-
Notifications
You must be signed in to change notification settings - Fork 53.3k
/
in2000.c
2337 lines (1949 loc) · 71.8 KB
/
in2000.c
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
/*
* in2000.c - Linux device driver for the
* Always IN2000 ISA SCSI card.
*
* Copyright (c) 1996 John Shifflett, GeoLog Consulting
* john@geolog.com
* jshiffle@netcom.com
*
* 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, or (at your option)
* any later version.
*
* This program is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
* GNU General Public License for more details.
*
* For the avoidance of doubt the "preferred form" of this code is one which
* is in an open non patent encumbered format. Where cryptographic key signing
* forms part of the process of creating an executable the information
* including keys needed to generate an equivalently functional executable
* are deemed to be part of the source code.
*
* Drew Eckhardt's excellent 'Generic NCR5380' sources provided
* much of the inspiration and some of the code for this driver.
* The Linux IN2000 driver distributed in the Linux kernels through
* version 1.2.13 was an extremely valuable reference on the arcane
* (and still mysterious) workings of the IN2000's fifo. It also
* is where I lifted in2000_biosparam(), the gist of the card
* detection scheme, and other bits of code. Many thanks to the
* talented and courageous people who wrote, contributed to, and
* maintained that driver (including Brad McLean, Shaun Savage,
* Bill Earnest, Larry Doolittle, Roger Sunshine, John Luckey,
* Matt Postiff, Peter Lu, zerucha@shell.portal.com, and Eric
* Youngdale). I should also mention the driver written by
* Hamish Macdonald for the (GASP!) Amiga A2091 card, included
* in the Linux-m68k distribution; it gave me a good initial
* understanding of the proper way to run a WD33c93 chip, and I
* ended up stealing lots of code from it.
*
* _This_ driver is (I feel) an improvement over the old one in
* several respects:
* - All problems relating to the data size of a SCSI request are
* gone (as far as I know). The old driver couldn't handle
* swapping to partitions because that involved 4k blocks, nor
* could it deal with the st.c tape driver unmodified, because
* that usually involved 4k - 32k blocks. The old driver never
* quite got away from a morbid dependence on 2k block sizes -
* which of course is the size of the card's fifo.
*
* - Target Disconnection/Reconnection is now supported. Any
* system with more than one device active on the SCSI bus
* will benefit from this. The driver defaults to what I'm
* calling 'adaptive disconnect' - meaning that each command
* is evaluated individually as to whether or not it should
* be run with the option to disconnect/reselect (if the
* device chooses), or as a "SCSI-bus-hog".
*
* - Synchronous data transfers are now supported. Because there
* are a few devices (and many improperly terminated systems)
* that choke when doing sync, the default is sync DISABLED
* for all devices. This faster protocol can (and should!)
* be enabled on selected devices via the command-line.
*
* - Runtime operating parameters can now be specified through
* either the LILO or the 'insmod' command line. For LILO do:
* "in2000=blah,blah,blah"
* and with insmod go like:
* "insmod /usr/src/linux/modules/in2000.o setup_strings=blah,blah"
* The defaults should be good for most people. See the comment
* for 'setup_strings' below for more details.
*
* - The old driver relied exclusively on what the Western Digital
* docs call "Combination Level 2 Commands", which are a great
* idea in that the CPU is relieved of a lot of interrupt
* overhead. However, by accepting a certain (user-settable)
* amount of additional interrupts, this driver achieves
* better control over the SCSI bus, and data transfers are
* almost as fast while being much easier to define, track,
* and debug.
*
* - You can force detection of a card whose BIOS has been disabled.
*
* - Multiple IN2000 cards might almost be supported. I've tried to
* keep it in mind, but have no way to test...
*
*
* TODO:
* tagged queuing. multiple cards.
*
*
* NOTE:
* When using this or any other SCSI driver as a module, you'll
* find that with the stock kernel, at most _two_ SCSI hard
* drives will be linked into the device list (ie, usable).
* If your IN2000 card has more than 2 disks on its bus, you
* might want to change the define of 'SD_EXTRA_DEVS' in the
* 'hosts.h' file from 2 to whatever is appropriate. It took
* me a while to track down this surprisingly obscure and
* undocumented little "feature".
*
*
* People with bug reports, wish-lists, complaints, comments,
* or improvements are asked to pah-leeez email me (John Shifflett)
* at john@geolog.com or jshiffle@netcom.com! I'm anxious to get
* this thing into as good a shape as possible, and I'm positive
* there are lots of lurking bugs and "Stupid Places".
