summaryrefslogtreecommitdiffstats
path: root/drivers/parisc/lba_pci.c
blob: 052fa230bc776b7d91b3377abcf947bf260e83f3 (plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
1001
1002
1003
1004
1005
1006
1007
1008
1009
1010
1011
1012
1013
1014
1015
1016
1017
1018
1019
1020
1021
1022
1023
1024
1025
1026
1027
1028
1029
1030
1031
1032
1033
1034
1035
1036
1037
1038
1039
1040
1041
1042
1043
1044
1045
1046
1047
1048
1049
1050
1051
1052
1053
1054
1055
1056
1057
1058
1059
1060
1061
1062
1063
1064
1065
1066
1067
1068
1069
1070
1071
1072
1073
1074
1075
1076
1077
1078
1079
1080
1081
1082
1083
1084
1085
1086
1087
1088
1089
1090
1091
1092
1093
1094
1095
1096
1097
1098
1099
1100
1101
1102
1103
1104
1105
1106
1107
1108
1109
1110
1111
1112
1113
1114
1115
1116
1117
1118
1119
1120
1121
1122
1123
1124
1125
1126
1127
1128
1129
1130
1131
1132
1133
1134
1135
1136
1137
1138
1139
1140
1141
1142
1143
1144
1145
1146
1147
1148
1149
1150
1151
1152
1153
1154
1155
1156
1157
1158
1159
1160
1161
1162
1163
1164
1165
1166
1167
1168
1169
1170
1171
1172
1173
1174
1175
1176
1177
1178
1179
1180
1181
1182
1183
1184
1185
1186
1187
1188
1189
1190
1191
1192
1193
1194
1195
1196
1197
1198
1199
1200
1201
1202
1203
1204
1205
1206
1207
1208
1209
1210
1211
1212
1213
1214
1215
1216
1217
1218
1219
1220
1221
1222
1223
1224
1225
1226
1227
1228
1229
1230
1231
1232
1233
1234
1235
1236
1237
1238
1239
1240
1241
1242
1243
1244
1245
1246
1247
1248
1249
1250
1251
1252
1253
1254
1255
1256
1257
1258
1259
1260
1261
1262
1263
1264
1265
1266
1267
1268
1269
1270
1271
1272
1273
1274
1275
1276
1277
1278
1279
1280
1281
1282
1283
1284
1285
1286
1287
1288
1289
1290
1291
1292
1293
1294
1295
1296
1297
1298
1299
1300
1301
1302
1303
1304
1305
1306
1307
1308
1309
1310
1311
1312
1313
1314
1315
1316
1317
1318
1319
1320
1321
1322
1323
1324
1325
1326
1327
1328
1329
1330
1331
1332
1333
1334
1335
1336
1337
1338
1339
1340
1341
1342
1343
1344
1345
1346
1347
1348
1349
1350
1351
1352
1353
1354
1355
1356
1357
1358
1359
1360
1361
1362
1363
1364
1365
1366
1367
1368
1369
1370
1371
1372
1373
1374
1375
1376
1377
1378
1379
1380
1381
1382
1383
1384
1385
1386
1387
1388
1389
1390
1391
1392
1393
1394
1395
1396
1397
1398
1399
1400
1401
1402
1403
1404
1405
1406
1407
1408
1409
1410
1411
1412
1413
1414
1415
1416
1417
1418
1419
1420
1421
1422
1423
1424
1425
1426
1427
1428
1429
1430
1431
1432
1433
1434
1435
1436
1437
1438
1439
1440
1441
1442
1443
1444
1445
1446
1447
1448
1449
1450
1451
1452
1453
1454
1455
1456
1457
1458
1459
1460
1461
1462
1463
1464
1465
1466
1467
1468
1469
1470
1471
1472
1473
1474
1475
1476
1477
1478
1479
1480
1481
1482
1483
1484
1485
1486
1487
1488
1489
1490
1491
1492
1493
1494
1495
1496
1497
1498
1499
1500
1501
1502
1503
1504
1505
1506
1507
1508
1509
1510
1511
1512
1513
1514
1515
1516
1517
1518
1519
1520
1521
1522
1523
1524
1525
1526
1527
1528
1529
1530
1531
1532
1533
1534
1535
1536
1537
1538
1539
1540
1541
1542
1543
1544
1545
1546
1547
1548
1549
1550
1551
1552
1553
1554
1555
1556
1557
1558
1559
1560
1561
1562
1563
1564
1565
1566
1567
1568
1569
1570
1571
1572
1573
1574
1575
1576
1577
1578
1579
1580
1581
1582
1583
1584
1585
1586
1587
1588
1589
1590
1591
1592
1593
/*
**
**  PCI Lower Bus Adapter (LBA) manager
**
**	(c) Copyright 1999,2000 Grant Grundler
**	(c) Copyright 1999,2000 Hewlett-Packard Company
**
**	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.
**
**
** This module primarily provides access to PCI bus (config/IOport
** spaces) on platforms with an SBA/LBA chipset. A/B/C/J/L/N-class
** with 4 digit model numbers - eg C3000 (and A400...sigh).
**
** LBA driver isn't as simple as the Dino driver because:
**   (a) this chip has substantial bug fixes between revisions
**       (Only one Dino bug has a software workaround :^(  )
**   (b) has more options which we don't (yet) support (DMA hints, OLARD)
**   (c) IRQ support lives in the I/O SAPIC driver (not with PCI driver)
**   (d) play nicely with both PAT and "Legacy" PA-RISC firmware (PDC).
**       (dino only deals with "Legacy" PDC)
**
** LBA driver passes the I/O SAPIC HPA to the I/O SAPIC driver.
** (I/O SAPIC is integratd in the LBA chip).
**
** FIXME: Add support to SBA and LBA drivers for DMA hint sets
** FIXME: Add support for PCI card hot-plug (OLARD).
*/

#include <linux/delay.h>
#include <linux/types.h>
#include <linux/kernel.h>
#include <linux/spinlock.h>
#include <linux/init.h>		/* for __init and __devinit */
#include <linux/pci.h>
#include <linux/ioport.h>
#include <linux/slab.h>

#include <asm/byteorder.h>
#include <asm/pdc.h>
#include <asm/pdcpat.h>
#include <asm/page.h>

#include <asm/ropes.h>
#include <asm/hardware.h>	/* for register_parisc_driver() stuff */
#include <asm/parisc-device.h>
#include <asm/io.h>		/* read/write stuff */

#undef DEBUG_LBA	/* general stuff */
#undef DEBUG_LBA_PORT	/* debug I/O Port access */
#undef DEBUG_LBA_CFG	/* debug Config Space Access (ie PCI Bus walk) */
#undef DEBUG_LBA_PAT	/* debug PCI Resource Mgt code - PDC PAT only */

#undef FBB_SUPPORT	/* Fast Back-Back xfers - NOT READY YET */


#ifdef DEBUG_LBA
#define DBG(x...)	printk(x)
#else
#define DBG(x...)
#endif

#ifdef DEBUG_LBA_PORT
#define DBG_PORT(x...)	printk(x)
#else
#define DBG_PORT(x...)
#endif

#ifdef DEBUG_LBA_CFG
#define DBG_CFG(x...)	printk(x)
#else
#define DBG_CFG(x...)
#endif

#ifdef DEBUG_LBA_PAT
#define DBG_PAT(x...)	printk(x)
#else
#define DBG_PAT(x...)
#endif


/*
** Config accessor functions only pass in the 8-bit bus number and not
** the 8-bit "PCI Segment" number. Each LBA will be assigned a PCI bus
** number based on what firmware wrote into the scratch register.
**
** The "secondary" bus number is set to this before calling
** pci_register_ops(). If any PPB's are present, the scan will
** discover them and update the "secondary" and "subordinate"
** fields in the pci_bus structure.
**
** Changes in the configuration *may* result in a different
** bus number for each LBA depending on what firmware does.
*/

#define MODULE_NAME "LBA"

/* non-postable I/O port space, densely packed */
#define LBA_PORT_BASE	(PCI_F_EXTEND | 0xfee00000UL)
static void __iomem *astro_iop_base __read_mostly;

static u32 lba_t32;

