1
2      specialix.txt  -- specialix IO8+ multiport serial driver readme.
3
4
5
6      Copyright (C) 1997  Roger Wolff (R.E.Wolff@BitWizard.nl)
7
8      Specialix pays for the development and support of this driver.
9      Please DO contact io8-linux@specialix.co.uk if you require
10      support.
11
12      This driver was developed in the BitWizard linux device
13      driver service. If you require a linux device driver for your
14      product, please contact devices@BitWizard.nl for a quote.
15
16      This code is firmly based on the riscom/8 serial driver,
17      written by Dmitry Gorodchanin. The specialix IO8+ card
18      programming information was obtained from the CL-CD1865 Data
19      Book, and Specialix document number 6200059: IO8+ Hardware
20      Functional Specification, augmented by document number 6200088:
21      Merak Hardware Functional Specification. (IO8+/PCI is also
22      called Merak)
23
24
25      This program is free software; you can redistribute it and/or
26      modify it under the terms of the GNU General Public License as
27      published by the Free Software Foundation; either version 2 of
28      the License, or (at your option) any later version.
29
30      This program is distributed in the hope that it will be
31      useful, but WITHOUT ANY WARRANTY; without even the implied
32      warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR
33      PURPOSE.  See the GNU General Public License for more details.
34
35      You should have received a copy of the GNU General Public
36      License along with this program; if not, write to the Free
37      Software Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139,
38      USA.
39
40
41Intro
42=====
43
44
45This file contains some random information, that I like to have online
46instead of in a manual that can get lost. Ever misplace your Linux
47kernel sources?  And the manual of one of the boards in your computer?
48
49
50Addresses and interrupts
51========================
52
53Address dip switch settings:
54The dip switch sets bits 2-9 of the IO address.
55
56       9 8 7 6 5 4 3 2
57     +-----------------+
58   0 | X   X X X X X X |
59     |                 |    =   IoBase = 0x100
60   1 |   X             |
61     +-----------------+          ------ RS232 connectors ---->
62
63         |    |    |
64       edge connector
65         |    |    |
66         V    V    V
67
68Base address 0x100 caused a conflict in one of my computers once.  I
69haven't the foggiest why. My Specialix card is now at 0x180.  My
70other computer runs just fine with the Specialix card at 0x100....
71The card occupies 4 addresses, but actually only two are really used.
72
73The PCI version doesn't have any dip switches. The BIOS assigns
74an IO address.
75
76The driver now still autoprobes at 0x100, 0x180, 0x250 and 0x260.  If
77that causes trouble for you, please report that. I'll remove
78autoprobing then.
79
80The driver will tell the card what IRQ to use, so you don't have to
81change any jumpers to change the IRQ. Just use a command line
82argument (irq=xx) to the insmod program to set the interrupt.
83
84The BIOS assigns the IRQ on the PCI version. You have no say in what
85IRQ to use in that case.
86
87If your specialix cards are not at the default locations, you can use
88the kernel command line argument "specialix=io0,irq0,io1,irq1...".
89Here "io0" is the io address for the first card, and "irq0" is the
90irq line that the first card should use. And so on.
91
92Examples.
93
94You use the driver as a module and have three cards at 0x100, 0x250
95and 0x180. And some way or another you want them detected in that
96order. Moreover irq 12 is taken (e.g. by your PS/2 mouse).
97
98  insmod specialix.o iobase=0x100,0x250,0x180 irq=9,11,15
99
100The same three cards, but now in the kernel would require you to
101add
102
103   specialix=0x100,9,0x250,11,0x180,15
104
105to the command line. This would become
106
107   append="specialix=0x100,9,0x250,11,0x180,15"
108
109in your /etc/lilo.conf file if you use lilo.
110
111The Specialix driver is slightly odd: It allows you to have the second
112or third card detected without having a first card. This has
113advantages and disadvantages. A slot that isn't filled by an ISA card,
114might be filled if a PCI card is detected. Thus if you have an ISA
115card at 0x250 and a PCI card, you would get:
116
117sx0: specialix IO8+ Board at 0x100 not found.
118sx1: specialix IO8+ Board at 0x180 not found.
119sx2: specialix IO8+ board detected at 0x250, IRQ 12, CD1865 Rev. B.
120sx3: specialix IO8+ Board at 0x260 not found.
121sx0: specialix IO8+ board detected at 0xd800, IRQ 9, CD1865 Rev. B.
122
123This would happen if you don't give any probe hints to the driver.
124If you would specify:
125
126   specialix=0x250,11
127
128you'd get the following messages:
129
130sx0: specialix IO8+ board detected at 0x250, IRQ 11, CD1865 Rev. B.
131sx1: specialix IO8+ board detected at 0xd800, IRQ 9, CD1865 Rev. B.
132
133ISA probing is aborted after the IO address you gave is exhausted, and
134the PCI card is now detected as the second card. The ISA card is now
135also forced to IRQ11....