*
* Updated for Linux 2.5 by Alan Cox <alan@lxorguk.ukuu.org.uk>
* - Using new_eh handler
* - Hopefully got all the locking right again
* See "FIXME" notes for items that could do with more work
*/
#include <linux/module.h>
#include <linux/blkdev.h>
#include <linux/interrupt.h>
#include <linux/string.h>
#include <linux/delay.h>
#include <linux/proc_fs.h>
#include <linux/ioport.h>
#include <linux/stat.h>
#include <asm/io.h>
#include <asm/system.h>
#include "scsi.h"
#include <scsi/scsi_host.h>
#define IN2000_VERSION "1.33-2.5"
#define IN2000_DATE "2002/11/03"
#include "in2000.h"
/*
* 'setup_strings' is a single string used to pass operating parameters and
* settings from the kernel/module command-line to the driver. 'setup_args[]'
* is an array of strings that define the compile-time default values for
* these settings. If Linux boots with a LILO or insmod command-line, those
* settings are combined with 'setup_args[]'. Note that LILO command-lines
* are prefixed with "in2000=" while insmod uses a "setup_strings=" prefix.
* The driver recognizes the following keywords (lower case required) and
* arguments:
*
* - ioport:addr -Where addr is IO address of a (usually ROM-less) card.
* - noreset -No optional args. Prevents SCSI bus reset at boot time.
* - nosync:x -x is a bitmask where the 1st 7 bits correspond with
* the 7 possible SCSI devices (bit 0 for device #0, etc).
* Set a bit to PREVENT sync negotiation on that device.
* The driver default is sync DISABLED on all devices.
* - period:ns -ns is the minimum # of nanoseconds in a SCSI data transfer
* period. Default is 500; acceptable values are 250 - 1000.
* - disconnect:x -x = 0 to never allow disconnects, 2 to always allow them.
* x = 1 does 'adaptive' disconnects, which is the default
* and generally the best choice.
* - debug:x -If 'DEBUGGING_ON' is defined, x is a bitmask that causes
* various types of debug output to printed - see the DB_xxx
* defines in in2000.h
* - proc:x -If 'PROC_INTERFACE' is defined, x is a bitmask that
* determines how the /proc interface works and what it
* does - see the PR_xxx defines in in2000.h
*
* Syntax Notes:
* - Numeric arguments can be decimal or the '0x' form of hex notation. There
* _must_ be a colon between a keyword and its numeric argument, with no
* spaces.
* - Keywords are separated by commas, no spaces, in the standard kernel
* command-line manner.
* - A keyword in the 'nth' comma-separated command-line member will overwrite
* the 'nth' element of setup_args[]. A blank command-line member (in
* other words, a comma with no preceding keyword) will _not_ overwrite
* the corresponding setup_args[] element.
*
* A few LILO examples (for insmod, use 'setup_strings' instead of 'in2000'):
* - in2000=ioport:0x220,noreset
* - in2000=period:250,disconnect:2,nosync:0x03
* - in2000=debug:0x1e
* - in2000=proc:3
*/
/* Normally, no defaults are specified... */
static char *setup_args[] = { "", "", "", "", "", "", "", "", "" };
/* filled in by 'insmod' */
static char *setup_strings;
module_param(setup_strings, charp, 0);
static inline uchar read_3393(struct IN2000_hostdata *hostdata, uchar reg_num)
{
write1_io(reg_num, IO_WD_ADDR);
return read1_io(IO_WD_DATA);
}
#define READ_AUX_STAT() read1_io(IO_WD_ASR)
static inline void write_3393(struct IN2000_hostdata *hostdata, uchar reg_num, uchar value)
{
write1_io(reg_num, IO_WD_ADDR);
write1_io(value, IO_WD_DATA);
}
static inline void write_3393_cmd(struct IN2000_hostdata *hostdata, uchar cmd)
{
/* while (READ_AUX_STAT() & ASR_CIP)
printk("|");*/
write1_io(WD_COMMAND, IO_WD_ADDR);
write1_io(cmd, IO_WD_DATA);
}
static uchar read_1_byte(struct IN2000_hostdata *hostdata)
{
uchar asr, x = 0;
write_3393(hostdata, WD_CONTROL, CTRL_IDI | CTRL_EDI | CTRL_POLLED);
write_3393_cmd(hostdata, WD_CMD_TRANS_INFO | 0x80);
do {
asr = READ_AUX_STAT();
if (asr & ASR_DBR)
x = read_3393(hostdata, WD_DATA);
} while (!(asr & ASR_INT));
return x;
}
static void write_3393_count(struct IN2000_hostdata *hostdata, unsigned long value)
{
write1_io(WD_TRANSFER_COUNT_MSB, IO_WD_ADDR);
write1_io((value >> 16), IO_WD_DATA);
write1_io((value >> 8), IO_WD_DATA);
write1_io(value, IO_WD_DATA);
}
static unsigned long read_3393_count(struct IN2000_hostdata *hostdata)
{
unsigned long value;
write1_io(WD_TRANSFER_COUNT_MSB, IO_WD_ADDR);
value = read1_io(IO_WD_DATA) << 16;
value |= read1_io(IO_WD_DATA) << 8;
value |= read1_io(IO_WD_DATA);
return value;
}
/* The 33c93 needs to be told which direction a command transfers its
* data; we use this function to figure it out. Returns true if there
* will be a DATA_OUT phase with this command, false otherwise.