/* lba flags */
#define LBA_FLAG_SKIP_PROBE	0x10

#define LBA_SKIP_PROBE(d) ((d)->flags & LBA_FLAG_SKIP_PROBE)


/* Looks nice and keeps the compiler happy */
#define LBA_DEV(d) ((struct lba_device *) (d))


/*
** Only allow 8 subsidiary busses per LBA
** Problem is the PCI bus numbering is globally shared.
*/
#define LBA_MAX_NUM_BUSES 8

/************************************
 * LBA register read and write support
 *
 * BE WARNED: register writes are posted.
 *  (ie follow writes which must reach HW with a read)
 */
#define READ_U8(addr)  __raw_readb(addr)
#define READ_U16(addr) __raw_readw(addr)
#define READ_U32(addr) __raw_readl(addr)
#define WRITE_U8(value, addr)  __raw_writeb(value, addr)
#define WRITE_U16(value, addr) __raw_writew(value, addr)
#define WRITE_U32(value, addr) __raw_writel(value, addr)

#define READ_REG8(addr)  readb(addr)
#define READ_REG16(addr) readw(addr)
#define READ_REG32(addr) readl(addr)
#define READ_REG64(addr) readq(addr)
#define WRITE_REG8(value, addr)  writeb(value, addr)
#define WRITE_REG16(value, addr) writew(value, addr)
#define WRITE_REG32(value, addr) writel(value, addr)


#define LBA_CFG_TOK(bus,dfn) ((u32) ((bus)<<16 | (dfn)<<8))
#define LBA_CFG_BUS(tok)  ((u8) ((tok)>>16))
#define LBA_CFG_DEV(tok)  ((u8) ((tok)>>11) & 0x1f)
#define LBA_CFG_FUNC(tok) ((u8) ((tok)>>8 ) & 0x7)


/*
** Extract LBA (Rope) number from HPA
** REVISIT: 16 ropes for Stretch/Ike?
*/
#define ROPES_PER_IOC	8
#define LBA_NUM(x)    ((((unsigned long) x) >> 13) & (ROPES_PER_IOC-1))


static void
lba_dump_res(struct resource *r, int d)
{
	int i;

	if (NULL == r)
		return;

	printk(KERN_DEBUG "(%p)", r->parent);
	for (i = d; i ; --i) printk(" ");
	printk(KERN_DEBUG "%p [%lx,%lx]/%lx\n", r,
		(long)r->start, (long)r->end, r->flags);
	lba_dump_res(r->child, d+2);
	lba_dump_res(r->sibling, d);
}


/*
** LBA rev 2.0, 2.1, 2.2, and 3.0 bus walks require a complex
** workaround for cfg cycles:
**	-- preserve  LBA state
**	-- prevent any DMA from occurring
**	-- turn on smart mode
**	-- probe with config writes before doing config reads
**	-- check ERROR_STATUS
**	-- clear ERROR_STATUS
**	-- restore LBA state
**
** The workaround is only used for device discovery.
*/

static int lba_device_present(u8 bus, u8 dfn, struct lba_device *d)
{
	u8 first_bus = d->hba.hba_bus->secondary;
	u8 last_sub_bus = d->hba.hba_bus->subordinate;

	if ((bus < first_bus) ||
	    (bus > last_sub_bus) ||
	    ((bus - first_bus) >= LBA_MAX_NUM_BUSES)) {
		return 0;
	}

	return 1;
}



#define LBA_CFG_SETUP(d, tok) {				\
    /* Save contents of error config register.  */			\
    error_config = READ_REG32(d->hba.base_addr + LBA_ERROR_CONFIG);		\
\
    /* Save contents of status control register.  */			\
    status_control = READ_REG32(d->hba.base_addr + LBA_STAT_CTL);		\
\
    /* For LBA rev 2.0, 2.1, 2.2, and 3.0, we must disable DMA		\
    ** arbitration for full bus walks.					\
    */									\
	/* Save contents of arb mask register. */			\
	arb_mask = READ_REG32(d->hba.base_addr + LBA_ARB_MASK);		\
\
	/*								\
	 * Turn off all device arbitration bits (i.e. everything	\
	 * except arbitration enable bit).				\
	 */								\
	WRITE_REG32(0x1, d->hba.base_addr + LBA_ARB_MASK);		\
\
    /*									\
     * Set the smart mode bit so that master aborts don't cause		\
     * LBA to go into PCI fatal mode (required).			\
     */									\
    WRITE_REG32(error_config | LBA_SMART_MODE, d->hba.base_addr + LBA_ERROR_CONFIG);	\
}


#define LBA_CFG_PROBE(d, tok) {				\
    /*									\
     * Setup Vendor ID write and read back the address register		\
     * to make sure that LBA is the bus master.				\
     */									\
    WRITE_REG32(tok | PCI_VENDOR_ID, (d)->hba.base_addr + LBA_PCI_CFG_ADDR);\
    /*									\
     * Read address register to ensure that LBA is the bus master,	\
     * which implies that DMA traffic has stopped when DMA arb is off.	\
     */									\
    lba_t32 = READ_REG32((d)->hba.base_addr + LBA_PCI_CFG_ADDR);	\
    /*									\
     * Generate a cfg write cycle (will have no affect on		\
     * Vendor ID register since read-only).				\
     */									\
    WRITE_REG32(~0, (d)->hba.base_addr + LBA_PCI_CFG_DATA);		\
    /*									\
     * Make sure write has completed before proceeding further,		\
     * i.e. before setting clear enable.				\
     */									\
    lba_t32 = READ_REG32((d)->hba.base_addr + LBA_PCI_CFG_ADDR);	\
}


/*
 * HPREVISIT:
 *   -- Can't tell if config cycle got the error.
 *
 *		OV bit is broken until rev 4.0, so can't use OV bit and
 *		LBA_ERROR_LOG_ADDR to tell if error belongs to config cycle.
 *
 *		As of rev 4.0, no longer need the error check.
 *
 *   -- Even if we could tell, we still want to return -1
 *	for **ANY** error (not just master abort).
 *
 *   -- Only clear non-fatal errors (we don't want to bring
 *	LBA out of pci-fatal mode).
 *
 *		Actually, there is still a race in which
 *		we could be clearing a fatal error.  We will
 *		live with this during our initial bus walk
 *		until rev 4.0 (no driver activity during
 *		initial bus walk).  The initial bus walk
 *		has race conditions concerning the use of
 *		smart mode as well.
 */

#define LBA_MASTER_ABORT_ERROR 0xc
#define LBA_FATAL_ERROR 0x10

#define LBA_CFG_MASTER_ABORT_CHECK(d, base, tok, error) {		\
    u32 error_status = 0;						\
    /*									\
     * Set clear enable (CE) bit. Unset by HW when new			\
     * errors are logged -- LBA HW ERS section 14.3.3).		\
     */									\
    WRITE_REG32(status_control | CLEAR_ERRLOG_ENABLE, base + LBA_STAT_CTL); \
    error_status = READ_REG32(base + LBA_ERROR_STATUS);		\
    if ((error_status & 0x1f) != 0) {					\
	/*								\
	 * Fail the config read request.				\
	 */								\
	error = 1;							\
	if ((error_status & LBA_FATAL_ERROR) == 0) {			\
	    /*								\
	     * Clear error status (if fatal bit not set) by setting	\
	     * clear error log bit (CL).				\
	     */								\
	    WRITE_REG32(status_control | CLEAR_ERRLOG, base + LBA_STAT_CTL); \
	}								\
    }									\
}

#define LBA_CFG_TR4_ADDR_SETUP(d, addr)					\
	WRITE_REG32(((addr) & ~3), (d)->hba.base_addr + LBA_PCI_CFG_ADDR);

#define LBA_CFG_ADDR_SETUP(d, addr) {					\
    WRITE_REG32(((addr) & ~3), (d)->hba.base_addr + LBA_PCI_CFG_ADDR);	\
    /*									\
     * Read address register to ensure that LBA is the bus master,	\
     * which implies that DMA traffic has stopped when DMA arb is off.	\
     */									\
    lba_t32 = READ_REG32((d)->hba.base_addr + LBA_PCI_CFG_ADDR);	\
}


