"'s version, which also has the problem.
>
> This is quite an old version (almost a year ago!), try getting the 3c59x
> driver from 2.4.3.
I am now using this one and am having exactly the same problem. It has
followed me through kernel reveisions quite nicely.
> Is your system SMP ? If so, you can try booting with the "noapic" option.
Yes, SMP, good suggestion on the no apic; will try that.
> For how much time the interface is functioning ?
Anywhere from days to weeks. The most is about two weeks. This
suggested to me that it was somehow tied to the number of Rx'd patckets.
> Are there any error messages in the logs around the time when the
> interface stops functioning ?
It will regularly complain about:
kernel: eth1: Too much work in interrupt, status e401.
But the rate of these complaints does not suddenly increase just before
it drops out.
The rate is often like this:
Apr 10 19:36:35 host kernel: eth1: Too much work in interrupt, status e401.
Apr 10 19:37:06 host last message repeated 5 times
Apr 10 19:38:30 host last message repeated 5 times
Apr 10 19:38:40 host last message repeated 2 times
> Can you change the interface connected to the Cisco port ? You can then
> isolate the problem as being with Cisco or with the 3Com card/driver.
Yes, I have done this, and it stays with the 3Com.
> One other very approximate method is to watch the "Activity" LED on the
> card when you are supposed to be receiving packets from the switch; if
> it's not flashing, there is no packet arriving to the card...
"The lights are on, but nobody is home" -- the Cisco thinks the link is
up, the lights on the Cisco continue to blink, but no traffic is
observed on the 3com side.
Thank you for the ideas and suggestions; I'll run some further
diagnostics.
Regards,
Scott
From bogdan.costescu@iwr.uni-heidelberg.de Wed, 18 Apr 2001 14:39:05 +0200 (CEST)
Date: Wed, 18 Apr 2001 14:39:05 +0200 (CEST)
From: Bogdan Costescu bogdan.costescu@iwr.uni-heidelberg.de
Subject: [vortex] 3c59x "just stops"
On Wed, 18 Apr 2001, Scott A. McIntyre wrote:
> Yes, SMP, good suggestion on the no apic; will try that.
AFAIK, only 2.4 has (or had) the problems with lost interrupts on SMP
systems. If 2.2 had the same problems, there might be something else. The
suggestion with "noapic" applies to 2.2 as well (if you still want to try
2.2).
> Anywhere from days to weeks. The most is about two weeks. This
> suggested to me that it was somehow tied to the number of Rx'd patckets.
After reading the rest of the message, my guess is that it's not related
to number of Rx packets, but to pps (packets per second).
> It will regularly complain about:
> kernel: eth1: Too much work in interrupt, status e401.
That normally means that you receive very many packets in a short time...
The problem migth be like this: the driver is exceeding number of
iterations through the interrupt routine and issues this message, then
disables the interrupt and waits for a timer to wake it. Then something
wrong is happening with this timer (similar to other SMP problems with
lost interrupts), so that the driver's interrupt is never re-enabled. Then
the card fills up the Rx ring and goes in an UpStall state (where no
packet is received) - I guess this explains why the card is no longer
flashing the LED.
Depending on what you do with the computer, it might worth increasing the
"max_interrupt_work" value, which determines the maximum number of
iterations through the interrupt routine, to try to eliminate the
exceeding situation. However, this means that the CPU might spend more
time inside the interrupt routine of this driver, processing packets; IOW,
you give more "priority" to receiving packets through this interface than
to other activities on the computer. If you have a fast enough CPU and
it's not loaded by other activities (including traffic on other NICs),
increasing "max_interrupt_work" to 64 or 100 might give good results.
You can modify this value by either modifying and recompiling the driver
or by passing it as a module option.
Sincerely,
Bogdan Costescu
IWR - Interdisziplinaeres Zentrum fuer Wissenschaftliches Rechnen
Universitaet Heidelberg, INF 368, D-69120 Heidelberg, GERMANY
Telephone: +49 6221 54 8869, Telefax: +49 6221 54 8868
E-mail: Bogdan.Costescu@IWR.Uni-Heidelberg.De
From becker@scyld.com Wed, 18 Apr 2001 13:31:45 -0400 (EDT)
Date: Wed, 18 Apr 2001 13:31:45 -0400 (EDT)
From: Donald Becker becker@scyld.com
Subject: [vortex] 3c59x "just stops"
On Wed, 18 Apr 2001, Bogdan Costescu wrote:
> On Wed, 18 Apr 2001, Scott A. McIntyre wrote:
>
> > Yes, SMP, good suggestion on the no apic; will try that.
>
> AFAIK, only 2.4 has (or had) the problems with lost interrupts on SMP
> systems. If 2.2 had the same problems, there might be something else. The
> suggestion with "noapic" applies to 2.2 as well (if you still want to try
> 2.2).
The 2.2 kernels also have problems with the APIC ceasing delivery of
interrupts. Shutting down and restarting the drivers using the IRQ
would clear the problem, making it appears as a driver bug.
Several of the drivers have checks or low-rate polling to alert the user
to the problem.
We use "noapic" with the Scyld Beowulf distribution to avoid this
problem.
> > Anywhere from days to weeks. The most is about two weeks. This
> > suggested to me that it was somehow tied to the number of Rx'd patckets.
This is consistent with the APIC problem.