* (Thanks to Joerg Dorchain for the research and suggestion.)
*/
static int is_dir_out(Scsi_Cmnd * cmd)
{
switch (cmd->cmnd[0]) {
case WRITE_6:
case WRITE_10:
case WRITE_12:
case WRITE_LONG:
case WRITE_SAME:
case WRITE_BUFFER:
case WRITE_VERIFY:
case WRITE_VERIFY_12:
case COMPARE:
case COPY:
case COPY_VERIFY:
case SEARCH_EQUAL:
case SEARCH_HIGH:
case SEARCH_LOW:
case SEARCH_EQUAL_12:
case SEARCH_HIGH_12:
case SEARCH_LOW_12:
case FORMAT_UNIT:
case REASSIGN_BLOCKS:
case RESERVE:
case MODE_SELECT:
case MODE_SELECT_10:
case LOG_SELECT:
case SEND_DIAGNOSTIC:
case CHANGE_DEFINITION:
case UPDATE_BLOCK:
case SET_WINDOW:
case MEDIUM_SCAN:
case SEND_VOLUME_TAG:
case 0xea:
return 1;
default:
return 0;
}
}
static struct sx_period sx_table[] = {
{1, 0x20},
{252, 0x20},
{376, 0x30},
{500, 0x40},
{624, 0x50},
{752, 0x60},
{876, 0x70},
{1000, 0x00},
{0, 0}
};
static int round_period(unsigned int period)
{
int x;
for (x = 1; sx_table[x].period_ns; x++) {
if ((period <= sx_table[x - 0].period_ns) && (period > sx_table[x - 1].period_ns)) {
return x;
}
}
return 7;
}
static uchar calc_sync_xfer(unsigned int period, unsigned int offset)
{
uchar result;
period *= 4; /* convert SDTR code to ns */
result = sx_table[round_period(period)].reg_value;
result |= (offset < OPTIMUM_SX_OFF) ? offset : OPTIMUM_SX_OFF;
return result;
}
static void in2000_execute(struct Scsi_Host *instance);
static int in2000_queuecommand_lck(Scsi_Cmnd * cmd, void (*done) (Scsi_Cmnd *))
{
struct Scsi_Host *instance;
struct IN2000_hostdata *hostdata;
Scsi_Cmnd *tmp;
instance = cmd->device->host;
hostdata = (struct IN2000_hostdata *) instance->hostdata;
DB(DB_QUEUE_COMMAND, scmd_printk(KERN_DEBUG, cmd, "Q-%02x(", cmd->cmnd[0]))
/* Set up a few fields in the Scsi_Cmnd structure for our own use:
* - host_scribble is the pointer to the next cmd in the input queue
* - scsi_done points to the routine we call when a cmd is finished
* - result is what you'd expect
*/
cmd->host_scribble = NULL;
cmd->scsi_done = done;
cmd->result = 0;
/* We use the Scsi_Pointer structure that's included with each command
* as a scratchpad (as it's intended to be used!). The handy thing about
* the SCp.xxx fields is that they're always associated with a given
* cmd, and are preserved across disconnect-reselect. This means we
* can pretty much ignore SAVE_POINTERS and RESTORE_POINTERS messages
* if we keep all the critical pointers and counters in SCp:
* - SCp.ptr is the pointer into the RAM buffer
* - SCp.this_residual is the size of that buffer
* - SCp.buffer points to the current scatter-gather buffer
* - SCp.buffers_residual tells us how many S.G. buffers there are
* - SCp.have_data_in helps keep track of >2048 byte transfers
* - SCp.sent_command is not used
* - SCp.phase records this command's SRCID_ER bit setting
*/
if (scsi_bufflen(cmd)) {
cmd->SCp.buffer = scsi_sglist(cmd);
cmd->SCp.buffers_residual = scsi_sg_count(cmd) - 1;
cmd->SCp.ptr = sg_virt(cmd->SCp.buffer);
cmd->SCp.this_residual = cmd->SCp.buffer->length;
} else {
cmd->SCp.buffer = NULL;
cmd->SCp.buffers_residual = 0;
cmd->SCp.ptr = NULL;
cmd->SCp.this_residual = 0;
}
cmd->SCp.have_data_in = 0;
/* We don't set SCp.phase here - that's done in in2000_execute() */
/* WD docs state that at the conclusion of a "LEVEL2" command, the
* status byte can be retrieved from the LUN register. Apparently,
* this is the case only for *uninterrupted* LEVEL2 commands! If
* there are any unexpected phases entered, even if they are 100%
* legal (different devices may choose to do things differently),
* the LEVEL2 command sequence is exited. This often occurs prior
* to receiving the status byte, in which case the driver does a
* status phase interrupt and gets the status byte on its own.