#define LBA_CFG_RESTORE(d, base) {					\
    /*									\
     * Restore status control register (turn off clear enable).		\
     */									\
    WRITE_REG32(status_control, base + LBA_STAT_CTL);			\
    /*									\
     * Restore error config register (turn off smart mode).		\
     */									\
    WRITE_REG32(error_config, base + LBA_ERROR_CONFIG);			\
	/*								\
	 * Restore arb mask register (reenables DMA arbitration).	\
	 */								\
	WRITE_REG32(arb_mask, base + LBA_ARB_MASK);			\
}



static unsigned int
lba_rd_cfg(struct lba_device *d, u32 tok, u8 reg, u32 size)
{
	u32 data = ~0U;
	int error = 0;
	u32 arb_mask = 0;	/* used by LBA_CFG_SETUP/RESTORE */
	u32 error_config = 0;	/* used by LBA_CFG_SETUP/RESTORE */
	u32 status_control = 0;	/* used by LBA_CFG_SETUP/RESTORE */

	LBA_CFG_SETUP(d, tok);
	LBA_CFG_PROBE(d, tok);
	LBA_CFG_MASTER_ABORT_CHECK(d, d->hba.base_addr, tok, error);
	if (!error) {
		void __iomem *data_reg = d->hba.base_addr + LBA_PCI_CFG_DATA;

		LBA_CFG_ADDR_SETUP(d, tok | reg);
		switch (size) {
		case 1: data = (u32) READ_REG8(data_reg + (reg & 3)); break;
		case 2: data = (u32) READ_REG16(data_reg+ (reg & 2)); break;
		case 4: data = READ_REG32(data_reg); break;
		}
	}
	LBA_CFG_RESTORE(d, d->hba.base_addr);
	return(data);
}


static int elroy_cfg_read(struct pci_bus *bus, unsigned int devfn, int pos, int size, u32 *data)
{
	struct lba_device *d = LBA_DEV(parisc_walk_tree(bus->bridge));
	u32 local_bus = (bus->parent == NULL) ? 0 : bus->secondary;
	u32 tok = LBA_CFG_TOK(local_bus, devfn);
	void __iomem *data_reg = d->hba.base_addr + LBA_PCI_CFG_DATA;

	if ((pos > 255) || (devfn > 255))
		return -EINVAL;

/* FIXME: B2K/C3600 workaround is always use old method... */
	/* if (!LBA_SKIP_PROBE(d)) */ {
		/* original - Generate config cycle on broken elroy
		  with risk we will miss PCI bus errors. */
		*data = lba_rd_cfg(d, tok, pos, size);
		DBG_CFG("%s(%x+%2x) -> 0x%x (a)\n", __func__, tok, pos, *data);
		return 0;
	}

	if (LBA_SKIP_PROBE(d) && !lba_device_present(bus->secondary, devfn, d)) {
		DBG_CFG("%s(%x+%2x) -> -1 (b)\n", __func__, tok, pos);
		/* either don't want to look or know device isn't present. */
		*data = ~0U;
		return(0);
	}

	/* Basic Algorithm
	** Should only get here on fully working LBA rev.
	** This is how simple the code should have been.
	*/
	LBA_CFG_ADDR_SETUP(d, tok | pos);
	switch(size) {
	case 1: *data = READ_REG8 (data_reg + (pos & 3)); break;
	case 2: *data = READ_REG16(data_reg + (pos & 2)); break;
	case 4: *data = READ_REG32(data_reg); break;
	}
	DBG_CFG("%s(%x+%2x) -> 0x%x (c)\n", __func__, tok, pos, *data);
	return 0;
}


static void
lba_wr_cfg(struct lba_device *d, u32 tok, u8 reg, u32 data, u32 size)
{
	int error = 0;
	u32 arb_mask = 0;
	u32 error_config = 0;
	u32 status_control = 0;
	void __iomem *data_reg = d->hba.base_addr + LBA_PCI_CFG_DATA;

	LBA_CFG_SETUP(d, tok);
	LBA_CFG_ADDR_SETUP(d, tok | reg);
	switch (size) {
	case 1: WRITE_REG8 (data, data_reg + (reg & 3)); break;
	case 2: WRITE_REG16(data, data_reg + (reg & 2)); break;
	case 4: WRITE_REG32(data, data_reg);             break;
	}
	LBA_CFG_MASTER_ABORT_CHECK(d, d->hba.base_addr, tok, error);
	LBA_CFG_RESTORE(d, d->hba.base_addr);
}


/*
 * LBA 4.0 config write code implements non-postable semantics
 * by doing a read of CONFIG ADDR after the write.
 */

static int elroy_cfg_write(struct pci_bus *bus, unsigned int devfn, int pos, int size, u32 data)
{
	struct lba_device *d = LBA_DEV(parisc_walk_tree(bus->bridge));
	u32 local_bus = (bus->parent == NULL) ? 0 : bus->secondary;
	u32 tok = LBA_CFG_TOK(local_bus,devfn);

	if ((pos > 255) || (devfn > 255))
		return -EINVAL;

	if (!LBA_SKIP_PROBE(d)) {
		/* Original Workaround */
		lba_wr_cfg(d, tok, pos, (u32) data, size);
		DBG_CFG("%s(%x+%2x) = 0x%x (a)\n", __func__, tok, pos,data);
		return 0;
	}

	if (LBA_SKIP_PROBE(d) && (!lba_device_present(bus->secondary, devfn, d))) {
		DBG_CFG("%s(%x+%2x) = 0x%x (b)\n", __func__, tok, pos,data);
		return 1; /* New Workaround */
	}

	DBG_CFG("%s(%x+%2x) = 0x%x (c)\n", __func__, tok, pos, data);

	/* Basic Algorithm */
	LBA_CFG_ADDR_SETUP(d, tok | pos);
	switch(size) {
	case 1: WRITE_REG8 (data, d->hba.base_addr + LBA_PCI_CFG_DATA + (pos & 3));
		   break;
	case 2: WRITE_REG16(data, d->hba.base_addr + LBA_PCI_CFG_DATA + (pos & 2));
		   break;
	case 4: WRITE_REG32(data, d->hba.base_addr + LBA_PCI_CFG_DATA);
		   break;
	}
	/* flush posted write */
	lba_t32 = READ_REG32(d->hba.base_addr + LBA_PCI_CFG_ADDR);
	return 0;
}


static struct pci_ops elroy_cfg_ops = {
	.read =		elroy_cfg_read,
	.write =	elroy_cfg_write,
};

/*
 * The mercury_cfg_ops are slightly misnamed; they're also used for Elroy
 * TR4.0 as no additional bugs were found in this areea between Elroy and
 * Mercury
 */

static int mercury_cfg_read(struct pci_bus *bus, unsigned int devfn, int pos, int size, u32 *data)
{
	struct lba_device *d = LBA_DEV(parisc_walk_tree(bus->bridge));
	u32 local_bus = (bus->parent == NULL) ? 0 : bus->secondary;
	u32 tok = LBA_CFG_TOK(local_bus, devfn);
	void __iomem *data_reg = d->hba.base_addr + LBA_PCI_CFG_DATA;

	if ((pos > 255) || (devfn > 255))
		return -EINVAL;

	LBA_CFG_TR4_ADDR_SETUP(d, tok | pos);
	switch(size) {
	case 1:
		*data = READ_REG8(data_reg + (pos & 3));
		break;
	case 2:
		*data = READ_REG16(data_reg + (pos & 2));
		break;
	case 4:
		*data = READ_REG32(data_reg);             break;
		break;
	}

	DBG_CFG("mercury_cfg_read(%x+%2x) -> 0x%x\n", tok, pos, *data);
	return 0;
}

/*
 * LBA 4.0 config write code implements non-postable semantics
 * by doing a read of CONFIG ADDR after the write.
 */

static int mercury_cfg_write(struct pci_bus *bus, unsigned int devfn, int pos, int size, u32 data)
{
	struct lba_device *d = LBA_DEV(parisc_walk_tree(bus->bridge));
	void __iomem *data_reg = d->hba.base_addr + LBA_PCI_CFG_DATA;
	u32 local_bus = (bus->parent == NULL) ? 0 : bus->secondary;
	u32 tok = LBA_CFG_TOK(local_bus,devfn);

	if ((pos > 255) || (devfn > 255))
		return -EINVAL;

