1.\"	$OpenBSD: dc.4,v 1.34 2004/03/21 19:50:44 miod Exp $
2.\"
3.\" Copyright (c) 1997, 1998, 1999
4.\"	Bill Paul <wpaul@ee.columbia.edu>. All rights reserved.
5.\"
6.\" Redistribution and use in source and binary forms, with or without
7.\" modification, are permitted provided that the following conditions
8.\" are met:
9.\" 1. Redistributions of source code must retain the above copyright
10.\"    notice, this list of conditions and the following disclaimer.
11.\" 2. Redistributions in binary form must reproduce the above copyright
12.\"    notice, this list of conditions and the following disclaimer in the
13.\"    documentation and/or other materials provided with the distribution.
14.\" 3. All advertising materials mentioning features or use of this software
15.\"    must display the following acknowledgement:
16.\"	This product includes software developed by Bill Paul.
17.\" 4. Neither the name of the author nor the names of any co-contributors
18.\"    may be used to endorse or promote products derived from this software
19.\"   without specific prior written permission.
20.\"
21.\" THIS SOFTWARE IS PROVIDED BY Bill Paul AND CONTRIBUTORS ``AS IS'' AND
22.\" ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
23.\" IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
24.\" ARE DISCLAIMED.  IN NO EVENT SHALL Bill Paul OR THE VOICES IN HIS HEAD
25.\" BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
26.\" CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
27.\" SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
28.\" INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
29.\" CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
30.\" ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF
31.\" THE POSSIBILITY OF SUCH DAMAGE.
32.\"
33.\" $FreeBSD: src/share/man/man4/dc.4,v 1.1 1999/12/04 17:41:24 wpaul Exp $
34.\"
35.Dd November 20, 1999
36.Dt DC 4
37.Os
38.Sh NAME
39.Nm dc
40.Nd DEC/Intel 21140, 21142, 21143, 21145 and clones 10/100 Ethernet driver
41.Sh SYNOPSIS
42.Cd "dc* at pci? dev ? function ?"
43.Cd "dc* at cardbus? dev ? function ?"
44.Cd "dcphy* at mii? phy ?"
45.Cd "nsphy* at mii? phy ?"
46.Cd "lxtphy* at mii? phy ?"
47.Sh DESCRIPTION
48The
49.Nm
50driver provides support for several PCI, MiniPCI, and CardBus Fast Ethernet
51adapters and embedded controllers based on the following chipsets:
52.Pp
53.Bl -bullet -compact -offset indent
54.It
55DEC 21140 PCI
56.It
57DEC/Intel 21143 PCI and CardBus
58.It
59Intel 21145 PCI
60.It
61Macronix 98713, 98713A, 98715, 98715A, 98725, 98727, and 98732
62.It
63Davicom DM9100, DM9102, and DM9102A
64.It
65ASIX Electronics AX88140A and AX88141
66.It
67ADMtek AL981 Comet PCI and AN983 Centaur-P PCI
68.It
69ADMtek AN985 Centaur-C CardBus
70.It
71Lite-On 82c168 and 82c169 PNIC
72.It
73Lite-On/Macronix 82c115 PNIC II
74.It
75Xircom X3201-based CardBus
76.El
77.Pp
78All of these chips have the same general register layout, DMA
79descriptor format and method of operation.
80All of the clone chips are based on the 21143 design with
81various modifications.
82(The 21140 is an older version of the 21143.)
83The 21143 itself has support for 10baseT, BNC, AUI, MII and symbol
84media attachments, 10 and 100Mbps speeds in full or half duplex,
85built-in NWAY autonegotiation and wake on LAN.
86The 21143 also offers several receive filter programming options including
87perfect filtering, inverse perfect filtering and hash table filtering.
88The 21145 seems to be 10 MBit/s only and has an additional (unsupported)
89HomePNA PHY.
90.Pp
91Some clone chips duplicate the 21143 fairly closely while others
92only maintain superficial similarities.
93Some support only MII media attachments.
94Others use different receiver filter programming mechanisms.
95At least one supports only chained DMA descriptors
96(most support both chained descriptors and contiguously allocated
97fixed size rings).
98Some chips (especially the PNIC) also have peculiar bugs.
99The
100.Nm
101driver does its best to provide generalized support for all
102of these chipsets in order to keep special case code to a minimum.
103.Pp
104These chips are used by many vendors, which makes it
105difficult to provide a complete list of all supported cards.