* While such a command can then be "resumed" (ie restarted to
* finish up as a LEVEL2 command), the LUN register will NOT be
* a valid status byte at the command's conclusion, and we must
* use the byte obtained during the earlier interrupt. Here, we
* preset SCp.Status to an illegal value (0xff) so that when
* this command finally completes, we can tell where the actual
* status byte is stored.
*/
cmd->SCp.Status = ILLEGAL_STATUS_BYTE;
/* We need to disable interrupts before messing with the input
* queue and calling in2000_execute().
*/
/*
* Add the cmd to the end of 'input_Q'. Note that REQUEST_SENSE
* commands are added to the head of the queue so that the desired
* sense data is not lost before REQUEST_SENSE executes.
*/
if (!(hostdata->input_Q) || (cmd->cmnd[0] == REQUEST_SENSE)) {
cmd->host_scribble = (uchar *) hostdata->input_Q;
hostdata->input_Q = cmd;
} else { /* find the end of the queue */
for (tmp = (Scsi_Cmnd *) hostdata->input_Q; tmp->host_scribble; tmp = (Scsi_Cmnd *) tmp->host_scribble);
tmp->host_scribble = (uchar *) cmd;
}
/* We know that there's at least one command in 'input_Q' now.
* Go see if any of them are runnable!
*/
in2000_execute(cmd->device->host);
DB(DB_QUEUE_COMMAND, printk(")Q "))
return 0;
}
static DEF_SCSI_QCMD(in2000_queuecommand)
/*
* This routine attempts to start a scsi command. If the host_card is
* already connected, we give up immediately. Otherwise, look through
* the input_Q, using the first command we find that's intended
* for a currently non-busy target/lun.
* Note that this function is always called with interrupts already
* disabled (either from in2000_queuecommand() or in2000_intr()).
*/
static void in2000_execute(struct Scsi_Host *instance)
{
struct IN2000_hostdata *hostdata;
Scsi_Cmnd *cmd, *prev;
int i;
unsigned short *sp;
unsigned short f;
unsigned short flushbuf[16];
hostdata = (struct IN2000_hostdata *) instance->hostdata;
DB(DB_EXECUTE, printk("EX("))
if (hostdata->selecting || hostdata->connected) {
DB(DB_EXECUTE, printk(")EX-0 "))
return;
}
/*
* Search through the input_Q for a command destined
* for an idle target/lun.
*/
cmd = (Scsi_Cmnd *) hostdata->input_Q;
prev = NULL;
while (cmd) {
if (!(hostdata->busy[cmd->device->id] & (1 << cmd->device->lun)))
break;
prev = cmd;
cmd = (Scsi_Cmnd *) cmd->host_scribble;
}
/* quit if queue empty or all possible targets are busy */
if (!cmd) {
DB(DB_EXECUTE, printk(")EX-1 "))
return;
}
/* remove command from queue */
if (prev)
prev->host_scribble = cmd->host_scribble;
else
hostdata->input_Q = (Scsi_Cmnd *) cmd->host_scribble;
#ifdef PROC_STATISTICS
hostdata->cmd_cnt[cmd->device->id]++;
#endif
/*
* Start the selection process
*/
if (is_dir_out(cmd))
write_3393(hostdata, WD_DESTINATION_ID, cmd->device->id);
else
write_3393(hostdata, WD_DESTINATION_ID, cmd->device->id | DSTID_DPD);
/* Now we need to figure out whether or not this command is a good
* candidate for disconnect/reselect. We guess to the best of our
* ability, based on a set of hierarchical rules. When several
* devices are operating simultaneously, disconnects are usually
* an advantage. In a single device system, or if only 1 device
* is being accessed, transfers usually go faster if disconnects
* are not allowed:
*
* + Commands should NEVER disconnect if hostdata->disconnect =
* DIS_NEVER (this holds for tape drives also), and ALWAYS
* disconnect if hostdata->disconnect = DIS_ALWAYS.
* + Tape drive commands should always be allowed to disconnect.
* + Disconnect should be allowed if disconnected_Q isn't empty.
* + Commands should NOT disconnect if input_Q is empty.
* + Disconnect should be allowed if there are commands in input_Q
* for a different target/lun. In this case, the other commands
* should be made disconnect-able, if not already.