	DBG_CFG("%s(%x+%2x) <- 0x%x (c)\n", __func__, tok, pos, data);

	LBA_CFG_TR4_ADDR_SETUP(d, tok | pos);
	switch(size) {
	case 1:
		WRITE_REG8 (data, data_reg + (pos & 3));
		break;
	case 2:
		WRITE_REG16(data, data_reg + (pos & 2));
		break;
	case 4:
		WRITE_REG32(data, data_reg);
		break;
	}

	/* flush posted write */
	lba_t32 = READ_U32(d->hba.base_addr + LBA_PCI_CFG_ADDR);
	return 0;
}

static struct pci_ops mercury_cfg_ops = {
	.read =		mercury_cfg_read,
	.write =	mercury_cfg_write,
};


static void
lba_bios_init(void)
{
	DBG(MODULE_NAME ": lba_bios_init\n");
}


#ifdef CONFIG_64BIT

/*
 * truncate_pat_collision:  Deal with overlaps or outright collisions
 *			between PAT PDC reported ranges.
 *
 *   Broken PA8800 firmware will report lmmio range that
 *   overlaps with CPU HPA. Just truncate the lmmio range.
 *
 *   BEWARE: conflicts with this lmmio range may be an
 *   elmmio range which is pointing down another rope.
 *
 *  FIXME: only deals with one collision per range...theoretically we
 *  could have several. Supporting more than one collision will get messy.
 */
static unsigned long
truncate_pat_collision(struct resource *root, struct resource *new)
{
	unsigned long start = new->start;
	unsigned long end = new->end;
	struct resource *tmp = root->child;

	if (end <= start || start < root->start || !tmp)
		return 0;

	/* find first overlap */
	while (tmp && tmp->end < start)
		tmp = tmp->sibling;

	/* no entries overlap */
	if (!tmp)  return 0;

	/* found one that starts behind the new one
	** Don't need to do anything.
	*/
	if (tmp->start >= end) return 0;

	if (tmp->start <= start) {
		/* "front" of new one overlaps */
		new->start = tmp->end + 1;

		if (tmp->end >= end) {
			/* AACCKK! totally overlaps! drop this range. */
			return 1;
		}
	} 

	if (tmp->end < end ) {
		/* "end" of new one overlaps */
		new->end = tmp->start - 1;
	}

	printk(KERN_WARNING "LBA: Truncating lmmio_space [%lx/%lx] "
					"to [%lx,%lx]\n",
			start, end,
			(long)new->start, (long)new->end );

	return 0;	/* truncation successful */
}

#else
#define truncate_pat_collision(r,n)  (0)
#endif

/*
** The algorithm is generic code.
** But it needs to access local data structures to get the IRQ base.
** Could make this a "pci_fixup_irq(bus, region)" but not sure
** it's worth it.
**
** Called by do_pci_scan_bus() immediately after each PCI bus is walked.
** Resources aren't allocated until recursive buswalk below HBA is completed.
*/
static void
lba_fixup_bus(struct pci_bus *bus)
{
	struct list_head *ln;
#ifdef FBB_SUPPORT
	u16 status;
#endif
	struct lba_device *ldev = LBA_DEV(parisc_walk_tree(bus->bridge));

	DBG("lba_fixup_bus(0x%p) bus %d platform_data 0x%p\n",
		bus, bus->secondary, bus->bridge->platform_data);

	/*
	** Properly Setup MMIO resources for this bus.
	** pci_alloc_primary_bus() mangles this.
	*/
	if (bus->parent) {
		int i;
		/* PCI-PCI Bridge */
		pci_read_bridge_bases(bus);
		for (i = PCI_BRIDGE_RESOURCES; i < PCI_NUM_RESOURCES; i++) {
			pci_claim_resource(bus->self, i);
		}
	} else {
		/* Host-PCI Bridge */
		int err;

		DBG("lba_fixup_bus() %s [%lx/%lx]/%lx\n",
			ldev->hba.io_space.name,
			ldev->hba.io_space.start, ldev->hba.io_space.end,
			ldev->hba.io_space.flags);
		DBG("lba_fixup_bus() %s [%lx/%lx]/%lx\n",
			ldev->hba.lmmio_space.name,
			ldev->hba.lmmio_space.start, ldev->hba.lmmio_space.end,
			ldev->hba.lmmio_space.flags);

		err = request_resource(&ioport_resource, &(ldev->hba.io_space));
		if (err < 0) {
			lba_dump_res(&ioport_resource, 2);
			BUG();
		}

		if (ldev->hba.elmmio_space.start) {
			err = request_resource(&iomem_resource,
					&(ldev->hba.elmmio_space));
			if (err < 0) {

				printk("FAILED: lba_fixup_bus() request for "
						"elmmio_space [%lx/%lx]\n",
						(long)ldev->hba.elmmio_space.start,
						(long)ldev->hba.elmmio_space.end);

				/* lba_dump_res(&iomem_resource, 2); */
				/* BUG(); */
			}
		}

		if (ldev->hba.lmmio_space.flags) {
			err = request_resource(&iomem_resource, &(ldev->hba.lmmio_space));
			if (err < 0) {
				printk(KERN_ERR "FAILED: lba_fixup_bus() request for "
					"lmmio_space [%lx/%lx]\n",
					(long)ldev->hba.lmmio_space.start,
					(long)ldev->hba.lmmio_space.end);
			}
		}

#ifdef CONFIG_64BIT
		/* GMMIO is  distributed range. Every LBA/Rope gets part it. */
		if (ldev->hba.gmmio_space.flags) {
			err = request_resource(&iomem_resource, &(ldev->hba.gmmio_space));
			if (err < 0) {
				printk("FAILED: lba_fixup_bus() request for "
					"gmmio_space [%lx/%lx]\n",
					(long)ldev->hba.gmmio_space.start,
					(long)ldev->hba.gmmio_space.end);
				lba_dump_res(&iomem_resource, 2);
				BUG();
			}
		}
#endif

	}

	list_for_each(ln, &bus->devices) {
		int i;
		struct pci_dev *dev = pci_dev_b(ln);

		DBG("lba_fixup_bus() %s\n", pci_name(dev));

		/* Virtualize Device/Bridge Resources. */
		for (i = 0; i < PCI_BRIDGE_RESOURCES; i++) {
			struct resource *res = &dev->resource[i];

			/* If resource not allocated - skip it */
			if (!res->start)
				continue;

			/*
			** FIXME: this will result in whinging for devices
			** that share expansion ROMs (think quad tulip), but
			** isn't harmful.
			*/
			pci_claim_resource(dev, i);
		}

#ifdef FBB_SUPPORT
		/*
		** If one device does not support FBB transfers,
		** No one on the bus can be allowed to use them.
		*/
		(void) pci_read_config_word(dev, PCI_STATUS, &status);
		bus->bridge_ctl &= ~(status & PCI_STATUS_FAST_BACK);
#endif

                /*
		** P2PB's have no IRQs. ignore them.
		*/
		if ((dev->class >> 8) == PCI_CLASS_BRIDGE_PCI)
			continue;

		/* Adjust INTERRUPT_LINE for this dev */
		iosapic_fixup_irq(ldev->iosapic_obj, dev);
	}

#ifdef FBB_SUPPORT
/* FIXME/REVISIT - finish figuring out to set FBB on both
** pci_setup_bridge() clobbers PCI_BRIDGE_CONTROL.
** Can't fixup here anyway....garr...
*/
	if (fbb_enable) {
		if (bus->parent) {
			u8 control;
			/* enable on PPB */
			(void) pci_read_config_byte(bus->self, PCI_BRIDGE_CONTROL, &control);
			(void) pci_write_config_byte(bus->self, PCI_BRIDGE_CONTROL, control | PCI_STATUS_FAST_BACK);

		} else {
			/* enable on LBA */
		}
		fbb_enable = PCI_COMMAND_FAST_BACK;
	}

	/* Lastly enable FBB/PERR/SERR on all devices too */
	list_for_each(ln, &bus->devices) {
		(void) pci_read_config_word(dev, PCI_COMMAND, &status);
		status |= PCI_COMMAND_PARITY | PCI_COMMAND_SERR | fbb_enable;
		(void) pci_write_config_word(dev, PCI_COMMAND, status);
	}
#endif
}


static struct pci_bios_ops lba_bios_ops = {
	.init =		lba_bios_init,
	.fixup_bus =	lba_fixup_bus,
};