Donald Becker becker@scyld.com
Scyld Computing Corporation http://www.scyld.com
410 Severn Ave. Suite 210 Second Generation Beowulf Clusters
Annapolis MD 21403 410-990-9993
From dohan@WPI.EDU Wed, 18 Apr 2001 22:23:52 -0400
Date: Wed, 18 Apr 2001 22:23:52 -0400
From: Dohan dohan@WPI.EDU
Subject: [vortex] 3c900 -COMBO Media Type Problems
This is a multi-part message in MIME format.
------=_NextPart_000_0018_01C0C856.3FAFC380
Content-Type: text/plain;
charset="iso-8859-1"
Content-Transfer-Encoding: quoted-printable
I have been trying to get this to work forever now. Hopefully someone =
can help me as I am getting desperate.
I have a 3Com 3c900-COMBO ethernet card. In windows my internet speed is =
extremly slow (100 bytes/sec). To fix this I have to goto setting and =
change media type to "10mbs half-duplex"
Now in linux I get the same 100 bytes/sec so I need to change the media =
type. I can't succesfully do this. I believe there is three ways I can =
do this, and all three don't work:
1) Upgrade to the 3c90x drivers.
Problem: When I try to insmod the 3c90x I just compiled, I get =
the error init_module: Device or resource busy
2) Use the 3c59x drivers and use "insmod 3c59x.o debug=3D1 =
full_duplex=3D0 options=3D0" I have also tried using options=3D8. =
Nothing works
3) Use vortex-diag. I use "vortex-diag -F 10baseT" doesn't change, even =
if I try to set it to 10base2 which I know won't work, it still doesn't =
change.
I am looking for ANY way to change this one setting. I'm not even =
positive of what it's supposed to be. I don't really know what I'm doing =
so please be basic. Here is some other infomation about my system:
lspci
00:0d.0 Ethernet controller: 3Com Corporation 3c900 Combo [Boomerang]
Flags: bus master, medium devsel, latency 64, IRQ 10
I/O ports at 1040
ifconfig
eth0 Link encap:Ethernet HWaddr 00:A0:24:D2:FF:E1
inet addr:130.215.229.79 Bcast:130.215.239.255 =
Mask:255.255.240.0
UP BROADCAST NOTRAILERS RUNNING MTU:1500 Metric:1
RX packets:15885 errors:4 dropped:0 overruns:0 frame:7
TX packets:93 errors:0 dropped:0 overruns:0 carrier:0
collisions:0 txqueuelen:100
Interrupt:10 Base address:0x1040
lo Link encap:Local Loopback
inet addr:127.0.0.1 Mask:255.0.0.0
UP LOOPBACK RUNNING MTU:3924 Metric:1
RX packets:86 errors:0 dropped:0 overruns:0 frame:0
TX packets:86 errors:0 dropped:0 overruns:0 carrier:0
collisions:0 txqueuelen:0=20
vortex-diag.c:v2.04 1/8/2001 Donald Becker (becker@scyld.com)
http://www.scyld.com/diag/index.html
Index #1: Found a 3c900 Boomerang 10Mbps Combo adapter at 0x1040.
Parsing the EEPROM of a 3Com Vortex/Boomerang:
3Com Node Address 00:A0:24:D2:FF:E1 (used as a unique ID only).
OEM Station address 00:A0:24:D2:FF:E1 (used as the ethernet address).
Manufacture date (MM/DD/YYYY) 7/25/1996, division 6, product FP.
Options: force full-duplex.
Vortex format checksum is correct (0047 vs. 0047).
Cyclone format checksum is incorrect (00 vs. 0xff).
Hurricane format checksum is incorrect (00 vs. 0xff).
------=_NextPart_000_0018_01C0C856.3FAFC380
Content-Type: text/html;
charset="iso-8859-1"
Content-Transfer-Encoding: quoted-printable
I have been trying to get this to work =
forever now.=20
Hopefully someone can help me as I am getting desperate.
I have a 3Com 3c900-COMBO ethernet =
card. In windows=20
my internet speed is extremly slow (100 bytes/sec). To fix this I have =
to goto=20
setting and change media type to "10mbs half-duplex"
Now in linux I get the same 100 =
bytes/sec so I need=20
to change the media type. I can't succesfully do this. I believe there =
is three=20
ways I can do this, and all three don't work:
1) Upgrade to the 3c90x =
drivers.
=
Problem: When=20
I try to insmod the 3c90x I just compiled, I get the error init_module: =
Device=20
or resource busy
2) Use the 3c59x drivers and use =
"insmod 3c59x.o=20
debug=3D1 full_duplex=3D0 options=3D0" I have also tried using =
options=3D8.=20
Nothing works
3) Use vortex-diag. I use "vortex-diag =
-F=20
10baseT" doesn't change, even if I try to set it to 10base2 which =
I know=20
won't work, it still doesn't change.
I am looking for ANY way to change this =
one=20
setting. I'm not even positive of what it's supposed to be. I don't =
really know=20
what I'm doing so please be basic. Here is some other infomation about =
my=20
system:
lspci
00:0d.0 Ethernet controller: 3Com =
Corporation 3c900=20
Combo [Boomerang]
Flags: =
bus=20
master, medium devsel, latency 64, IRQ=20
10
I/O ports at =
1040
ifconfig
eth0 Link =
encap:Ethernet HWaddr=20
00:A0:24:D2:FF:E1
&nbs=
p; inet=20
addr:130.215.229.79 Bcast:130.215.239.255 =20
Mask:255.255.240.0
&nb=
sp; UP=20
BROADCAST NOTRAILERS RUNNING MTU:1500 =20
Metric:1
RX=20
packets:15885 errors:4 dropped:0 overruns:0=20
frame:7
TX =
packets:93=20
errors:0 dropped:0 overruns:0=20
carrier:0
=
collisions:0=20
txqueuelen:100
=
Interrupt:10 Base address:0x1040
lo Link=20
encap:Local =
Loopback
=20
inet addr:127.0.0.1 =20
Mask:255.0.0.0
=
UP=20
LOOPBACK RUNNING MTU:3924 =20
Metric:1
RX =
packets:86=20
errors:0 dropped:0 overruns:0=20
frame:0
TX =
packets:86=20
errors:0 dropped:0 overruns:0=20
carrier:0
=
collisions:0=20
txqueuelen:0
vortex-diag.c:v2.04 1/8/2001 Donald =
Becker (becker@scyld.com)
http://www.scyld.com/diag/i=
ndex.html
Index=20
#1: Found a 3c900 Boomerang 10Mbps Combo adapter at 0x1040.
