA_IN) != 0);
		udelay(ACE_SHORT_DELAY);
		mb();
		local = readl(&regs->LocalCtrl);
		local &= ~EEPROM_CLK_OUT;
		writel(local, &regs->LocalCtrl);
		readl(&regs->LocalCtrl);
		udelay(ACE_SHORT_DELAY);
		mb();
		if (i == 7) {
			local |= EEPROM_WRITE_ENABLE;
			writel(local, &regs->LocalCtrl);
			readl(&regs->LocalCtrl);
			mb();
			udelay(ACE_SHORT_DELAY);
		}
	}

	local |= EEPROM_DATA_OUT;
	writel(local, &regs->LocalCtrl);
	readl(&regs->LocalCtrl);
	mb();
	udelay(ACE_SHORT_DELAY);
	writel(readl(&regs->LocalCtrl) | EEPROM_CLK_OUT, &regs->LocalCtrl);
	readl(&regs->LocalCtrl);
	udelay(ACE_LONG_DELAY);
	writel(readl(&regs->LocalCtrl) & ~EEPROM_CLK_OUT, &regs->LocalCtrl);
	readl(&regs->LocalCtrl);
	mb();
	udelay(ACE_SHORT_DELAY);
	eeprom_stop(regs);

	local_irq_restore(flags);
 out:
	return result;

 eeprom_read_error:
	printk(KERN_ERR "%s: Unable to read eeprom byte 0x%02lx\n",
	       ap->name, offset);
	goto out;
}


/*
 * Local variables:
 * compile-command: "gcc -D__SMP__ -D__KERNEL__ -DMODULE -I../../include -Wall -Wstrict-prototypes -O2 -fomit-frame-pointer -pipe -fno-strength-reduce -DMODVERSIONS -include ../../include/linux/modversions.h   -c -o acenic.o acenic.c"
 * End:
 */