/*******************************************************
**
** LBA Sprockets "I/O Port" Space Accessor Functions
**
** This set of accessor functions is intended for use with
** "legacy firmware" (ie Sprockets on Allegro/Forte boxes).
**
** Many PCI devices don't require use of I/O port space (eg Tulip,
** NCR720) since they export the same registers to both MMIO and
** I/O port space. In general I/O port space is slower than
** MMIO since drivers are designed so PIO writes can be posted.
**
********************************************************/

#define LBA_PORT_IN(size, mask) \
static u##size lba_astro_in##size (struct pci_hba_data *d, u16 addr) \
{ \
	u##size t; \
	t = READ_REG##size(astro_iop_base + addr); \
	DBG_PORT(" 0x%x\n", t); \
	return (t); \
}

LBA_PORT_IN( 8, 3)
LBA_PORT_IN(16, 2)
LBA_PORT_IN(32, 0)



/*
** BUG X4107:  Ordering broken - DMA RD return can bypass PIO WR
**
** Fixed in Elroy 2.2. The READ_U32(..., LBA_FUNC_ID) below is
** guarantee non-postable completion semantics - not avoid X4107.
** The READ_U32 only guarantees the write data gets to elroy but
** out to the PCI bus. We can't read stuff from I/O port space
** since we don't know what has side-effects. Attempting to read
** from configuration space would be suicidal given the number of
** bugs in that elroy functionality.
**
**      Description:
**          DMA read results can improperly pass PIO writes (X4107).  The
**          result of this bug is that if a processor modifies a location in
**          memory after having issued PIO writes, the PIO writes are not
**          guaranteed to be completed before a PCI device is allowed to see
**          the modified data in a DMA read.
**
**          Note that IKE bug X3719 in TR1 IKEs will result in the same
**          symptom.
**
**      Workaround:
**          The workaround for this bug is to always follow a PIO write with
**          a PIO read to the same bus before starting DMA on that PCI bus.
**
*/
#define LBA_PORT_OUT(size, mask) \
static void lba_astro_out##size (struct pci_hba_data *d, u16 addr, u##size val) \
{ \
	DBG_PORT("%s(0x%p, 0x%x, 0x%x)\n", __func__, d, addr, val); \
	WRITE_REG##size(val, astro_iop_base + addr); \
	if (LBA_DEV(d)->hw_rev < 3) \
		lba_t32 = READ_U32(d->base_addr + LBA_FUNC_ID); \
}

LBA_PORT_OUT( 8, 3)
LBA_PORT_OUT(16, 2)
LBA_PORT_OUT(32, 0)


static struct pci_port_ops lba_astro_port_ops = {
	.inb =	lba_astro_in8,
	.inw =	lba_astro_in16,
	.inl =	lba_astro_in32,
	.outb =	lba_astro_out8,
	.outw =	lba_astro_out16,
	.outl =	lba_astro_out32
};


#ifdef CONFIG_64BIT
#define PIOP_TO_GMMIO(lba, addr) \
	((lba)->iop_base + (((addr)&0xFFFC)<<10) + ((addr)&3))

/*******************************************************
**
** LBA PAT "I/O Port" Space Accessor Functions
**
** This set of accessor functions is intended for use with
** "PAT PDC" firmware (ie Prelude/Rhapsody/Piranha boxes).
**
** This uses the PIOP space located in the first 64MB of GMMIO.
** Each rope gets a full 64*KB* (ie 4 bytes per page) this way.
** bits 1:0 stay the same.  bits 15:2 become 25:12.
** Then add the base and we can generate an I/O Port cycle.
********************************************************/
#undef LBA_PORT_IN
#define LBA_PORT_IN(size, mask) \
static u##size lba_pat_in##size (struct pci_hba_data *l, u16 addr) \
{ \
	u##size t; \
	DBG_PORT("%s(0x%p, 0x%x) ->", __func__, l, addr); \
	t = READ_REG##size(PIOP_TO_GMMIO(LBA_DEV(l), addr)); \
	DBG_PORT(" 0x%x\n", t); \
	return (t); \
}

LBA_PORT_IN( 8, 3)
LBA_PORT_IN(16, 2)
LBA_PORT_IN(32, 0)


#undef LBA_PORT_OUT
#define LBA_PORT_OUT(size, mask) \
static void lba_pat_out##size (struct pci_hba_data *l, u16 addr, u##size val) \
{ \
	void __iomem *where = PIOP_TO_GMMIO(LBA_DEV(l), addr); \
	DBG_PORT("%s(0x%p, 0x%x, 0x%x)\n", __func__, l, addr, val); \
	WRITE_REG##size(val, where); \
	/* flush the I/O down to the elroy at least */ \
	lba_t32 = READ_U32(l->base_addr + LBA_FUNC_ID); \
}

LBA_PORT_OUT( 8, 3)
LBA_PORT_OUT(16, 2)
LBA_PORT_OUT(32, 0)


static struct pci_port_ops lba_pat_port_ops = {
	.inb =	lba_pat_in8,
	.inw =	lba_pat_in16,
	.inl =	lba_pat_in32,
	.outb =	lba_pat_out8,
	.outw =	lba_pat_out16,
	.outl =	lba_pat_out32
};



/*
** make range information from PDC available to PCI subsystem.
** We make the PDC call here in order to get the PCI bus range
** numbers. The rest will get forwarded in pcibios_fixup_bus().
** We don't have a struct pci_bus assigned to us yet.
*/
static void
lba_pat_resources(struct parisc_device *pa_dev, struct lba_device *lba_dev)
{
	unsigned long bytecnt;
	long io_count;
	long status;	/* PDC return status */
	long pa_count;
	pdc_pat_cell_mod_maddr_block_t *pa_pdc_cell;	/* PA_VIEW */
	pdc_pat_cell_mod_maddr_block_t *io_pdc_cell;	/* IO_VIEW */
	int i;

	pa_pdc_cell = kzalloc(sizeof(pdc_pat_cell_mod_maddr_block_t), GFP_KERNEL);
	if (!pa_pdc_cell)
		return;

	io_pdc_cell = kzalloc(sizeof(pdc_pat_cell_mod_maddr_block_t), GFP_KERNEL);
	if (!io_pdc_cell) {
		kfree(pa_pdc_cell);
		return;
	}

	/* return cell module (IO view) */
	status = pdc_pat_cell_module(&bytecnt, pa_dev->pcell_loc, pa_dev->mod_index,
				PA_VIEW, pa_pdc_cell);
	pa_count = pa_pdc_cell->mod[1];

	status |= pdc_pat_cell_module(&bytecnt, pa_dev->pcell_loc, pa_dev->mod_index,
				IO_VIEW, io_pdc_cell);
	io_count = io_pdc_cell->mod[1];

	/* We've already done this once for device discovery...*/
	if (status != PDC_OK) {
		panic("pdc_pat_cell_module() call failed for LBA!\n");
	}

	if (PAT_GET_ENTITY(pa_pdc_cell->mod_info) != PAT_ENTITY_LBA) {
		panic("pdc_pat_cell_module() entity returned != PAT_ENTITY_LBA!\n");
	}

	/*
	** Inspect the resources PAT tells us about
	*/
	for (i = 0; i < pa_count; i++) {
		struct {
			unsigned long type;
			unsigned long start;
			unsigned long end;	/* aka finish */
		} *p, *io;
		struct resource *r;

		p = (void *) &(pa_pdc_cell->mod[2+i*3]);
		io = (void *) &(io_pdc_cell->mod[2+i*3]);