Parsing =
the=20
EEPROM of a 3Com Vortex/Boomerang:
3Com Node Address =
00:A0:24:D2:FF:E1=20
(used as a unique ID only).
OEM Station address =
00:A0:24:D2:FF:E1 (used=20
as the ethernet address).
Manufacture date (MM/DD/YYYY) =
7/25/1996,=20
division 6, product FP.
Options: force full-duplex.
Vortex =
format=20
checksum is correct (0047 vs. 0047).
Cyclone format checksum =
is=20
incorrect (00 vs. 0xff).
Hurricane format checksum is =
incorrect (00=20
vs. 0xff).
------=_NextPart_000_0018_01C0C856.3FAFC380--
From becker@scyld.com Thu, 19 Apr 2001 02:47:23 -0400 (EDT)
Date: Thu, 19 Apr 2001 02:47:23 -0400 (EDT)
From: Donald Becker becker@scyld.com
Subject: [vortex] 3c900 -COMBO Media Type Problems
On Wed, 18 Apr 2001, Dohan wrote:
> I have a 3Com 3c900-COMBO ethernet card. In windows my internet speed
> is extremly slow (100 bytes/sec). To fix this I have to goto setting
> and change media type to "10mbs half-duplex"
>
> Now in linux I get the same 100 bytes/sec so I need to change the
> media type. I can't succesfully do this. I believe there is three ways
> I can do this, and all three don't work:
>
> 1) Upgrade to the 3c90x drivers.
> Problem: When I try to insmod the 3c90x I just compiled, I get
> the error init_module: Device or resource busy
That's because the 3Com driver only supports their recent cards. It's
much easier to write a driver that only has to support a small number of
cards.
> 2) Use the 3c59x drivers and use "insmod 3c59x.o debug=1 full_duplex=0
> options=0" I have also tried using options=8. Nothing works
Try options=2.
> 3) Use vortex-diag. I use "vortex-diag -F 10baseT" doesn't change,
> even if I try to set it to 10base2 which I know won't work, it still
> doesn't change.
The media setting code isn't including in the released vortex-diag program.
> I am looking for ANY way to change this one setting. I'm not even
> positive of what it's supposed to be. I don't really know what I'm
> doing so please be basic. Here is some other infomation about my
> system:
...
> vortex-diag.c:v2.04 1/8/2001 Donald Becker (becker@scyld.com)
> http://www.scyld.com/diag/index.html
> Index #1: Found a 3c900 Boomerang 10Mbps Combo adapter at 0x1040.
> Parsing the EEPROM of a 3Com Vortex/Boomerang:
> 3Com Node Address 00:A0:24:D2:FF:E1 (used as a unique ID only).
> OEM Station address 00:A0:24:D2:FF:E1 (used as the ethernet address).
> Manufacture date (MM/DD/YYYY) 7/25/1996, division 6, product FP.
> Options: force full-duplex.
^^^^^^^^^^^
How did this get set?
Donald Becker becker@scyld.com
Scyld Computing Corporation http://www.scyld.com
410 Severn Ave. Suite 210 Second Generation Beowulf Clusters
Annapolis MD 21403 410-990-9993
From the_h1ghlander@yahoo.com Wed, 18 Apr 2001 23:51:11 -0700 (PDT)
Date: Wed, 18 Apr 2001 23:51:11 -0700 (PDT)
From: Ben Taylor the_h1ghlander@yahoo.com
Subject: [vortex] 3Com Model numbers vs config ids?
Well, I bought a 3CXEFEM656 10/100 modem card, only to find out that
the modem is a winmodem (damnit). so I'm on the hunt for a 3com
10/100 real modem card. The reason for 3com is I'm running solaris
on my laptop and the 3com will run with win2k and SuSE 7.1.
Anyone have a definitive list of 3Com model numbers vs. the list of
"config ids from /etc/pcmcia/config.opts".
Thanks in advance,
Ben
__________________________________________________
Do You Yahoo!?
Yahoo! Auctions - buy the things you want at great prices
http://auctions.yahoo.com/
From bogdan.costescu@iwr.uni-heidelberg.de Thu, 19 Apr 2001 12:31:02 +0200 (CEST)
Date: Thu, 19 Apr 2001 12:31:02 +0200 (CEST)
From: Bogdan Costescu bogdan.costescu@iwr.uni-heidelberg.de
Subject: [vortex] 3c59x "just stops"
On Wed, 18 Apr 2001, Donald Becker wrote:
> The 2.2 kernels also have problems with the APIC ceasing delivery of
> interrupts.
Then maybe I'm just lucky! I have SMP cluster nodes and a SMP file-server
and never used "noapic" with 2.2 kernels.
> Several of the drivers have checks or low-rate polling to alert the user
> to the problem.