*
* I know, I know - this code would flunk me out of any
* "C Programming 101" class ever offered. But it's easy
* to change around and experiment with for now.
*/
cmd->SCp.phase = 0; /* assume no disconnect */
if (hostdata->disconnect == DIS_NEVER)
goto no;
if (hostdata->disconnect == DIS_ALWAYS)
goto yes;
if (cmd->device->type == 1) /* tape drive? */
goto yes;
if (hostdata->disconnected_Q) /* other commands disconnected? */
goto yes;
if (!(hostdata->input_Q)) /* input_Q empty? */
goto no;
for (prev = (Scsi_Cmnd *) hostdata->input_Q; prev; prev = (Scsi_Cmnd *) prev->host_scribble) {
if ((prev->device->id != cmd->device->id) || (prev->device->lun != cmd->device->lun)) {
for (prev = (Scsi_Cmnd *) hostdata->input_Q; prev; prev = (Scsi_Cmnd *) prev->host_scribble)
prev->SCp.phase = 1;
goto yes;
}
}
goto no;
yes:
cmd->SCp.phase = 1;
#ifdef PROC_STATISTICS
hostdata->disc_allowed_cnt[cmd->device->id]++;
#endif
no:
write_3393(hostdata, WD_SOURCE_ID, ((cmd->SCp.phase) ? SRCID_ER : 0));
write_3393(hostdata, WD_TARGET_LUN, cmd->device->lun);
write_3393(hostdata, WD_SYNCHRONOUS_TRANSFER, hostdata->sync_xfer[cmd->device->id]);
hostdata->busy[cmd->device->id] |= (1 << cmd->device->lun);
if ((hostdata->level2 <= L2_NONE) || (hostdata->sync_stat[cmd->device->id] == SS_UNSET)) {
/*
* Do a 'Select-With-ATN' command. This will end with
* one of the following interrupts:
* CSR_RESEL_AM: failure - can try again later.
* CSR_TIMEOUT: failure - give up.
* CSR_SELECT: success - proceed.
*/
hostdata->selecting = cmd;
/* Every target has its own synchronous transfer setting, kept in
* the sync_xfer array, and a corresponding status byte in sync_stat[].
* Each target's sync_stat[] entry is initialized to SS_UNSET, and its
* sync_xfer[] entry is initialized to the default/safe value. SS_UNSET
* means that the parameters are undetermined as yet, and that we
* need to send an SDTR message to this device after selection is
* complete. We set SS_FIRST to tell the interrupt routine to do so,
* unless we don't want to even _try_ synchronous transfers: In this
* case we set SS_SET to make the defaults final.
*/
if (hostdata->sync_stat[cmd->device->id] == SS_UNSET) {
if (hostdata->sync_off & (1 << cmd->device->id))
hostdata->sync_stat[cmd->device->id] = SS_SET;
else
hostdata->sync_stat[cmd->device->id] = SS_FIRST;
}
hostdata->state = S_SELECTING;
write_3393_count(hostdata, 0); /* this guarantees a DATA_PHASE interrupt */
write_3393_cmd(hostdata, WD_CMD_SEL_ATN);
}
else {
/*
* Do a 'Select-With-ATN-Xfer' command. This will end with
* one of the following interrupts:
* CSR_RESEL_AM: failure - can try again later.
* CSR_TIMEOUT: failure - give up.
* anything else: success - proceed.
*/
hostdata->connected = cmd;
write_3393(hostdata, WD_COMMAND_PHASE, 0);
/* copy command_descriptor_block into WD chip
* (take advantage of auto-incrementing)
*/
write1_io(WD_CDB_1, IO_WD_ADDR);
for (i = 0; i < cmd->cmd_len; i++)
write1_io(cmd->cmnd[i], IO_WD_DATA);
/* The wd33c93 only knows about Group 0, 1, and 5 commands when
* it's doing a 'select-and-transfer'. To be safe, we write the
* size of the CDB into the OWN_ID register for every case. This
* way there won't be problems with vendor-unique, audio, etc.
*/
write_3393(hostdata, WD_OWN_ID, cmd->cmd_len);
/* When doing a non-disconnect command, we can save ourselves a DATA
* phase interrupt later by setting everything up now. With writes we
* need to pre-fill the fifo; if there's room for the 32 flush bytes,
* put them in there too - that'll avoid a fifo interrupt. Reads are
* somewhat simpler.
* KLUDGE NOTE: It seems that you can't completely fill the fifo here:
* This results in the IO_FIFO_COUNT register rolling over to zero,
* and apparently the gate array logic sees this as empty, not full,
* so the 3393 chip is never signalled to start reading from the
* fifo. Or maybe it's seen as a permanent fifo interrupt condition.