136
137
138Baud rates
139==========
140
141The rev 1.2 and below boards use a CL-CD1864. These chips can only
142do 64kbit. The rev 1.3 and newer boards use a CL-CD1865. These chips
143are officially capable of 115k2.
144
145The Specialix card uses a 25MHz crystal (in times two mode, which in
146fact is a divided by two mode). This is not enough to reach the rated
147115k2 on all ports at the same time. With this clock rate you can only
148do 37% of this rate. This means that at 115k2 on all ports you are
149going to lose characters (The chip cannot handle that many incoming
150bits at this clock rate.) (Yes, you read that correctly: there is a
151limit to the number of -=bits=- per second that the chip can handle.)
152
153If you near the "limit" you will first start to see a graceful
154degradation in that the chip cannot keep the transmitter busy at all
155times. However with a central clock this slow, you can also get it to
156miss incoming characters. The driver will print a warning message when
157you are outside the official specs. The messages usually show up in
158the file /var/log/messages .
159
160The specialix card cannot reliably do 115k2. If you use it, you have
161to do "extensive testing" (*) to verify if it actually works.
162
163When "mgetty" communicates with my modem at 115k2 it reports:
164got: +++[0d]ATQ0V1H0[0d][0d][8a]O[cb][0d][8a]
165                            ^^^^ ^^^^    ^^^^
166
167The three characters that have the "^^^" under them have suffered a
168bit error in the highest bit. In conclusion: I've tested it, and found
169that it simply DOESN'T work for me. I also suspect that this is also
170caused by the baud rate being just a little bit out of tune.
171
172I upgraded the crystal to 66Mhz on one of my Specialix cards. Works
173great! Contact me for details. (Voids warranty, requires a steady hand
174and more such restrictions....)
175
176
177(*) Cirrus logic CD1864 databook, page 40.
178
179
180Cables for the Specialix IO8+
181=============================
182
183The pinout of the connectors on the IO8+ is:
184
185     pin    short    direction    long name
186            name
187    Pin 1   DCD      input        Data Carrier Detect
188    Pin 2   RXD      input        Receive
189    Pin 3   DTR/RTS  output       Data Terminal Ready/Ready To Send
190    Pin 4   GND      -            Ground
191    Pin 5   TXD      output       Transmit
192    Pin 6   CTS      input        Clear To Send
193
194
195             -- 6  5  4  3  2  1 --
196             |                    |
197             |                    |
198             |                    |
199             |                    |
200             +-----          -----+
201                  |__________|
202                      clip
203
204    Front view of an RJ12 connector. Cable moves "into" the paper.
205    (the plug is ready to plug into your mouth this way...)
206
207
208    NULL cable. I don't know who is going to use these except for
209    testing purposes, but I tested the cards with this cable. (It
210    took quite a while to figure out, so I'm not going to delete
211    it. So there! :-)
212
213
214    This end goes               This end needs
215    straight into the           some twists in
216    RJ12 plug.                  the wiring.
217       IO8+ RJ12                   IO8+ RJ12
218        1  DCD       white          -
219        -             -             1 DCD
220        2  RXD       black          5 TXD
221        3  DTR/RTS   red            6 CTS
222        4  GND       green          4 GND
223        5  TXD       yellow         2 RXD
224        6  CTS       blue           3 DTR/RTS
225
226
227    Same NULL cable, but now sorted on the second column.
228
229        1  DCD       white          -
230        -             -             1 DCD
231        5  TXD       yellow         2 RXD
232        6  CTS       blue           3 DTR/RTS
233        4  GND       green          4 GND
234        2  RXD       black          5 TXD
235        3  DTR/RTS   red            6 CTS
236
237
238
239    This is a modem cable usable for hardware handshaking:
240        RJ12                        DB25           DB9
241        1   DCD      white          8 DCD          1 DCD
242        2   RXD      black          3 RXD          2 RXD
243        3   DTR/RTS  red            4 RTS          7 RTS
244        4   GND      green          7 GND          5 GND
245        5   TXD      yellow         2 TXD          3 TXD
246        6   CTS      blue           5 CTS          8 CTS
247                            +----   6 DSR          6 DSR
248                            +----  20 DTR          4 DTR
249
250    This is a modem cable usable for software handshaking:
251    It allows you to reset the modem using the DTR ioctls.
252    I (REW) have never tested this, "but xxxxxxxxxxxxx
253    says that it works." If you test this, please
254    tell me and I'll fill in your name on the xxx's.