I guess alerting is better than nothing but still not sufficient. Would it
be possible to _do_ something in this case ? Like trying again (or more
aggressively) to enable the interrupt... however this might not work well
for shared interrupts.
[ Well, I'm not happy with the fact that the drivers have to deal with the
problem of distributing interrupts, but this exists for some time now and
from what I gathered from different mailing lists, it's not clear if/how
this should be solved. And the fact that 2.4 seems to exhibit this
problem more often than 2.2 is one of the reasons for which I keep my
clusters with 2.2...]
> We use "noapic" with the Scyld Beowulf distribution to avoid this
> problem.
If my cluster nodes would have this problem, I would do this too. That's
because the NIC generates several orders of magnitute more interrupts than
anything else in the system. But for a (relatively small) file-server with
1-2 SCSI interfaces and 2-4 (bonded) NICs, that's not really a solution...
Sincerely,
Bogdan Costescu
IWR - Interdisziplinaeres Zentrum fuer Wissenschaftliches Rechnen
Universitaet Heidelberg, INF 368, D-69120 Heidelberg, GERMANY
Telephone: +49 6221 54 8869, Telefax: +49 6221 54 8868
E-mail: Bogdan.Costescu@IWR.Uni-Heidelberg.De
From bogdan.costescu@iwr.uni-heidelberg.de Thu, 19 Apr 2001 12:34:50 +0200 (CEST)
Date: Thu, 19 Apr 2001 12:34:50 +0200 (CEST)
From: Bogdan Costescu bogdan.costescu@iwr.uni-heidelberg.de
Subject: [vortex] 3c900 -COMBO Media Type Problems
On Thu, 19 Apr 2001, Donald Becker wrote:
> > 3) Use vortex-diag. I use "vortex-diag -F 10baseT" doesn't change,
> > even if I try to set it to 10base2 which I know won't work, it still
> > doesn't change.
>
> The media setting code isn't including in the released vortex-diag program.
Why not using mii-diag ? It's more-or-less specifically designed for media
handling.
You might also have some luck by running the DOS-based setup program. My
guess is that you have the card's EEPROM set to something like "force
100MBit" or "force full-duplex".
Sincerely,
Bogdan Costescu
IWR - Interdisziplinaeres Zentrum fuer Wissenschaftliches Rechnen
Universitaet Heidelberg, INF 368, D-69120 Heidelberg, GERMANY
Telephone: +49 6221 54 8869, Telefax: +49 6221 54 8868
E-mail: Bogdan.Costescu@IWR.Uni-Heidelberg.De
From scyld.vortex@econos.de Thu, 19 Apr 2001 12:37:46 +0200
Date: Thu, 19 Apr 2001 12:37:46 +0200
From: Stefan Hoffmeister scyld.vortex@econos.de
Subject: [vortex] 905C: 3c59x unusably slow / 3c90x fine
Hi,
I am experiencing massive performance problems with the 3c59x driver
(module) shipping with the 2.4.2 kernel of Caldera OpenLinux 3.1 (beta).
The same problem seems to exist with the module delivered with RedHat 7.0
(2.2.16) and Mandrake 7.2 (2.2.17, IIRC).
"Tiny" transmits will work fine - e.g. ssh / telnet to a host on the LAN
works as expected -, but any attempt to transmit some more bits (10 MB
file, remote X session) will result in throughput of about 1-2 KB/s.
The 3c90x module delivered with SuSE 7.0 / 7.1 (and RedHat 7.0 / Mandrake
7.2) works perfectly fine, yielding throughput of about 8000 KB/s, all
other things being equal.
There do not seem to be any warnings anywhere related to this.
I am going through a 10/100 hub; the other machine is running a 2.2.19
kernel with an Intel eepro100 module.
I have pasted the boot messages from the 3c59x module below - any other
information I could provide?
How can I get the 3c59x module show acceptable performance?
TIA,
Stefan
PS: In case you wonder about that plethora of distributions - it's part of
my job.
****************
Apr 19 11:43:12 stefan-bt kernel: 3c59x.c:LK1.1.13 27 Jan 2001 Donald
Becker and others. http://www.scyld.com/network/vortex.html
Apr 19 11:43:12 stefan-bt kernel: See Documentation/networking/vortex.txt
Apr 19 11:43:12 stefan-bt kernel: eth0: 3Com PCI 3c905C Tornado at 0xb000,
PCI: Setting latency timer of device 00:0b.0 to 64
Apr 19 11:43:12 stefan-bt kernel: 00:01:02:19:5d:33, IRQ 10
Apr 19 11:43:12 stefan-bt kernel: product code 4552 rev 00.13 date
03-27-00
Apr 19 11:43:12 stefan-bt kernel: Full duplex capable
Apr 19 11:43:12 stefan-bt kernel: 8K byte-wide RAM 5:3 Rx:Tx split,
autoselect/Autonegotiate interface.
Apr 19 11:43:12 stefan-bt kernel: MII transceiver found at address 24,
status 782d.
Apr 19 11:43:12 stefan-bt kernel: Enabling bus-master transmits and
whole-frame receives.