* Regardless, we fix this by temporarily pretending that the fifo
* is 16 bytes smaller. (I see now that the old driver has a comment
* about "don't fill completely" in an analogous place - must be the
* same deal.) This results in CDROM, swap partitions, and tape drives
* needing an extra interrupt per write command - I think we can live
* with that!
*/
if (!(cmd->SCp.phase)) {
write_3393_count(hostdata, cmd->SCp.this_residual);
write_3393(hostdata, WD_CONTROL, CTRL_IDI | CTRL_EDI | CTRL_BUS);
write1_io(0, IO_FIFO_WRITE); /* clear fifo counter, write mode */
if (is_dir_out(cmd)) {
hostdata->fifo = FI_FIFO_WRITING;
if ((i = cmd->SCp.this_residual) > (IN2000_FIFO_SIZE - 16))
i = IN2000_FIFO_SIZE - 16;
cmd->SCp.have_data_in = i; /* this much data in fifo */
i >>= 1; /* Gulp. Assuming modulo 2. */
sp = (unsigned short *) cmd->SCp.ptr;
f = hostdata->io_base + IO_FIFO;
#ifdef FAST_WRITE_IO
FAST_WRITE2_IO();
#else
while (i--)
write2_io(*sp++, IO_FIFO);
#endif
/* Is there room for the flush bytes? */
if (cmd->SCp.have_data_in <= ((IN2000_FIFO_SIZE - 16) - 32)) {
sp = flushbuf;
i = 16;
#ifdef FAST_WRITE_IO
FAST_WRITE2_IO();
#else
while (i--)
write2_io(0, IO_FIFO);
#endif
}
}
else {
write1_io(0, IO_FIFO_READ); /* put fifo in read mode */
hostdata->fifo = FI_FIFO_READING;
cmd->SCp.have_data_in = 0; /* nothing transferred yet */
}
} else {
write_3393_count(hostdata, 0); /* this guarantees a DATA_PHASE interrupt */
}
hostdata->state = S_RUNNING_LEVEL2;
write_3393_cmd(hostdata, WD_CMD_SEL_ATN_XFER);
}
/*
* Since the SCSI bus can handle only 1 connection at a time,
* we get out of here now. If the selection fails, or when
* the command disconnects, we'll come back to this routine
* to search the input_Q again...
*/
DB(DB_EXECUTE, printk("%s)EX-2 ", (cmd->SCp.phase) ? "d:" : ""))
}
static void transfer_pio(uchar * buf, int cnt, int data_in_dir, struct IN2000_hostdata *hostdata)
{
uchar asr;
DB(DB_TRANSFER, printk("(%p,%d,%s)", buf, cnt, data_in_dir ? "in" : "out"))
write_3393(hostdata, WD_CONTROL, CTRL_IDI | CTRL_EDI | CTRL_POLLED);
write_3393_count(hostdata, cnt);
write_3393_cmd(hostdata, WD_CMD_TRANS_INFO);
if (data_in_dir) {
do {
asr = READ_AUX_STAT();
if (asr & ASR_DBR)
*buf++ = read_3393(hostdata, WD_DATA);
} while (!(asr & ASR_INT));
} else {
do {
asr = READ_AUX_STAT();
if (asr & ASR_DBR)
write_3393(hostdata, WD_DATA, *buf++);
} while (!(asr & ASR_INT));
}
/* Note: we are returning with the interrupt UN-cleared.
* Since (presumably) an entire I/O operation has
* completed, the bus phase is probably different, and
* the interrupt routine will discover this when it
* responds to the uncleared int.
*/
}
static void transfer_bytes(Scsi_Cmnd * cmd, int data_in_dir)
{
struct IN2000_hostdata *hostdata;
unsigned short *sp;
unsigned short f;
int i;
hostdata = (struct IN2000_hostdata *) cmd->device->host->hostdata;
/* Normally, you'd expect 'this_residual' to be non-zero here.
* In a series of scatter-gather transfers, however, this
* routine will usually be called with 'this_residual' equal
* to 0 and 'buffers_residual' non-zero. This means that a
* previous transfer completed, clearing 'this_residual', and
* now we need to setup the next scatter-gather buffer as the
* source or destination for THIS transfer.
*/
if (!cmd->SCp.this_residual && cmd->SCp.buffers_residual) {
++cmd->SCp.buffer;
--cmd->SCp.buffers_residual;
cmd->SCp.this_residual = cmd->SCp.buffer->length;
cmd->SCp.ptr = sg_virt(cmd->SCp.buffer);
}
/* Set up hardware registers */
write_3393(hostdata, WD_SYNCHRONOUS_TRANSFER, hostdata->sync_xfer[cmd->device->id]);
write_3393_count(hostdata, cmd->SCp.this_residual);
write_3393(hostdata, WD_CONTROL, CTRL_IDI | CTRL_EDI | CTRL_BUS);
write1_io(0, IO_FIFO_WRITE); /* zero counter, assume write */
/* Reading is easy. Just issue the command and return - we'll
* get an interrupt later when we have actual data to worry about.