255
256        RJ12                        DB25           DB9
257        1   DCD      white          8 DCD          1 DCD
258        2   RXD      black          3 RXD          2 RXD
259        3   DTR/RTS  red           20 DTR          4 DTR
260        4   GND      green          7 GND          5 GND
261        5   TXD      yellow         2 TXD          3 TXD
262        6   CTS      blue           5 CTS          8 CTS
263                            +----   6 DSR          6 DSR
264                            +----   4 RTS          7 RTS
265
266   I bought a 6 wire flat cable. It was colored as indicated.
267   Check that yours is the same before you trust me on this.
268
269
270Hardware handshaking issues.
271============================
272
273The driver can be compiled in two different ways. The default
274("Specialix DTR/RTS pin is RTS" is off) the pin behaves as DTR when
275hardware handshaking is off. It behaves as the RTS hardware
276handshaking signal when hardware handshaking is selected.
277
278When you use this, you have to use the appropriate cable. The
279cable will either be compatible with hardware handshaking or with
280software handshaking. So switching on the fly is not really an
281option.
282
283I actually prefer to use the "Specialix DTR/RTS pin is RTS" option.
284This makes the DTR/RTS pin always an RTS pin, and ioctls to
285change DTR are always ignored. I have a cable that is configured
286for this.
287
288
289Ports and devices
290=================
291
292Port 0 is the one furthest from the card-edge connector.
293
294Devices:
295
296You should make the devices as follows:
297
298bash
299cd /dev
300for i in  0  1  2  3  4  5  6  7  8  9 10 11 12 13 14 15 \
301         16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31
302do
303  echo -n "$i "
304  mknod /dev/ttyW$i c 75 $i
305  mknod /dev/cuw$i c 76 $i
306done
307echo ""
308
309If your system doesn't come with these devices preinstalled, bug your
310linux-vendor about this. They have had ample time to get this
311implemented by now.
312
313You cannot have more than 4 boards in one computer. The card only
314supports 4 different interrupts. If you really want this, contact me
315about this and I'll give you a few tips (requires soldering iron)....
316
317If you have enough PCI slots, you can probably use more than 4 PCI
318versions of the card though....
319
320The PCI version of the card cannot adhere to the mechanical part of
321the PCI spec because the 8 serial connectors are simply too large. If
322it doesn't fit in your computer, bring back the card.
323
324
325------------------------------------------------------------------------
326
327
328  Fixed bugs and restrictions:
329       - During intialization, interrupts are blindly turned on.
330            Having a shadow variable would cause an extra memory
331            access on every IO instruction.
332       - The interrupt (on the card) should be disabled when we
333         don't allocate the Linux end of the interrupt. This allows
334         a different driver/card to use it while all ports are not in
335         use..... (a la standard serial port)
336       == An extra _off variant of the sx_in and sx_out macros are
337          now available. They don't set the interrupt enable bit.
338          These are used during initialization. Normal operation uses
339          the old variant which enables the interrupt line.
340       - RTS/DTR issue needs to be implemented according to
341         specialix' spec.
342            I kind of like the "determinism" of the current
343            implementation. Compile time flag?
344       == Ok. Compile time flag! Default is how Specialix likes it.
345       == Now a config time flag! Gets saved in your config file. Neat!
346       - Can you set the IO address from the lilo command line?
347            If you need this, bug me about it, I'll make it.
348       == Hah! No bugging needed. Fixed! :-)
349       - Cirrus logic hasn't gotten back to me yet why the CD1865 can
350            and the CD1864 can't do 115k2. I suspect that this is
351            because the CD1864 is not rated for 33MHz operation.
352            Therefore the CD1864 versions of the card can't do 115k2 on
353            all ports just like the CD1865 versions. The driver does
354            not block 115k2 on CD1864 cards.
355        == I called the Cirrus Logic representative here in Holland.
356           The CD1864 databook is identical to the CD1865 databook,
357           except for an extra warning at the end. Similar Bit errors
358           have been observed in testing at 115k2 on both an 1865 and
359           a 1864 chip. I see no reason why I would prohibit 115k2 on
360           1864 chips and not do it on 1865 chips. Actually there is
361           reason to prohibit it on BOTH chips. I print a warning.
362           If you use 115k2, you're on your own.
363       - A spiky CD may send spurious HUPs. Also in CLOCAL???
364         -- A fix for this turned out to be counter productive.
365            Different fix? Current behaviour is acceptable?
366         -- Maybe the current implementation is correct. If anybody
367            gets bitten by this, please report, and it will get fixed.
368
369         -- Testing revealed that when in CLOCAL, the problem doesn't
370            occur. As warned for in the CD1865 manual, the chip may
371            send modem intr's on a spike. We could filter those out,
372            but that would be a cludge anyway (You'd still risk getting
373            a spurious HUP when two spikes occur.).....
374
375
376
377  Bugs & restrictions:
378       - This is a difficult card to autoprobe.
379            You have to WRITE to the address register to even
380            read-probe a CD186x register. Disable autodetection?
381         -- Specialix: any suggestions?
382       - Arbitrary baud rates are not implemented yet.
383            If you need this, bug me about it.
384
385
386