Apr 19 11:43:12 stefan-bt kernel: eth0: scatter/gather disabled. h/w
checksums enabled
From becker@scyld.com Thu, 19 Apr 2001 09:36:48 -0400 (EDT)
Date: Thu, 19 Apr 2001 09:36:48 -0400 (EDT)
From: Donald Becker becker@scyld.com
Subject: [vortex] 3c900 -COMBO Media Type Problems
On Thu, 19 Apr 2001, Bogdan Costescu wrote:
> Date: Thu, 19 Apr 2001 12:34:50 +0200 (CEST)
> From: Bogdan Costescu
> To: vortex@scyld.com
> Cc: Dohan
> Subject: Re: [vortex] 3c900 -COMBO Media Type Problems
>
> On Thu, 19 Apr 2001, Donald Becker wrote:
>
> > > 3) Use vortex-diag. I use "vortex-diag -F 10baseT" doesn't change,
> > > even if I try to set it to 10base2 which I know won't work, it still
> > > doesn't change.
> >
> > The media setting code isn't including in the released vortex-diag program.
>
> Why not using mii-diag ? It's more-or-less specifically designed for media
> handling.
>
> You might also have some luck by running the DOS-based setup program. My
> guess is that you have the card's EEPROM set to something like "force
> 100MBit" or "force full-duplex".
My final comment was too subtle:
>> vortex-diag.c:v2.04 1/8/2001 Donald Becker (becker@scyld.com)
..
>> Index #1: Found a 3c900 Boomerang 10Mbps Combo adapter at 0x1040.
>> Parsing the EEPROM of a 3Com Vortex/Boomerang:
...
>> Options: force full-duplex.
^^^^^^^^^^^
>How did this get set?
The EEPROM has the board set to forced full duplex.
He can't use vortex-diag to turn this off because I stripped out the
media type setting code. That code required the user to know the
specifics of the transceiver connection e.g. that the 3c900 had a
on-chip "10baseT" transceiver, while the 3c905 used an external MII
transceiver. The 'B' model changed both to internal "NWay" transceiver.
Donald Becker becker@scyld.com
Scyld Computing Corporation http://www.scyld.com
410 Severn Ave. Suite 210 Second Generation Beowulf Clusters
Annapolis MD 21403 410-990-9993
From bogdan.costescu@iwr.uni-heidelberg.de Thu, 19 Apr 2001 19:16:28 +0200 (CEST)
Date: Thu, 19 Apr 2001 19:16:28 +0200 (CEST)
From: Bogdan Costescu bogdan.costescu@iwr.uni-heidelberg.de
Subject: [vortex] 905C: 3c59x unusably slow / 3c90x fine
On Thu, 19 Apr 2001, Stefan Hoffmeister wrote:
> "Tiny" transmits will work fine - e.g. ssh / telnet to a host on the LAN
> works as expected -, but any attempt to transmit some more bits (10 MB
> file, remote X session) will result in throughput of about 1-2 KB/s.
My guess is that there is a media mismatch. Please read
http://www.scyld.com/network/vortex.html
and post the results of the mii-diag tool.
However, I would have expected some messages in /var/log/messages...
> I am going through a 10/100 hub;
What kind of hub ? Is it just a repeater (hub) or switch ?
Sincerely,
Bogdan Costescu
IWR - Interdisziplinaeres Zentrum fuer Wissenschaftliches Rechnen
Universitaet Heidelberg, INF 368, D-69120 Heidelberg, GERMANY
Telephone: +49 6221 54 8869, Telefax: +49 6221 54 8868
E-mail: Bogdan.Costescu@IWR.Uni-Heidelberg.De
From andrewm@uow.edu.au Thu, 19 Apr 2001 10:18:36 -0700
Date: Thu, 19 Apr 2001 10:18:36 -0700
From: Andrew Morton andrewm@uow.edu.au
Subject: [vortex] 3c900 -COMBO Media Type Problems
Donald Becker wrote:
>
> >> Options: force full-duplex.
> ^^^^^^^^^^^
> >How did this get set?
>
> The EEPROM has the board set to forced full duplex.
3com have a DOS-based setup tool which allows
you to alter the EEPROM contents.
ftp://ftp.3com.com/pub/nic/3c90x/3c90xx2.exe
From dohan@WPI.EDU Thu, 19 Apr 2001 15:00:00 -0400
Date: Thu, 19 Apr 2001 15:00:00 -0400
From: Dohan dohan@WPI.EDU
Subject: [vortex] 3c900 -COMBO Media Type Problems
I have tried a few setting with the mii-diag program. Am I supposed to be
using the -A or the -F flag? I can't seem to change the line that says
"Speed fixed at 10 mbps, half-duplex". I am also trying to get the DOS
config working, no luck with that yet...figures :) Anyways, this is the
feedback from mii-diag. Maybe this will help.
root@linux:/home/dohan/3c90x-1.0.0 > ./mii-diag
Using the default interface 'eth0'.
Basic registers of MII PHY #0: 0000 ffff 0000 ffff 0000 ffff 0000 ffff.
Basic mode control register 0x0000: Auto-negotiation disabled, with
Speed fixed at 10 mbps, half-duplex.
Basic mode status register 0xffff ... ffff.
Link status: established.
Remote fault detected!
*** Link Jabber! ***
Your link partner advertised ffff: Flow-control 100baseT4 100baseTx-FD
100baseTx 10baseT-FD 10baseT, w/ 802.3X flow control.
> On Thu, 19 Apr 2001, Bogdan Costescu wrote:
>
> > Date: Thu, 19 Apr 2001 12:34:50 +0200 (CEST)
> > From: Bogdan Costescu
> > To: vortex@scyld.com
> > Cc: Dohan
> > Subject: Re: [vortex] 3c900 -COMBO Media Type Problems
> >
> > On Thu, 19 Apr 2001, Donald Becker wrote:
> >
> > > > 3) Use vortex-diag. I use "vortex-diag -F 10baseT" doesn't change,
> > > > even if I try to set it to 10base2 which I know won't work, it still
> > > > doesn't change.