		/* Convert the PAT range data to PCI "struct resource" */
		switch(p->type & 0xff) {
		case PAT_PBNUM:
			lba_dev->hba.bus_num.start = p->start;
			lba_dev->hba.bus_num.end   = p->end;
			break;

		case PAT_LMMIO:
			/* used to fix up pre-initialized MEM BARs */
			if (!lba_dev->hba.lmmio_space.start) {
				sprintf(lba_dev->hba.lmmio_name,
						"PCI%02x LMMIO",
						(int)lba_dev->hba.bus_num.start);
				lba_dev->hba.lmmio_space_offset = p->start -
					io->start;
				r = &lba_dev->hba.lmmio_space;
				r->name = lba_dev->hba.lmmio_name;
			} else if (!lba_dev->hba.elmmio_space.start) {
				sprintf(lba_dev->hba.elmmio_name,
						"PCI%02x ELMMIO",
						(int)lba_dev->hba.bus_num.start);
				r = &lba_dev->hba.elmmio_space;
				r->name = lba_dev->hba.elmmio_name;
			} else {
				printk(KERN_WARNING MODULE_NAME
					" only supports 2 LMMIO resources!\n");
				break;
			}

			r->start  = p->start;
			r->end    = p->end;
			r->flags  = IORESOURCE_MEM;
			r->parent = r->sibling = r->child = NULL;
			break;

		case PAT_GMMIO:
			/* MMIO space > 4GB phys addr; for 64-bit BAR */
			sprintf(lba_dev->hba.gmmio_name, "PCI%02x GMMIO",
					(int)lba_dev->hba.bus_num.start);
			r = &lba_dev->hba.gmmio_space;
			r->name  = lba_dev->hba.gmmio_name;
			r->start  = p->start;
			r->end    = p->end;
			r->flags  = IORESOURCE_MEM;
			r->parent = r->sibling = r->child = NULL;
			break;

		case PAT_NPIOP:
			printk(KERN_WARNING MODULE_NAME
				" range[%d] : ignoring NPIOP (0x%lx)\n",
				i, p->start);
			break;

		case PAT_PIOP:
			/*
			** Postable I/O port space is per PCI host adapter.
			** base of 64MB PIOP region
			*/
			lba_dev->iop_base = ioremap_nocache(p->start, 64 * 1024 * 1024);

			sprintf(lba_dev->hba.io_name, "PCI%02x Ports",
					(int)lba_dev->hba.bus_num.start);
			r = &lba_dev->hba.io_space;
			r->name  = lba_dev->hba.io_name;
			r->start  = HBA_PORT_BASE(lba_dev->hba.hba_num);
			r->end    = r->start + HBA_PORT_SPACE_SIZE - 1;
			r->flags  = IORESOURCE_IO;
			r->parent = r->sibling = r->child = NULL;
			break;

		default:
			printk(KERN_WARNING MODULE_NAME
				" range[%d] : unknown pat range type (0x%lx)\n",
				i, p->type & 0xff);
			break;
		}
	}

	kfree(pa_pdc_cell);
	kfree(io_pdc_cell);
}
#else
/* keep compiler from complaining about missing declarations */
#define lba_pat_port_ops lba_astro_port_ops
#define lba_pat_resources(pa_dev, lba_dev)
#endif	/* CONFIG_64BIT */


extern void sba_distributed_lmmio(struct parisc_device *, struct resource *);
extern void sba_directed_lmmio(struct parisc_device *, struct resource *);


static void
lba_legacy_resources(struct parisc_device *pa_dev, struct lba_device *lba_dev)
{
	struct resource *r;
	int lba_num;

	lba_dev->hba.lmmio_space_offset = PCI_F_EXTEND;

	/*
	** With "legacy" firmware, the lowest byte of FW_SCRATCH
	** represents bus->secondary and the second byte represents
	** bus->subsidiary (i.e. highest PPB programmed by firmware).
	** PCI bus walk *should* end up with the same result.
	** FIXME: But we don't have sanity checks in PCI or LBA.
	*/
	lba_num = READ_REG32(lba_dev->hba.base_addr + LBA_FW_SCRATCH);
	r = &(lba_dev->hba.bus_num);
	r->name = "LBA PCI Busses";
	r->start = lba_num & 0xff;
	r->end = (lba_num>>8) & 0xff;

	/* Set up local PCI Bus resources - we don't need them for
	** Legacy boxes but it's nice to see in /proc/iomem.
	*/
	r = &(lba_dev->hba.lmmio_space);
	sprintf(lba_dev->hba.lmmio_name, "PCI%02x LMMIO",
					(int)lba_dev->hba.bus_num.start);
	r->name  = lba_dev->hba.lmmio_name;

#if 1
	/* We want the CPU -> IO routing of addresses.
	 * The SBA BASE/MASK registers control CPU -> IO routing.
	 * Ask SBA what is routed to this rope/LBA.
	 */
	sba_distributed_lmmio(pa_dev, r);
#else
	/*
	 * The LBA BASE/MASK registers control IO -> System routing.
	 *
	 * The following code works but doesn't get us what we want.
	 * Well, only because firmware (v5.0) on C3000 doesn't program
	 * the LBA BASE/MASE registers to be the exact inverse of 
	 * the corresponding SBA registers. Other Astro/Pluto
	 * based platform firmware may do it right.
	 *
	 * Should someone want to mess with MSI, they may need to
	 * reprogram LBA BASE/MASK registers. Thus preserve the code
	 * below until MSI is known to work on C3000/A500/N4000/RP3440.
	 *
	 * Using the code below, /proc/iomem shows:
	 * ...
	 * f0000000-f0ffffff : PCI00 LMMIO
	 *   f05d0000-f05d0000 : lcd_data
	 *   f05d0008-f05d0008 : lcd_cmd
	 * f1000000-f1ffffff : PCI01 LMMIO
	 * f4000000-f4ffffff : PCI02 LMMIO
	 *   f4000000-f4001fff : sym53c8xx
	 *   f4002000-f4003fff : sym53c8xx
	 *   f4004000-f40043ff : sym53c8xx
	 *   f4005000-f40053ff : sym53c8xx
	 *   f4007000-f4007fff : ohci_hcd
	 *   f4008000-f40083ff : tulip
	 * f6000000-f6ffffff : PCI03 LMMIO
	 * f8000000-fbffffff : PCI00 ELMMIO
	 *   fa100000-fa4fffff : stifb mmio
	 *   fb000000-fb1fffff : stifb fb
	 *
	 * But everything listed under PCI02 actually lives under PCI00.
	 * This is clearly wrong.
	 *
	 * Asking SBA how things are routed tells the correct story:
	 * LMMIO_BASE/MASK/ROUTE f4000001 fc000000 00000000
	 * DIR0_BASE/MASK/ROUTE fa000001 fe000000 00000006
	 * DIR1_BASE/MASK/ROUTE f9000001 ff000000 00000004
	 * DIR2_BASE/MASK/ROUTE f0000000 fc000000 00000000
	 * DIR3_BASE/MASK/ROUTE f0000000 fc000000 00000000
	 *
	 * Which looks like this in /proc/iomem:
	 * f4000000-f47fffff : PCI00 LMMIO
	 *   f4000000-f4001fff : sym53c8xx
	 *   ...[deteled core devices - same as above]...
	 *   f4008000-f40083ff : tulip
	 * f4800000-f4ffffff : PCI01 LMMIO
	 * f6000000-f67fffff : PCI02 LMMIO
	 * f7000000-f77fffff : PCI03 LMMIO
	 * f9000000-f9ffffff : PCI02 ELMMIO
	 * fa000000-fbffffff : PCI03 ELMMIO
	 *   fa100000-fa4fffff : stifb mmio
	 *   fb000000-fb1fffff : stifb fb
	 *
	 * ie all Built-in core are under now correctly under PCI00.
	 * The "PCI02 ELMMIO" directed range is for:
	 *  +-[02]---03.0  3Dfx Interactive, Inc. Voodoo 2
	 *
	 * All is well now.
	 */
	r->start = READ_REG32(lba_dev->hba.base_addr + LBA_LMMIO_BASE);
	if (r->start & 1) {
		unsigned long rsize;

		r->flags = IORESOURCE_MEM;
		/* mmio_mask also clears Enable bit */
		r->start &= mmio_mask;
		r->start = PCI_HOST_ADDR(HBA_DATA(lba_dev), r->start);
		rsize = ~ READ_REG32(lba_dev->hba.base_addr + LBA_LMMIO_MASK);

		/*
		** Each rope only gets part of the distributed range.
		** Adjust "window" for this rope.
		*/
		rsize /= ROPES_PER_IOC;
		r->start += (rsize + 1) * LBA_NUM(pa_dev->hpa.start);
		r->end = r->start + rsize;
	} else {
		r->end = r->start = 0;	/* Not enabled. */
	}
#endif

