| blob | 6eb6f3a3331c2656b69156b3c308308500b58d98 |
2 specialix.txt -- specialix IO8+ multiport serial driver readme.
6 Copyright (C) 1997 Roger Wolff (R.E.Wolff@BitWizard.nl)
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.
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.
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)
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.
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.
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.
41 Intro
42 =====
45 This file contains some random information, that I like to have online
46 instead of in a manual that can get lost. Ever misplace your Linux
47 kernel sources? And the manual of one of the boards in your computer?
50 Addresses and interrupts
51 ========================
53 Address dip switch settings:
54 The dip switch sets bits 2-9 of the IO address.
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 ---->
63 | | |
64 edge connector
65 | | |
66 V V V
68 Base address 0x100 caused a conflict in one of my computers once. I
69 haven't the foggiest why. My Specialix card is now at 0x180. My
70 other computer runs just fine with the Specialix card at 0x100....
71 The card occupies 4 addresses, but actually only two are really used.
73 The PCI version doesn't have any dip switches. The BIOS assigns
74 an IO address.
76 The driver now still autoprobes at 0x100, 0x180, 0x250 and 0x260. If
77 that causes trouble for you, please report that. I'll remove
78 autoprobing then.
80 The driver will tell the card what IRQ to use, so you don't have to
81 change any jumpers to change the IRQ. Just use a command line
82 argument (irq=xx) to the insmod program to set the interrupt.
84 The BIOS assigns the IRQ on the PCI version. You have no say in what
85 IRQ to use in that case.
87 If your specialix cards are not at the default locations, you can use
88 the kernel command line argument "specialix=io0,irq0,io1,irq1...".
89 Here "io0" is the io address for the first card, and "irq0" is the
90 irq line that the first card should use. And so on.
92 Examples.
94 You use the driver as a module and have three cards at 0x100, 0x250
95 and 0x180. And some way or another you want them detected in that
96 order. Moreover irq 12 is taken (e.g. by your PS/2 mouse).
98 insmod specialix.o iobase=0x100,0x250,0x180 irq=9,11,15
100 The same three cards, but now in the kernel would require you to
101 add
103 specialix=0x100,9,0x250,11,0x180,15
105 to the command line. This would become
107 append="specialix=0x100,9,0x250,11,0x180,15"
109 in your /etc/lilo.conf file if you use lilo.
111 The Specialix driver is slightly odd: It allows you to have the second
112 or third card detected without having a first card. This has
113 advantages and disadvantages. A slot that isn't filled by an ISA card,
114 might be filled if a PCI card is detected. Thus if you have an ISA
115 card at 0x250 and a PCI card, you would get:
117 sx0: specialix IO8+ Board at 0x100 not found.
118 sx1: specialix IO8+ Board at 0x180 not found.
119 sx2: specialix IO8+ board detected at 0x250, IRQ 12, CD1865 Rev. B.
120 sx3: specialix IO8+ Board at 0x260 not found.
121 sx0: specialix IO8+ board detected at 0xd800, IRQ 9, CD1865 Rev. B.
123 This would happen if you don't give any probe hints to the driver.
124 If you would specify:
126 specialix=0x250,11
128 you'd get the following messages:
130 sx0: specialix IO8+ board detected at 0x250, IRQ 11, CD1865 Rev. B.
131 sx1: specialix IO8+ board detected at 0xd800, IRQ 9, CD1865 Rev. B.
133 ISA probing is aborted after the IO address you gave is exhausted, and
134 the PCI card is now detected as the second card. The ISA card is now
135 also forced to IRQ11....
138 Baud rates
139 ==========
141 The rev 1.2 and below boards use a CL-CD1864. These chips can only
142 do 64kbit. The rev 1.3 and newer boards use a CL-CD1865. These chips
143 are officially capable of 115k2.
145 The Specialix card uses a 25MHz crystal (in times two mode, which in
146 fact is a divided by two mode). This is not enough to reach the rated
147 115k2 on all ports at the same time. With this clock rate you can only
148 do 37% of this rate. This means that at 115k2 on all ports you are
149 going to lose characters (The chip cannot handle that many incoming
150 bits at this clock rate.) (Yes, you read that correctly: there is a
151 limit to the number of -=bits=- per second that the chip can handle.)
153 If you near the "limit" you will first start to see a graceful
154 degradation in that the chip cannot keep the transmitter busy at all
155 times. However with a central clock this slow, you can also get it to
156 miss incoming characters. The driver will print a warning message when
157 you are outside the official specs. The messages usually show up in
158 the file /var/log/messages .
160 The specialix card cannot reliably do 115k2. If you use it, you have
161 to do "extensive testing" (*) to verify if it actually works.
163 When "mgetty" communicates with my modem at 115k2 it reports:
164 got: +++[0d]ATQ0V1H0[0d][0d][8a]O[cb][0d][8a]
165 ^^^^ ^^^^ ^^^^
167 The three characters that have the "^^^" under them have suffered a
168 bit error in the highest bit. In conclusion: I've tested it, and found
169 that it simply DOESN'T work for me. I also suspect that this is also
170 caused by the baud rate being just a little bit out of tune.
172 I upgraded the crystal to 66Mhz on one of my Specialix cards. Works
173 great! Contact me for details. (Voids warranty, requires a steady hand
174 and more such restrictions....)
177 (*) Cirrus logic CD1864 databook, page 40.
180 Cables for the Specialix IO8+
181 =============================
183 The pinout of the connectors on the IO8+ is:
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
195 -- 6 5 4 3 2 1 --
196 | |
197 | |
198 | |
199 | |
200 +----- -----+
201 |__________|
202 clip
204 Front view of an RJ12 connector. Cable moves "into" the paper.
205 (the plug is ready to plug into your mouth this way...)
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! :-)
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
227 Same NULL cable, but now sorted on the second column.
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
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
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.
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
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.
270 Hardware handshaking issues.
271 ============================
273 The driver can be told to operate in two different ways. The default
274 behaviour is specialix.sx_rtscts = 0 where the pin behaves as DTR when
275 hardware handshaking is off. It behaves as the RTS hardware
276 handshaking signal when hardware handshaking is selected.
278 When you use this, you have to use the appropriate cable. The
279 cable will either be compatible with hardware handshaking or with
280 software handshaking. So switching on the fly is not really an
281 option.
283 I actually prefer to use the "specialix.sx_rtscts=1" option.
284 This makes the DTR/RTS pin always an RTS pin, and ioctls to
285 change DTR are always ignored. I have a cable that is configured
286 for this.
289 Ports and devices
290 =================
292 Port 0 is the one furthest from the card-edge connector.
294 Devices:
296 You should make the devices as follows:
298 bash
299 cd /dev
300 for 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
302 do
303 echo -n "$i "
304 mknod /dev/ttyW$i c 75 $i
305 mknod /dev/cuw$i c 76 $i
306 done
307 echo ""
309 If your system doesn't come with these devices preinstalled, bug your
310 linux-vendor about this. They have had ample time to get this
311 implemented by now.
313 You cannot have more than 4 boards in one computer. The card only
314 supports 4 different interrupts. If you really want this, contact me
315 about this and I'll give you a few tips (requires soldering iron)....
317 If you have enough PCI slots, you can probably use more than 4 PCI
318 versions of the card though....
320 The PCI version of the card cannot adhere to the mechanical part of
321 the PCI spec because the 8 serial connectors are simply too large. If
322 it doesn't fit in your computer, bring back the card.
325 ------------------------------------------------------------------------
328 Fixed bugs and restrictions:
329 - During initialization, 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.
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.).....
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?