> > >
> > > The media setting code isn't including in the released vortex-diag
program.
> >
> > Why not using mii-diag ? It's more-or-less specifically designed for
media
> > handling.
> >
> > You might also have some luck by running the DOS-based setup program. My
> > guess is that you have the card's EEPROM set to something like "force
> > 100MBit" or "force full-duplex".
>
> My final comment was too subtle:
>
> >> vortex-diag.c:v2.04 1/8/2001 Donald Becker (becker@scyld.com)
> ..
> >> Index #1: Found a 3c900 Boomerang 10Mbps Combo adapter at 0x1040.
> >> Parsing the EEPROM of a 3Com Vortex/Boomerang:
> ...
> >> Options: force full-duplex.
> ^^^^^^^^^^^
> >How did this get set?
>
> The EEPROM has the board set to forced full duplex.
>
> He can't use vortex-diag to turn this off because I stripped out the
> media type setting code. That code required the user to know the
> specifics of the transceiver connection e.g. that the 3c900 had a
> on-chip "10baseT" transceiver, while the 3c905 used an external MII
> transceiver. The 'B' model changed both to internal "NWay" transceiver.
>
> Donald Becker becker@scyld.com
> Scyld Computing Corporation http://www.scyld.com
> 410 Severn Ave. Suite 210 Second Generation Beowulf Clusters
> Annapolis MD 21403 410-990-9993
>
>
From becker@scyld.com Thu, 19 Apr 2001 15:28:09 -0400 (EDT)
Date: Thu, 19 Apr 2001 15:28:09 -0400 (EDT)
From: Donald Becker becker@scyld.com
Subject: [vortex] 3c900 -COMBO Media Type Problems
On Thu, 19 Apr 2001, Dohan wrote:
> Subject: Re: [vortex] 3c900 -COMBO Media Type Problems
>
> I have tried a few setting with the mii-diag program. Am I supposed to be
The 3c900-combo does not have an MII transceiver.
The transceiver doesn't even do autonegotiation, so I didn't write code
to emulate MII management register. It could be done (except for
forcing 10base2 or AUI), it just hasn't been written.
Your near-term solution is to remove the forced-full-duplex setting from
the EEPROM.
Donald Becker becker@scyld.com
Scyld Computing Corporation http://www.scyld.com
410 Severn Ave. Suite 210 Second Generation Beowulf Clusters
Annapolis MD 21403 410-990-9993
From Stefan.Hoffmeister@Econos.de Thu, 19 Apr 2001 22:33:51 +0200
Date: Thu, 19 Apr 2001 22:33:51 +0200
From: Stefan Hoffmeister Stefan.Hoffmeister@Econos.de
Subject: [vortex] 905C: 3c59x unusably slow / 3c90x fine
: On Thu, 19 Apr 2001 19:16:28 +0200 (CEST), Bogdan Costescu wrote:
>On Thu, 19 Apr 2001, Stefan Hoffmeister wrote:
>
>> "Tiny" transmits will work fine - e.g. ssh / telnet to a host on the LAN
>> works as expected -, but any attempt to transmit some more bits (10 MB
>> file, remote X session) will result in throughput of about 1-2 KB/s.
>
>My guess is that there is a media mismatch. Please read
>
>http://www.scyld.com/network/vortex.html
>
>and post the results of the mii-diag tool.
At the end of this message - together with the output of vortex-diag.
What irritates me is:
Options: force full-duplex.
Vortex format checksum is incorrect (0086 vs. 10b7).
Cyclone format checksum is incorrect (0x16 vs. 0x3f).
Hurricane format checksum is correct (0x3f vs. 0x3f).
Note that I haven't fiddled with that card at all. Plain, bulk 3com card,
dropped out of the bag into the system.
>However, I would have expected some messages in /var/log/messages...
modprobe 3c59x debug=7
modprobe 3c59x debug=99
Same result - none.
>> I am going through a 10/100 hub;
>
>What kind of hub ? Is it just a repeater (hub) or switch ?
The cheapest 10/100 repeater hub around.
***************
Basic registers of MII PHY #24: 3000 782d 0040 6176 05e1 40a1 0003 0000.
Basic mode control register 0x3000: Auto-negotiation enabled.
You have link beat, and everything is working OK.
Your link partner advertised 40a1: 100baseTx 10baseT.
***************
vortex-diag.c:v2.04 1/8/2001 Donald Becker (becker@scyld.com)
http://www.scyld.com/diag/index.html
Index #1: Found a 3c905C Tornado 100baseTx adapter at 0xb000.
Initial window 4, registers values by window:
Window 0: 0000 0000 0000 0000 0000 00bf ffff 0000.
Window 1: 6300 0026 0700 0000 0000 007f 0000 2000.
Window 2: 0100 1902 335d 0000 0000 0000 0052 4000.
Window 3: 0000 0380 05ea 0020 000a 0800 0800 6000.
Window 4: 0000 0000 8000 0cfa 0001 88c0 0000 8000.
Window 5: 1ffc 0000 0000 0600 0807 06ce 06c6 a000.
Window 6: 0000 0000 0000 0000 0000 0000 0000 c000.
Window 7: 0000 0000 0000 0000 0000 0000 0000 e000.
Vortex chip registers at 0xb000
0xB010: 00000000 00000000 0000000a 00000000
0xB020: 00000020 00000000 00080000 00000004
0xB030: 00000000 69689698 172351f0 00080004
Indication enable is 06c6, interrupt enable is 06ce.
No interrupt sources are pending.
Transceiver/media interfaces available: 100baseTx 10baseT.
