xref: /freebsd/share/man/man4/dc.4 (revision daf1cffce2e07931f27c6c6998652e90df6ba87e)
1.\" Copyright (c) 1997, 1998, 1999
2.\"	Bill Paul <wpaul@ee.columbia.edu>. All rights reserved.
3.\"
4.\" Redistribution and use in source and binary forms, with or without
5.\" modification, are permitted provided that the following conditions
6.\" are met:
7.\" 1. Redistributions of source code must retain the above copyright
8.\"    notice, this list of conditions and the following disclaimer.
9.\" 2. Redistributions in binary form must reproduce the above copyright
10.\"    notice, this list of conditions and the following disclaimer in the
11.\"    documentation and/or other materials provided with the distribution.
12.\" 3. All advertising materials mentioning features or use of this software
13.\"    must display the following acknowledgement:
14.\"	This product includes software developed by Bill Paul.
15.\" 4. Neither the name of the author nor the names of any co-contributors
16.\"    may be used to endorse or promote products derived from this software
17.\"   without specific prior written permission.
18.\"
19.\" THIS SOFTWARE IS PROVIDED BY Bill Paul AND CONTRIBUTORS ``AS IS'' AND
20.\" ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
21.\" IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
22.\" ARE DISCLAIMED.  IN NO EVENT SHALL Bill Paul OR THE VOICES IN HIS HEAD
23.\" BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
24.\" CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
25.\" SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
26.\" INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
27.\" CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
28.\" ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF
29.\" THE POSSIBILITY OF SUCH DAMAGE.
30.\"
31.\" $FreeBSD$
32.\"
33.Dd November 20, 1999
34.Dt DC 4 i386
35.Os FreeBSD
36.Sh NAME
37.Nm dc
38.Nd
39DEC/Intel 21143 and clone 10/100 ethernet driver
40.Sh SYNOPSIS
41.Cd "device miibus"
42.Cd "device dc"
43.Sh DESCRIPTION
44The
45.Nm
46driver provides support for several PCI fast ethernet adapters and
47embedded controllers based on the following chipsets:
48.Pp
49.Bl -bullet -compact -offset indent
50.It
51DEC/Intel 21143
52.It
53Macronix 98713, 98713A, 98715, 98715A and 98725
54.It
55Davicom DM9100, DM9102 and DM9102A
56.It
57ASIX Electronics AX88140A and AX88141
58.It
59ADMtek AL981 Comet and AN985 Centaur
60.It
61Lite-On 82c168 and 82c169 PNIC
62.It
63Lite-On/Macronix 82c115 PNIC II
64.El
65.Pp
66All of these chips have the same general register layout, DMA
67descriptor format and method of operation. All of the clone chips
68are based on the 21143 design with various modifications. The
6921143 itself has support for 10baseT, BNC, AUI, MII and symbol
70media attachments, 10 and 100Mbps speeds in full or half duplex,
71built in NWAY autonegotiation and wake on LAN. The 21143 also
72offers several receive filter programming options including
73perfect filtering, inverse perfect filtering and hash table
74filtering.
75.Pp
76Some clone chips duplicate the 21143 fairly closely while others
77only maintain superficial simularities. Some support only MII
78media attachments. Others use different receiver filter programming
79mechanisms. At least one supports only chained DMA descriptors
80(most support both chained descriptors and contiguously allocated
81fixed size rings). Some chips (especially the PNIC) also have
82peculiar bugs. The
83.Nm
84driver does its best to provide generalized support for all
85of these chipsets in order to keep special case code to a minimun.