106The following NICs are known to work with the
107.Nm
108driver at this time:
109.Pp
110.Bl -bullet -compact -offset indent
111.It
112Digital DE500-BA 10/100 (21143, non-MII)
113.It
114Built-in DE500-BA on DEC Alpha workstations (21143, non-MII)
115.It
116Built-in Ethernet on LinkSys EtherFast 10/100 Instant GigaDrive (DM9102, MII)
117.It
118Kingston KNE100TX (21143, MII)
119.It
120D-Link DFE-570TX (21143, MII, quad port)
121.It
122NDC SOHOware SFA110A (98713A)
123.It
124NDC SOHOware SFA110A Rev B4 (98715AEC-C)
125.It
126SVEC PN102-TX (98713)
127.It
128CNet Pro120A (98715A or 98713A) and CNet Pro120B (98715)
129.It
130Compex RL100-TX (98713 or 98713A)
131.It
132LinkSys LNE100TX (PNIC 82c168, 82c169)
133.It
134NetGear FA310-TX Rev. D1, D2 or D3 (PNIC 82c169)
135.It
136Matrox FastNIC 10/100 (PNIC 82c168, 82c169)
137.It
138Kingston KNE110TX (PNIC 82c169)
139.It
140LinkSys LNE100TX v2.0 (PNIC II 82c115)
141.It
142Jaton XpressNet (Davicom DM9102)
143.It
144Alfa Inc GFC2204 (ASIX AX88140A)
145.It
146CNet Pro110B (ASIX AX88140A)
147.It
148LinkSys LNE100TX v4.x (ADMtek AN983 Centaur-P)
149.It
150Xircom CardBus, including RealPort models (Xircom X3201)
151.It
152IBM EtherJet 10/100 CardBus (Intel 21143)
153.It
154Accton EN1217 (98715) and EN2242 (ADMtek Centaur)
155.It
156Mototech ME316 (ADMtek Centaur)
157.It
158Conexant LANfinity RS7112 MiniPCI
159.El
160.Pp
161The
162.Nm
163driver supports the following media types:
164.Bl -tag -width full-duplex
165.It autoselect
166Enable autoselection of the media type and options.
167The user can manually override
168the autoselected mode by adding media options to the
169.Xr hostname.if 5
170file.
171.Pp
172Note: the built-in NWAY autonegotiation on the original PNIC 82c168
173chip is horribly broken and is not supported by the
174.Nm
175driver at this time: the chip will operate in any speed or duplex
176mode, however these must be set manually.
177The original 82c168 appears on very early revisions of the LinkSys LNE100TX
178and Matrox FastNIC.
179.It 10baseT
180Set 10Mbps operation.
181The
182.Ar mediaopt
183option can also be used to enable
184.Ar full-duplex
185operation.
186Not specifying
187.Ar full duplex
188implies
189.Ar half-duplex
190mode.
191.It 100baseTX
192Set 100Mbps (Fast Ethernet) operation.
193The
194.Ar mediaopt
195option can also be used to enable
196.Ar full-duplex
197operation.
198Not specifying
199.Ar full duplex
200implies
201.Ar half-duplex
202mode.
203.El
204.Pp
205The
206.Nm
207driver supports the following media options:
208.Bl -tag -width full-duplex
209.It full-duplex
210Force full duplex operation.
211The interface will operate in half duplex mode if this media option
212is not specified.
213.El
214.Pp
215Note that the 100baseTX media type may not be available on certain
216Intel 21143 adapters which support 10Mbps media attachments only.
217The Intel 21145 supports 10 MBit/s half-duplex only.
218For more information on configuring this device, see
219.Xr ifconfig 8 .
220.Sh DIAGNOSTICS
221.Bl -diag
222.It "dc%d: couldn't map ports/memory"
223A fatal initialization error has occurred.
224.It "dc%d: couldn't map interrupt"
225A fatal initialization error has occurred.
226.It "dc%d: watchdog timeout"
227A packet was queued for transmission and a transmit command was
228issued, however the device failed to acknowledge the transmission
229before a timeout expired.
230This can happen if the device is unable to deliver interrupts for some
231reason, or if there is a problem with the network connection (cable).
232.It "dc%d: no memory for rx list"
233The driver failed to allocate an mbuf for the receiver ring.
234.It "dc%d: TX underrun -- increasing TX threshold"
235The device generated a transmit underrun error while attempting to
236DMA and transmit a packet.
237This happens if the host is not able to DMA the packet data into the NIC's
238FIFO fast enough.
239The driver will dynamically increase the transmit start threshold so that
240more data must be DMAed into the FIFO before the NIC will start
241transmitting it onto the wire.
242.It "dc%d: TX underrun -- using store and forward mode"
243The device continued to generate transmit underruns even after all
244possible transmit start threshold settings had been tried, so the
245driver programmed the chip for store and forward mode.
246In this mode, the NIC will not begin transmission until the entire packet
247has been transferred into its FIFO memory.
248.It "dc%d: chip is in D3 power state -- setting to D0"
249This message applies only to adapters which support power management.
250Some operating systems place the controller in low power
251mode when shutting down, and some PCI BIOSes fail to bring the chip
252out of this state before configuring it.
253The controller loses all of its PCI configuration in the D3 state,
254so if the BIOS does not set it back to full power mode in time,
255it won't be able to configure it correctly.
256The driver tries to detect this condition and bring
257the adapter back to the D0 (full power) state, but this may not be
258enough to return the driver to a fully operational condition.
259If you see this message at boot time and the driver fails to attach
260the device as a network interface, you will have to perform a second
261warm boot to have the device properly configured.