Transceiver type in use: Autonegotiate.
MAC settings: full-duplex.
Station address set to 00:01:02:19:5d:33.
Configuration options 0052.
Parsing the EEPROM of a 3Com Vortex/Boomerang:
3Com Node Address 00:01:02:19:5D:33 (used as a unique ID only).
OEM Station address 00:01:02:19:5D:33 (used as the ethernet address).
Manufacture date (MM/DD/YYYY) 3/27/2000, division H, product ER.
Options: force full-duplex.
Vortex format checksum is incorrect (0086 vs. 10b7).
Cyclone format checksum is incorrect (0x16 vs. 0x3f).
Hurricane format checksum is correct (0x3f vs. 0x3f).
MII PHY found at address 24, status 782d.
MII PHY 0 at #24 transceiver registers:
3000 782d 0040 6176 05e1 40a1 0003 0000
0000 0000 0000 0000 0000 0000 0000 0000
1000 0300 0000 0000 0000 009b 0300 0000
003e 833e 0f00 ff40 002e 0000 20a0 000b.
From Stefan.Hoffmeister@Econos.de Fri, 20 Apr 2001 02:37:49 +0200
Date: Fri, 20 Apr 2001 02:37:49 +0200
From: Stefan Hoffmeister Stefan.Hoffmeister@Econos.de
Subject: [vortex] 905C: 3c59x unusably slow / 3c90x fine
: On Thu, 19 Apr 2001 22:33:51 +0200, Stefan Hoffmeister wrote:
>What irritates me is:
>
>Options: force full-duplex.
It irritated the 3c59x driver, too.
After tinkering with the 3com DOS configuration tool, the "force
full-duplex" option is gone and
* the 3c59x driver has a throughput of about 8,000 KB/s.
* the dual-booted W2K SP1 with a 3com driver is still its usual
"performance wonder" at 5,300 KB/s, everything else being equal.
* the 3c90x driver continues to work fine
I still find the following irritating:
* I never touched that card with a configuration tool.
It has only ever seen W2K and Linux.
* 3com's own 3c90x driver handled the situation where the 3c59x driver
failed perfectly fine.
Perhaps this could be added to the documentation / mentioned somewhere on
the web site(s)?
From andrewm@uow.edu.au Thu, 19 Apr 2001 19:35:26 -0700
Date: Thu, 19 Apr 2001 19:35:26 -0700
From: Andrew Morton andrewm@uow.edu.au
Subject: [vortex] 905C: 3c59x unusably slow / 3c90x fine
Stefan Hoffmeister wrote:
>
> * 3com's own 3c90x driver handled the situation where the 3c59x driver
> failed perfectly fine.
>
This is becoming a problem, particularly if 3com
are shipping shrink-wrapped NICs with these EEPROM
settings.
A lot of people are migrating from 2.2 kernels
with vendor-provided 3c90x over to 2.4+3c59x
and they are getting bitten by this.
I'll sit down and review our interpretation of the EEPROM
bits versus 3com's driver and versus the datasheet.
And yes, I agree that additional documentation is
needed. It'd probably be useful to blurt out a
warning message at detection time.
From gav@nlr.ru Fri, 20 Apr 2001 10:36:34 +0400
Date: Fri, 20 Apr 2001 10:36:34 +0400
From: Alex Guryanow gav@nlr.ru
Subject: [vortex] Byte counter on 3Com905B
Hi,
I have RedHat 7.0 (kernel 2.4.2) box with 3 3Com cards: 2 3Com905B Cyclone and one 3Com905
Boomerang. The problem is following: statistic gathred from Cisco-router where eth0 is connected
differs from statistic gathered from eth0. /proc/net/dev shows incomming traffic two times greather than
shows Cisco. The outgoing traffic also differs but not so much. The same situation is at eth1
connected to another Cisco-switch. The data about traffic from Cisco and Linux box is gathered using
MRTG software and I have checked that it shows same data as in /proc/net/dev.
3c59x.c:LK1.1.12 06 Jan 2000 Donald Becker and others. http://www.scyld.com/network/vortex.html $Revis
ion: 1.102.2.46 $
See Documentation/networking/vortex.txt
eth0: 3Com PCI 3c905B Cyclone 100baseTx at 0x7880, 00:50:da:35:5d:8f, IRQ 15
product code 'WZ' rev 00.9 date 09-22-99
8K byte-wide RAM 5:3 Rx:Tx split, 10baseT interface.
Enabling bus-master transmits and whole-frame receives.
PCI: Found IRQ 9 for device 00:14.0
PCI: The same IRQ used for device 01:00.0
eth1: 3Com PCI 3c905B Cyclone 100baseTx at 0x7c00, 00:10:5a:6f:92:f6, IRQ 9
product code 'QP' rev 00.12 date 03-18-99
8K byte-wide RAM 5:3 Rx:Tx split, autoselect/Autonegotiate interface.
MII transceiver found at address 24, status 786d.
Enabling bus-master transmits and whole-frame receives.
PCI: Found IRQ 11 for device 00:12.0
eth2: 3Com PCI 3c905 Boomerang 100baseTx at 0x7840, 00:60:08:74:dd:78, IRQ 11
product code 'KK' rev 00.0 date 11-24-97
8K word-wide RAM 3:5 Rx:Tx split, autoselect/MII interface.
MII transceiver found at address 24, status 786f.
Enabling bus-master transmits and whole-frame receives.