*/
if (data_in_dir) {
write1_io(0, IO_FIFO_READ);
if ((hostdata->level2 >= L2_DATA) || (hostdata->level2 == L2_BASIC && cmd->SCp.phase == 0)) {
write_3393(hostdata, WD_COMMAND_PHASE, 0x45);
write_3393_cmd(hostdata, WD_CMD_SEL_ATN_XFER);
hostdata->state = S_RUNNING_LEVEL2;
} else
write_3393_cmd(hostdata, WD_CMD_TRANS_INFO);
hostdata->fifo = FI_FIFO_READING;
cmd->SCp.have_data_in = 0;
return;
}
/* Writing is more involved - we'll start the WD chip and write as
* much data to the fifo as we can right now. Later interrupts will
* write any bytes that don't make it at this stage.
*/
if ((hostdata->level2 >= L2_DATA) || (hostdata->level2 == L2_BASIC && cmd->SCp.phase == 0)) {
write_3393(hostdata, WD_COMMAND_PHASE, 0x45);
write_3393_cmd(hostdata, WD_CMD_SEL_ATN_XFER);
hostdata->state = S_RUNNING_LEVEL2;
} else
write_3393_cmd(hostdata, WD_CMD_TRANS_INFO);
hostdata->fifo = FI_FIFO_WRITING;
sp = (unsigned short *) cmd->SCp.ptr;
if ((i = cmd->SCp.this_residual) > IN2000_FIFO_SIZE)
i = IN2000_FIFO_SIZE;
cmd->SCp.have_data_in = i;
i >>= 1; /* Gulp. We assume this_residual is modulo 2 */
f = hostdata->io_base + IO_FIFO;
#ifdef FAST_WRITE_IO
FAST_WRITE2_IO();
#else
while (i--)
write2_io(*sp++, IO_FIFO);
#endif
}
/* We need to use spin_lock_irqsave() & spin_unlock_irqrestore() in this
* function in order to work in an SMP environment. (I'd be surprised
* if the driver is ever used by anyone on a real multi-CPU motherboard,
* but it _does_ need to be able to compile and run in an SMP kernel.)
*/
static irqreturn_t in2000_intr(int irqnum, void *dev_id)
{
struct Scsi_Host *instance = dev_id;
struct IN2000_hostdata *hostdata;
Scsi_Cmnd *patch, *cmd;
uchar asr, sr, phs, id, lun, *ucp, msg;
int i, j;
unsigned long length;
unsigned short *sp;
unsigned short f;
unsigned long flags;
hostdata = (struct IN2000_hostdata *) instance->hostdata;
/* Get the spin_lock and disable further ints, for SMP */
spin_lock_irqsave(instance->host_lock, flags);
#ifdef PROC_STATISTICS
hostdata->int_cnt++;
#endif
/* The IN2000 card has 2 interrupt sources OR'ed onto its IRQ line - the
* WD3393 chip and the 2k fifo (which is actually a dual-port RAM combined
* with a big logic array, so it's a little different than what you might
* expect). As far as I know, there's no reason that BOTH can't be active
* at the same time, but there's a problem: while we can read the 3393
* to tell if _it_ wants an interrupt, I don't know of a way to ask the
* fifo the same question. The best we can do is check the 3393 and if
* it _isn't_ the source of the interrupt, then we can be pretty sure
* that the fifo is the culprit.
* UPDATE: I have it on good authority (Bill Earnest) that bit 0 of the
* IO_FIFO_COUNT register mirrors the fifo interrupt state. I
* assume that bit clear means interrupt active. As it turns
* out, the driver really doesn't need to check for this after
* all, so my remarks above about a 'problem' can safely be
* ignored. The way the logic is set up, there's no advantage
* (that I can see) to worrying about it.
*
* It seems that the fifo interrupt signal is negated when we extract
* bytes during read or write bytes during write.
* - fifo will interrupt when data is moving from it to the 3393, and
* there are 31 (or less?) bytes left to go. This is sort of short-
* sighted: what if you don't WANT to do more? In any case, our
* response is to push more into the fifo - either actual data or
* dummy bytes if need be. Note that we apparently have to write at
* least 32 additional bytes to the fifo after an interrupt in order
* to get it to release the ones it was holding on to - writing fewer
* than 32 will result in another fifo int.