	/*
	** "Directed" ranges are used when the "distributed range" isn't
	** sufficient for all devices below a given LBA.  Typically devices
	** like graphics cards or X25 may need a directed range when the
	** bus has multiple slots (ie multiple devices) or the device
	** needs more than the typical 4 or 8MB a distributed range offers.
	**
	** The main reason for ignoring it now frigging complications.
	** Directed ranges may overlap (and have precedence) over
	** distributed ranges. Or a distributed range assigned to a unused
	** rope may be used by a directed range on a different rope.
	** Support for graphics devices may require fixing this
	** since they may be assigned a directed range which overlaps
	** an existing (but unused portion of) distributed range.
	*/
	r = &(lba_dev->hba.elmmio_space);
	sprintf(lba_dev->hba.elmmio_name, "PCI%02x ELMMIO",
					(int)lba_dev->hba.bus_num.start);
	r->name  = lba_dev->hba.elmmio_name;

#if 1
	/* See comment which precedes call to sba_directed_lmmio() */
	sba_directed_lmmio(pa_dev, r);
#else
	r->start = READ_REG32(lba_dev->hba.base_addr + LBA_ELMMIO_BASE);

	if (r->start & 1) {
		unsigned long rsize;
		r->flags = IORESOURCE_MEM;
		/* mmio_mask also clears Enable bit */
		r->start &= mmio_mask;
		r->start = PCI_HOST_ADDR(HBA_DATA(lba_dev), r->start);
		rsize = READ_REG32(lba_dev->hba.base_addr + LBA_ELMMIO_MASK);
		r->end = r->start + ~rsize;
	}
#endif

	r = &(lba_dev->hba.io_space);
	sprintf(lba_dev->hba.io_name, "PCI%02x Ports",
					(int)lba_dev->hba.bus_num.start);
	r->name  = lba_dev->hba.io_name;
	r->flags = IORESOURCE_IO;
	r->start = READ_REG32(lba_dev->hba.base_addr + LBA_IOS_BASE) & ~1L;
	r->end   = r->start + (READ_REG32(lba_dev->hba.base_addr + LBA_IOS_MASK) ^ (HBA_PORT_SPACE_SIZE - 1));

	/* Virtualize the I/O Port space ranges */
	lba_num = HBA_PORT_BASE(lba_dev->hba.hba_num);
	r->start |= lba_num;
	r->end   |= lba_num;
}


/**************************************************************************
**
**   LBA initialization code (HW and SW)
**
**   o identify LBA chip itself
**   o initialize LBA chip modes (HardFail)
**   o FIXME: initialize DMA hints for reasonable defaults
**   o enable configuration functions
**   o call pci_register_ops() to discover devs (fixup/fixup_bus get invoked)
**
**************************************************************************/

static int __init
lba_hw_init(struct lba_device *d)
{
	u32 stat;
	u32 bus_reset;	/* PDC_PAT_BUG */

#if 0
	printk(KERN_DEBUG "LBA %lx  STAT_CTL %Lx  ERROR_CFG %Lx  STATUS %Lx DMA_CTL %Lx\n",
		d->hba.base_addr,
		READ_REG64(d->hba.base_addr + LBA_STAT_CTL),
		READ_REG64(d->hba.base_addr + LBA_ERROR_CONFIG),
		READ_REG64(d->hba.base_addr + LBA_ERROR_STATUS),
		READ_REG64(d->hba.base_addr + LBA_DMA_CTL) );
	printk(KERN_DEBUG "	ARB mask %Lx  pri %Lx  mode %Lx  mtlt %Lx\n",
		READ_REG64(d->hba.base_addr + LBA_ARB_MASK),
		READ_REG64(d->hba.base_addr + LBA_ARB_PRI),
		READ_REG64(d->hba.base_addr + LBA_ARB_MODE),
		READ_REG64(d->hba.base_addr + LBA_ARB_MTLT) );
	printk(KERN_DEBUG "	HINT cfg 0x%Lx\n",
		READ_REG64(d->hba.base_addr + LBA_HINT_CFG));
	printk(KERN_DEBUG "	HINT reg ");
	{ int i;
	for (i=LBA_HINT_BASE; i< (14*8 + LBA_HINT_BASE); i+=8)
		printk(" %Lx", READ_REG64(d->hba.base_addr + i));
	}
	printk("\n");
#endif	/* DEBUG_LBA_PAT */

#ifdef CONFIG_64BIT
/*
 * FIXME add support for PDC_PAT_IO "Get slot status" - OLAR support
 * Only N-Class and up can really make use of Get slot status.
 * maybe L-class too but I've never played with it there.
 */
#endif

	/* PDC_PAT_BUG: exhibited in rev 40.48  on L2000 */
	bus_reset = READ_REG32(d->hba.base_addr + LBA_STAT_CTL + 4) & 1;
	if (bus_reset) {
		printk(KERN_DEBUG "NOTICE: PCI bus reset still asserted! (clearing)\n");
	}

	stat = READ_REG32(d->hba.base_addr + LBA_ERROR_CONFIG);
	if (stat & LBA_SMART_MODE) {
		printk(KERN_DEBUG "NOTICE: LBA in SMART mode! (cleared)\n");
		stat &= ~LBA_SMART_MODE;
		WRITE_REG32(stat, d->hba.base_addr + LBA_ERROR_CONFIG);
	}

	/* Set HF mode as the default (vs. -1 mode). */
        stat = READ_REG32(d->hba.base_addr + LBA_STAT_CTL);
	WRITE_REG32(stat | HF_ENABLE, d->hba.base_addr + LBA_STAT_CTL);

	/*
	** Writing a zero to STAT_CTL.rf (bit 0) will clear reset signal
	** if it's not already set. If we just cleared the PCI Bus Reset
	** signal, wait a bit for the PCI devices to recover and setup.
	*/
	if (bus_reset)
		mdelay(pci_post_reset_delay);

	if (0 == READ_REG32(d->hba.base_addr + LBA_ARB_MASK)) {
		/*
		** PDC_PAT_BUG: PDC rev 40.48 on L2000.
		** B2000/C3600/J6000 also have this problem?
		** 
		** Elroys with hot pluggable slots don't get configured
		** correctly if the slot is empty.  ARB_MASK is set to 0
		** and we can't master transactions on the bus if it's
		** not at least one. 0x3 enables elroy and first slot.
		*/
		printk(KERN_DEBUG "NOTICE: Enabling PCI Arbitration\n");
		WRITE_REG32(0x3, d->hba.base_addr + LBA_ARB_MASK);
	}

	/*
	** FIXME: Hint registers are programmed with default hint
	** values by firmware. Hints should be sane even if we
	** can't reprogram them the way drivers want.
	*/
	return 0;
}

/*
 * Unfortunately, when firmware numbers busses, it doesn't take into account
 * Cardbus bridges.  So we have to renumber the busses to suit ourselves.
 * Elroy/Mercury don't actually know what bus number they're attached to;
 * we use bus 0 to indicate the directly attached bus and any other bus
 * number will be taken care of by the PCI-PCI bridge.
 */
static unsigned int lba_next_bus = 0;

/*
 * Determine if lba should claim this chip (return 0) or not (return 1).
 * If so, initialize the chip and tell other partners in crime they
 * have work to do.
 */
static int __init
lba_driver_probe(struct parisc_device *dev)
{
	struct lba_device *lba_dev;
	LIST_HEAD(resources);
	struct pci_bus *lba_bus;
	struct pci_ops *cfg_ops;
	u32 func_class;
	void *tmp_obj;
	char *version;
	void __iomem *addr = ioremap_nocache(dev->hpa.start, 4096);

	/* Read HW Rev First */
	func_class = READ_REG32(addr + LBA_FCLASS);

	if (IS_ELROY(dev)) {	
		func_class &= 0xf;
		switch (func_class) {
		case 0:	version = "TR1.0"; break;
		case 1:	version = "TR2.0"; break;
		case 2:	version = "TR2.1"; break;
		case 3:	version = "TR2.2"; break;
		case 4:	version = "TR3.0"; break;
		case 5:	version = "TR4.0"; break;
		default: version = "TR4+";
		}

		printk(KERN_INFO "Elroy version %s (0x%x) found at 0x%lx\n",
		       version, func_class & 0xf, (long)dev->hpa.start);

		if (func_class < 2) {
			printk(KERN_WARNING "Can't support LBA older than "
				"TR2.1 - continuing under adversity.\n");
		}