86.Pp
87These chips are used by many vendors which makes it
88difficult provide a complete list of all supported cards. The
89following NICs are known to work with the
90.Nm
91driver at this time:
92.Pp
93.Bl -bullet -compact -offset indent
94.It
95Digital DE500-BA 10/100 (21143, non-MII)
96.It
97Built in DE500-BA on DEC Alpha workstations (21143, non-MII)
98.It
99Built in 10Mbps only ethernet on Compaq Presario 7900 series
100desktops (21143, non-MII)
101.It
102Built in ethernet on LinkSys EtherFast 10/100 Instant GigaDrive (DM9102, MII)
103.It
104Kingston KNE100TX (21143, MII)
105.It
106D-Link DFE-570TX (21143, MII, quad port)
107.It
108NDC SOHOware SFA110 (98713A)
109.It
110SVEC PN102-TX (98713)
111.It
112CNet Pro120A (98715A or 98713A) and CNet Pro120B (98715)
113.It
114Compex RL100-TX (98713 or 98713A)
115.It
116LinkSys LNE100TX (PNIC 82c168, 82c169)
117.It
118NetGear FA310-TX Rev. D1, D2 or D3 (PNIC 82c169)
119.It
120Matrox FastNIC 10/100 (PNIC 82c168, 82c169)
121.It
122Kingston KNE110TX (PNIC 82c169)
123.It
124LinkSys LNE100TX v2.0 (PNIC II 82c115)
125.It
126Jaton XpressNet (Davicom DM9102)
127.It
128Alfa Inc GFC2204 (ASIX AX88140A)
129.It
130CNet Pro110B (ASIX AX88140A)
131.El
132.Pp
133The
134.Nm
135driver supports the following media types:
136.Pp
137.Bl -tag -width xxxxxxxxxxxxxxxxxxxx
138.It autoselect
139Enable autoselection of the media type and options.
140The user can manually override
141the autoselected mode by adding media options to the
142.Pa /etc/rc.conf
143file.
144.Pp
145Note: the built-in NWAY autonegotiation on the original PNIC 82c168
146chip is horribly broken and is not supported by the
147.Nm
148driver at this time: the chip will operate in any speed or duplex
149mode, however these must be set manually. The original 82c168 appears
150on very early revisions of the LinkSys LNE100TX and Matrox FastNIC.
151.It 10baseT/UTP
152Set 10Mbps operation. The
153.Ar mediaopt
154option can also be used to enable
155.Ar full-duplex
156operation. Not specifying
157.Ar full duplex
158implies
159.Ar half-duplex
160mode.
161.It 100baseTX
162Set 100Mbps (fast ethernet) operation. The
163.Ar mediaopt
164option can also be used to enable
165.Ar full-duplex
166operation. Not specifying
167.Ar full duplex
168implies
169.Ar half-duplex
170mode.
171.El
172.Pp
173The
174.Nm
175driver supports the following media options:
176.Pp
177.Bl -tag -width xxxxxxxxxxxxxxxxxxxx
178.It full-duplex
179Force full duplex operation. The interface will operate in
180half duplex mode if this media option is not specified.
181.El
182.Pp
183Note that the 100baseTX media type may not be available on certain
184Intel 21143 adapters which support 10mbps media attachments only.
185For more information on configuring this device, see
186.Xr ifconfig 8 .
187.Sh DIAGNOSTICS
188.Bl -diag
189.It "dc%d: couldn't map ports/memory"
190A fatal initialization error has occurred.
191.It "dc%d: couldn't map interrupt"
192A fatal initialization error has occurred.
193.It "dc%d: watchdog timeout"
194A packet was queued for transmission and a transmit command was
195issued, however the device failed to acknowledge the transmission
196before a timeout expired. This can happen if the device is unable
197to deliver interrupts for some reason, of if there is a problem with
198the network connection (cable).
199.It "dc%d: no memory for rx list"
200The driver failed to allocate an mbuf for the receiver ring.
201.It "dc%d: TX underrun -- increasing TX threshold"
202The device generated a transmit underrun error while attempting to
203DMA and transmit a packet. This happens if the host is not able to
204DMA the packet data into the NIC's FIFO fast enough. The driver
205will dynamically increase the transmit start threshold so that
206more data must be DMAed into the FIFO before the NIC will start
207transmitting it onto the wire.
208.It "dc%d: TX underrun -- using store and forward mode"
209The device continued to generate transmit underruns even after all
210possible transmit start threshold settings had been tried, so the
211driver programmed the chip for store and forward mode. In this mode,
212the NIC will not begin transmission until the entire packet has been
213transfered into its FIFO memory.