* UPDATE: Again, info from Bill Earnest makes this more understandable:
* 32 bytes = two counts of the fifo counter register. He tells
* me that the fifo interrupt is a non-latching signal derived
* from a straightforward boolean interpretation of the 7
* highest bits of the fifo counter and the fifo-read/fifo-write
* state. Who'd a thought?
*/
write1_io(0, IO_LED_ON);
asr = READ_AUX_STAT();
if (!(asr & ASR_INT)) { /* no WD33c93 interrupt? */
/* Ok. This is definitely a FIFO-only interrupt.
*
* If FI_FIFO_READING is set, there are up to 2048 bytes waiting to be read,
* maybe more to come from the SCSI bus. Read as many as we can out of the
* fifo and into memory at the location of SCp.ptr[SCp.have_data_in], and
* update have_data_in afterwards.
*
* If we have FI_FIFO_WRITING, the FIFO has almost run out of bytes to move
* into the WD3393 chip (I think the interrupt happens when there are 31
* bytes left, but it may be fewer...). The 3393 is still waiting, so we
* shove some more into the fifo, which gets things moving again. If the
* original SCSI command specified more than 2048 bytes, there may still
* be some of that data left: fine - use it (from SCp.ptr[SCp.have_data_in]).
* Don't forget to update have_data_in. If we've already written out the
* entire buffer, feed 32 dummy bytes to the fifo - they're needed to
* push out the remaining real data.
* (Big thanks to Bill Earnest for getting me out of the mud in here.)
*/
cmd = (Scsi_Cmnd *) hostdata->connected; /* assume we're connected */
CHECK_NULL(cmd, "fifo_int")
if (hostdata->fifo == FI_FIFO_READING) {
DB(DB_FIFO, printk("{R:%02x} ", read1_io(IO_FIFO_COUNT)))
sp = (unsigned short *) (cmd->SCp.ptr + cmd->SCp.have_data_in);
i = read1_io(IO_FIFO_COUNT) & 0xfe;
i <<= 2; /* # of words waiting in the fifo */
f = hostdata->io_base + IO_FIFO;
#ifdef FAST_READ_IO
FAST_READ2_IO();
#else
while (i--)
*sp++ = read2_io(IO_FIFO);
#endif
i = sp - (unsigned short *) (cmd->SCp.ptr + cmd->SCp.have_data_in);
i <<= 1;
cmd->SCp.have_data_in += i;
}
else if (hostdata->fifo == FI_FIFO_WRITING) {
DB(DB_FIFO, printk("{W:%02x} ", read1_io(IO_FIFO_COUNT)))
/* If all bytes have been written to the fifo, flush out the stragglers.
* Note that while writing 16 dummy words seems arbitrary, we don't
* have another choice that I can see. What we really want is to read
* the 3393 transfer count register (that would tell us how many bytes
* needed flushing), but the TRANSFER_INFO command hasn't completed
* yet (not enough bytes!) and that register won't be accessible. So,
* we use 16 words - a number obtained through trial and error.
* UPDATE: Bill says this is exactly what Always does, so there.
* More thanks due him for help in this section.
*/
if (cmd->SCp.this_residual == cmd->SCp.have_data_in) {
i = 16;
while (i--) /* write 32 dummy bytes */
write2_io(0, IO_FIFO);
}
/* If there are still bytes left in the SCSI buffer, write as many as we
* can out to the fifo.
*/
else {
sp = (unsigned short *) (cmd->SCp.ptr + cmd->SCp.have_data_in);
i = cmd->SCp.this_residual - cmd->SCp.have_data_in; /* bytes yet to go */
j = read1_io(IO_FIFO_COUNT) & 0xfe;
j <<= 2; /* how many words the fifo has room for */
if ((j << 1) > i)
j = (i >> 1);
while (j--)
write2_io(*sp++, IO_FIFO);
i = sp - (unsigned short *) (cmd->SCp.ptr + cmd->SCp.have_data_in);
i <<= 1;
cmd->SCp.have_data_in += i;
}
}
else {
printk("*** Spurious FIFO interrupt ***");
}
write1_io(0, IO_LED_OFF);
/* release the SMP spin_lock and restore irq state */
spin_unlock_irqrestore(instance->host_lock, flags);
return IRQ_HANDLED;
}
/* This interrupt was triggered by the WD33c93 chip. The fifo interrupt
* may also be asserted, but we don't bother to check it: we get more
* detailed info from FIFO_READING and FIFO_WRITING (see below).
*/
cmd = (Scsi_Cmnd *) hostdata->connected; /* assume we're connected */
sr = read_3393(hostdata, WD_SCSI_STATUS); /* clear the interrupt */
phs = read_3393(hostdata, WD_COMMAND_PHASE);
if (!cmd && (sr != CSR_RESEL_AM && sr != CSR_TIMEOUT && sr != CSR_SELECT)) {