#if 0
/* Elroy TR4.0 should work with simple algorithm.
   But it doesn't.  Still missing something. *sigh*
*/
		if (func_class > 4) {
			cfg_ops = &mercury_cfg_ops;
		} else
#endif
		{
			cfg_ops = &elroy_cfg_ops;
		}

	} else if (IS_MERCURY(dev) || IS_QUICKSILVER(dev)) {
		int major, minor;

		func_class &= 0xff;
		major = func_class >> 4, minor = func_class & 0xf;

		/* We could use one printk for both Elroy and Mercury,
                 * but for the mask for func_class.
                 */ 
		printk(KERN_INFO "%s version TR%d.%d (0x%x) found at 0x%lx\n",
		       IS_MERCURY(dev) ? "Mercury" : "Quicksilver", major,
		       minor, func_class, (long)dev->hpa.start);

		cfg_ops = &mercury_cfg_ops;
	} else {
		printk(KERN_ERR "Unknown LBA found at 0x%lx\n",
			(long)dev->hpa.start);
		return -ENODEV;
	}

	/* Tell I/O SAPIC driver we have a IRQ handler/region. */
	tmp_obj = iosapic_register(dev->hpa.start + LBA_IOSAPIC_BASE);

	/* NOTE: PCI devices (e.g. 103c:1005 graphics card) which don't
	**	have an IRT entry will get NULL back from iosapic code.
	*/
	
	lba_dev = kzalloc(sizeof(struct lba_device), GFP_KERNEL);
	if (!lba_dev) {
		printk(KERN_ERR "lba_init_chip - couldn't alloc lba_device\n");
		return(1);
	}


	/* ---------- First : initialize data we already have --------- */

	lba_dev->hw_rev = func_class;
	lba_dev->hba.base_addr = addr;
	lba_dev->hba.dev = dev;
	lba_dev->iosapic_obj = tmp_obj;  /* save interrupt handle */
	lba_dev->hba.iommu = sba_get_iommu(dev);  /* get iommu data */
	parisc_set_drvdata(dev, lba_dev);

	/* ------------ Second : initialize common stuff ---------- */
	pci_bios = &lba_bios_ops;
	pcibios_register_hba(HBA_DATA(lba_dev));
	spin_lock_init(&lba_dev->lba_lock);

	if (lba_hw_init(lba_dev))
		return(1);

	/* ---------- Third : setup I/O Port and MMIO resources  --------- */

	if (is_pdc_pat()) {
		/* PDC PAT firmware uses PIOP region of GMMIO space. */
		pci_port = &lba_pat_port_ops;
		/* Go ask PDC PAT what resources this LBA has */
		lba_pat_resources(dev, lba_dev);
	} else {
		if (!astro_iop_base) {
			/* Sprockets PDC uses NPIOP region */
			astro_iop_base = ioremap_nocache(LBA_PORT_BASE, 64 * 1024);
			pci_port = &lba_astro_port_ops;
		}

		/* Poke the chip a bit for /proc output */
		lba_legacy_resources(dev, lba_dev);
	}

	if (lba_dev->hba.bus_num.start < lba_next_bus)
		lba_dev->hba.bus_num.start = lba_next_bus;

	/*   Overlaps with elmmio can (and should) fail here.
	 *   We will prune (or ignore) the distributed range.
	 *
	 *   FIXME: SBA code should register all elmmio ranges first.
	 *      that would take care of elmmio ranges routed
	 *	to a different rope (already discovered) from
	 *	getting registered *after* LBA code has already
	 *	registered it's distributed lmmio range.
	 */
	if (truncate_pat_collision(&iomem_resource,
				   &(lba_dev->hba.lmmio_space))) {
		printk(KERN_WARNING "LBA: lmmio_space [%lx/%lx] duplicate!\n",
				(long)lba_dev->hba.lmmio_space.start,
				(long)lba_dev->hba.lmmio_space.end);
		lba_dev->hba.lmmio_space.flags = 0;
	}

	pci_add_resource_offset(&resources, &lba_dev->hba.io_space,
				HBA_PORT_BASE(lba_dev->hba.hba_num));
	if (lba_dev->hba.elmmio_space.start)
		pci_add_resource_offset(&resources, &lba_dev->hba.elmmio_space,
					lba_dev->hba.lmmio_space_offset);
	if (lba_dev->hba.lmmio_space.flags)
		pci_add_resource_offset(&resources, &lba_dev->hba.lmmio_space,
					lba_dev->hba.lmmio_space_offset);
	if (lba_dev->hba.gmmio_space.flags)
		pci_add_resource(&resources, &lba_dev->hba.gmmio_space);

	dev->dev.platform_data = lba_dev;
	lba_bus = lba_dev->hba.hba_bus =
		pci_create_root_bus(&dev->dev, lba_dev->hba.bus_num.start,
				    cfg_ops, NULL, &resources);
	if (!lba_bus) {
		pci_free_resource_list(&resources);
		return 0;
	}

	lba_bus->subordinate = pci_scan_child_bus(lba_bus);

	/* This is in lieu of calling pci_assign_unassigned_resources() */
	if (is_pdc_pat()) {
		/* assign resources to un-initialized devices */

		DBG_PAT("LBA pci_bus_size_bridges()\n");
		pci_bus_size_bridges(lba_bus);

		DBG_PAT("LBA pci_bus_assign_resources()\n");
		pci_bus_assign_resources(lba_bus);

#ifdef DEBUG_LBA_PAT
		DBG_PAT("\nLBA PIOP resource tree\n");
		lba_dump_res(&lba_dev->hba.io_space, 2);
		DBG_PAT("\nLBA LMMIO resource tree\n");
		lba_dump_res(&lba_dev->hba.lmmio_space, 2);
#endif
	}
	pci_enable_bridges(lba_bus);

	/*
	** Once PCI register ops has walked the bus, access to config
	** space is restricted. Avoids master aborts on config cycles.
	** Early LBA revs go fatal on *any* master abort.
	*/
	if (cfg_ops == &elroy_cfg_ops) {
		lba_dev->flags |= LBA_FLAG_SKIP_PROBE;
	}

	lba_next_bus = lba_bus->subordinate + 1;
	pci_bus_add_devices(lba_bus);

	/* Whew! Finally done! Tell services we got this one covered. */
	return 0;
}

static struct parisc_device_id lba_tbl[] = {
	{ HPHW_BRIDGE, HVERSION_REV_ANY_ID, ELROY_HVERS, 0xa },
	{ HPHW_BRIDGE, HVERSION_REV_ANY_ID, MERCURY_HVERS, 0xa },
	{ HPHW_BRIDGE, HVERSION_REV_ANY_ID, QUICKSILVER_HVERS, 0xa },
	{ 0, }
};

static struct parisc_driver lba_driver = {
	.name =		MODULE_NAME,
	.id_table =	lba_tbl,
	.probe =	lba_driver_probe,
};

/*
** One time initialization to let the world know the LBA was found.
** Must be called exactly once before pci_init().
*/
void __init lba_init(void)
{
	register_parisc_driver(&lba_driver);
}

/*
** Initialize the IBASE/IMASK registers for LBA (Elroy).
** Only called from sba_iommu.c in order to route ranges (MMIO vs DMA).
** sba_iommu is responsible for locking (none needed at init time).
*/
void lba_set_iregs(struct parisc_device *lba, u32 ibase, u32 imask)
{
	void __iomem * base_addr = ioremap_nocache(lba->hpa.start, 4096);

	imask <<= 2;	/* adjust for hints - 2 more bits */

	/* Make sure we aren't trying to set bits that aren't writeable. */
	WARN_ON((ibase & 0x001fffff) != 0);
	WARN_ON((imask & 0x001fffff) != 0);
	
	DBG("%s() ibase 0x%x imask 0x%x\n", __func__, ibase, imask);
	WRITE_REG32( imask, base_addr + LBA_IMASK);
	WRITE_REG32( ibase, base_addr + LBA_IBASE);
	iounmap(base_addr);
}