214.It "dc%d: chip is in D3 power state -- setting to D0"
215This message applies only to adapters which support power
216management. Some operating systems place the controller in low power
217mode when shutting down, and some PCI BIOSes fail to bring the chip
218out of this state before configuring it. The controller loses all of
219its PCI configuration in the D3 state, so if the BIOS does not set
220it back to full power mode in time, it won't be able to configure it
221correctly. The driver tries to detect this condition and bring
222the adapter back to the D0 (full power) state, but this may not be
223enough to return the driver to a fully operational condition. If
224you see this message at boot time and the driver fails to attach
225the device as a network interface, you will have to perform second
226warm boot to have the device properly configured.
227.Pp
228Note that this condition only occurs when warm booting from another
229operating system. If you power down your system prior to booting
230.Fx ,
231the card should be configured correctly.
232.El
233.Sh SEE ALSO
234.Xr arp 4 ,
235.Xr netintro 4 ,
236.Xr ifconfig 8 ,
237.Xr ng_ether 8
238.Rs
239.%T ADMtek AL981, AL983 and AL985 data sheets
240.%O http://www.admtek.com.tw
241.Re
242.Rs
243.%T ASIX Electronics AX88140A and AX88141 data sheets
244.%O http://www.asix.com.tw
245.Re
246.Rs
247.%T Davicom DM9102 data sheet
248.%O http://www.davicom8.com
249.Re
250.Rs
251.%T Intel 21143 Hardware Reference Manual
252.%O http://developer.intel.com
253.Re
254.Rs
255.%T Macronix 98713/A, 98715/A and 98725 data sheets
256.%O http://www.macronix.com
257.Re
258.Rs
259.%T Macronix 98713/A and 98715/A app notes
260.%O http://www.macronix.com
261.Re
262.Sh HISTORY
263The
264.Nm
265device driver first appeared in
266.Fx 4.0 .
267.Sh AUTHORS
268The
269.Nm
270driver was written by
271.An Bill Paul Aq wpaul@ee.columbia.edu .
272.Sh BUGS
273The Macronix application notes claim that in order to put the
274chips in normal operation, the driver must write a certian magic
275number into the CSR16 register. The numbers are documented in
276the app notes, but the exact meaning of the bits is not.
277.Pp
278The 98713A seems to have a problem with 10Mbps full duplex mode.
279The transmitter works but the receiver tends to produce many
280unexplained errors leading to very poor overall performance. The
28198715A does not exhibit this problem. All other modes on the
28298713A seem to work correctly.
283.Pp
284The original 82c168 PNIC chip has built in NWAY support which is
285used on certain early LinkSys LNE100TX and Matrox FastNIC cards,
286however it is horribly broken and difficult to use reliably.
287Consequently, autonegotiation is not currently supported for this
288chipset: the driver defaults the NIC to 10baseT half duplex, and it's
289up to the operator to manually select a different mode if necessary.
290(Later cards use an external MII transceiver to implement NWAY
291autonegotiation and work correctly.)
292.Pp
293The
294.Nm
295driver programs 82c168 and 82c169 PNIC chips to use the store and
296forward setting for the transmit start threshold by default. This
297is to work around problems with some NIC/PCI bus combinations where
298the PNIC can transmit corrupt frames when operating at 100Mbps,
299probably due to PCI DMA burst transfer errors.
300.Pp
301The 82c168 and 82c169 PNIC chips also have a receiver bug that
302sometimes manifests during periods of heavy receive and transmit
303activity, where the chip will improperly DMA received frames to
304the host. The chips appear to upload several kilobytes of garbage
305data along with the received frame data, dirtying several RX buffers
306instead of just the expected one. The
307.Nm
308driver detects this condition and will salvage the frame, however
309it incurs a serious performance penalty in the process.
310.Pp
311The PNIC chips also sometimes generate a transmit underrun error when
312the driver attempts to download the receiver filter setup frame, which
313can result in the receive filter being incorrectly programmed. The
314.Nm
315driver will watch for this condition and requeue the setup frame until
316it is transfered successfully.
317.Pp
318The ADMtek AL981 chip (and possibly the AN985 as well) has been observed
319to sometimes wedge on transmit: this appears to happen when the driver
320queues a sequence of frames which cause it to wrap from the end of the
321the transmit descriptor ring back to the beginning. The
322.Nm
323driver attempts to avoid this condition by not queing any frames past
324the end of the transmit ring during a single invocation of the
325.Fn dc_start
326routine. This workaround has a negligible impact on transmit performance.
327
328
329