262.Pp
263Note that this condition only occurs when warm booting from another
264operating system.
265If you power down your system prior to booting
266.Ox ,
267the card should be configured correctly.
268.El
269.Sh SEE ALSO
270.Xr arp 4 ,
271.Xr cardbus 4 ,
272.Xr dcphy 4 ,
273.Xr ifmedia 4 ,
274.Xr intro 4 ,
275.Xr lxtphy 4 ,
276.Xr netintro 4 ,
277.Xr nsphy 4 ,
278.Xr pci 4 ,
279.Xr hostname.if 5 ,
280.Xr ifconfig 8
281.Rs
282.%T ADMtek AL981 and AL983 data sheets
283.%O http://www.admtek.com.tw
284.Re
285.Rs
286.%T ASIX Electronics AX88140A and AX88141 data sheets
287.%O http://www.asix.com.tw
288.Re
289.Rs
290.%T Davicom DM9102 data sheet
291.%O http://www.davicom8.com
292.Re
293.Rs
294.%T Intel 21143 Hardware Reference Manual
295.%O http://developer.intel.com
296.Re
297.Rs
298.%T Macronix 98713/A, 98715/A and 98725 data sheets
299.%O http://www.macronix.com
300.Re
301.Rs
302.%T Macronix 98713/A and 98715/A app notes
303.%O http://www.macronix.com
304.Re
305.Sh HISTORY
306The
307.Nm
308device driver first appeared in
309.Fx 4.0 .
310.Ox
311support was added in
312.Ox 2.7 .
313.Sh AUTHORS
314The
315.Nm
316driver was written by
317.An Bill Paul Aq wpaul@ee.columbia.edu
318and ported to
319.Ox
320by
321.An Aaron Campbell Aq aaron@openbsd.org .
322.Sh BUGS
323The Macronix application notes claim that in order to put the
324chips in normal operation, the driver must write a certain magic
325number into the CSR16 register.
326The numbers are documented in the app notes, but the exact meaning of the
327bits is not.
328.Pp
329The 98713A seems to have a problem with 10Mbps full duplex mode.
330The transmitter works but the receiver tends to produce many
331unexplained errors leading to very poor overall performance.
332The 98715A does not exhibit this problem.
333All other modes on the 98713A seem to work correctly.
334.Pp
335The original 82c168 PNIC chip has built-in NWAY support which is
336used on certain early LinkSys LNE100TX and Matrox FastNIC cards,
337however it is horribly broken and difficult to use reliably.
338Consequently, autonegotiation is not currently supported for this
339chipset: the driver defaults the NIC to 10baseT half duplex, and it's
340up to the operator to manually select a different mode if necessary.
341(Later cards use an external MII transceiver to implement NWAY
342autonegotiation and work correctly.)
343.Pp
344The
345.Nm
346driver programs 82c168 and 82c169 PNIC chips to use the store and
347forward setting for the transmit start threshold by default.
348This is to work around problems with some NIC/PCI bus combinations where
349the PNIC can transmit corrupt frames when operating at 100Mbps,
350probably due to PCI DMA burst transfer errors.
351.Pp
352The 82c168 and 82c169 PNIC chips also have a receiver bug that
353sometimes manifests during periods of heavy receive and transmit
354activity, where the chip will improperly DMA received frames to
355the host.
356The chips appear to upload several kilobytes of garbage
357data along with the received frame data, dirtying several RX buffers
358instead of just the expected one.
359The
360.Nm
361driver detects this condition and will salvage the frame, however
362it incurs a serious performance penalty in the process.
363.Pp
364The PNIC chips also sometimes generate a transmit underrun error when
365the driver attempts to download the receiver filter setup frame, which
366can result in the receive filter being incorrectly programmed.
367The
368.Nm
369driver will watch for this condition and requeue the setup frame until
370it is transferred successfully.
371.Pp
372The ADMtek AL981 chip (and possibly the AN983 as well) has been observed
373to sometimes wedge on transmit: this appears to happen when the driver
374queues a sequence of frames which cause it to wrap from the end of
375the transmit descriptor ring back to the beginning.
376The
377.Nm
378driver attempts to avoid this condition by not queuing any frames past
379the end of the transmit ring during a single invocation of the
380.Fn dc_start
381routine.
382This workaround has a negligible impact on transmit performance.
383.Pp
384The
385.Fn mii_tick
386function does not currently run for ASIX boards, meaning cable disconnects
387and reconnects can go unnoticed.
388The AX88140A and AX88141 data sheets indicate that they don't have RX or TX
389state registers (the bits are reserved).
390Therefore, we can't seem to reliably detect when the adapter is idle.
391.Pp
392The Davicom interfaces require a grossly high PCI latency timer value to
393function properly.
394This means when a Davicom adapter is present in the machine, it is given
395an unfairly high amount of bandwidth on the PCI bus, unnecessarily taking
396time away from other devices.
397Therefore, Davicom network cards are not recommended for use with
398.Ox .
399Be careful; some motherboards have Davicom interfaces built-in.
400