I have installed iproute2 and netfilter, but think they shouldn't influence this problem
From devdrvr@davis.com Fri, 20 Apr 2001 08:37:58 -0700
Date: Fri, 20 Apr 2001 08:37:58 -0700
From: devdrvr devdrvr@davis.com
Subject: [vortex] Gigabit NSC NIC
Is anyone working on, or no about the new gigabit card and chip from
National Semi (what is the name, and where did the ASIC come from)? Is
there a Linux driver for it? What about performance; I hear it knocks the
socks off the Tigon-2 based (now) 3COM card (including my MacOS driver,
ouch). Supposedly, Netgear bought Alteon and because they had a partnership
deal with NSC for this new gig chip, that is why they sold the division to
3COM; discarding it because they didn't want it.
--Lu
Device Driver Programmer
Davis, CA
01 (530) 400-5692
From becker@scyld.com Fri, 20 Apr 2001 13:24:30 -0400 (EDT)
Date: Fri, 20 Apr 2001 13:24:30 -0400 (EDT)
From: Donald Becker becker@scyld.com
Subject: [vortex] Re: Gigabit NSC NIC
On Fri, 20 Apr 2001, devdrvr wrote:
> Is anyone working on, or no about the new gigabit card and chip from
> National Semi (what is the name, and where did the ASIC come from)?
The chips are the 820 series, the DP83820 (prototype only?) and 83821.
It's NatSemi's own design.
It's based on the 810 series. The 83815 was used on the Netgear FA311
and 312. Some registers are moved, and addresses can be 64 bits in some
cases, so a common driver doesn't make sense.
> Is there a Linux driver for it?
I've written a driver named 'ns820.c', and there is reportedly a driver
included with the D-Link version of the board.
> What about performance; I hear it knocks the
> socks off the Tigon-2 based (now) 3COM card (including my MacOS driver,
> ouch).
It has on-chip checksumming, albeit implemented a little awkwardly.
Like the 810 series, it can only receive into aligned buffers so IP
headers will be misaligned.
It's a hardware implementation and thus has latency advantage over the
Acenic firmware.
> Supposedly, Netgear bought Alteon and because they had a partnership
> deal with NSC for this new gig chip, that is why they sold the division to
> 3COM; discarding it because they didn't want it.
It's a little more complicated than that. 3Com bought only the design,
not the people (that presumably would know how to update it).
Donald Becker becker@scyld.com
Scyld Computing Corporation http://www.scyld.com
410 Severn Ave. Suite 210 Second Generation Beowulf Clusters
Annapolis MD 21403 410-990-9993
From jes@linuxcare.com 20 Apr 2001 20:06:30 +0200
Date: 20 Apr 2001 20:06:30 +0200
From: Jes Sorensen jes@linuxcare.com
Subject: [vortex] Re: Gigabit NSC NIC
>>>>> "Donald" == Donald Becker writes:
Donald> On Fri, 20 Apr 2001, devdrvr wrote:
>> What about performance; I hear it knocks the socks off the Tigon-2
>> based (now) 3COM card (including my MacOS driver, ouch).
Donald> It has on-chip checksumming, albeit implemented a little
Donald> awkwardly. Like the 810 series, it can only receive into
Donald> aligned buffers so IP headers will be misaligned.
Oh dear, not again ;-( Looks like another failure.
Jes
From bib@artamis.org Mon, 23 Apr 2001 06:07:52 +0200
Date: Mon, 23 Apr 2001 06:07:52 +0200
From: Bertrand Habib bib@artamis.org
Subject: [vortex] Cannot get 3c905c-tx running
Hello,
I've a 3c905c-tx installed on a Asus A7Pro. It is reconised by the
kernel, ie: dmsg, ifconfig okay,
but no way to get it running on 2.2.17, nor 2.2.18 linux kernel
(debian). It don't ping/pong at all and
and dhcp server on that host distribute meaningless IPs.
Firmware setting of the device is N-Way autodetect for both, media and
full-duplex mode.
This device was connected first to a 100/10 switch (officeconnect), and
then on a 10 MB hub without behaviour change.
I've tried "no option" as well as "16" option for the driver module
without any behaviour change.
I've checked this card with 3com's 3c90xcfg diag under dos. First
internal check, then send check and echo server check.
No problem has been reported.
Changing this card with a 3c905c, everythings works fine (but I have to
give back 3c905c)
Could somebody help me to find out what's append ?
Many thanks in advance
Kindest regards
Bertrand Habib
From bib@artamis.org Mon, 23 Apr 2001 06:17:31 +0200
Date: Mon, 23 Apr 2001 06:17:31 +0200
From: Bertrand Habib bib@artamis.org
Subject: [vortex] ERRATA - Cannot get 3c905c-tx running
Sorry for mistake in product name.
The problem which i describe belong to 3c905CX-TXM adapter!!
Bertrand Habib wrote:
> Hello,
>
> I've a 3c905c-tx installed on a Asus A7Pro. It is reconised by the
> kernel, ie: dmsg, ifconfig okay,
> but no way to get it running on 2.2.17, nor 2.2.18 linux kernel
> (debian). It don't ping/pong at all and
> and dhcp server on that host distribute meaningless IPs.
>
> Firmware setting of the device is N-Way autodetect for both, media and
> full-duplex mode.
> This device was connected first to a 100/10 switch (officeconnect), and
> then on a 10 MB hub without behaviour change.
> I've tried "no option" as well as "16" option for the driver module
> without any behaviour change.
> I've checked this card with 3com's 3c90xcfg diag under dos. First
> internal check, then send check and echo server check.
> No problem has been reported.
>
> Changing this card with a 3c905c, everythings works fine (but I have to
> give back 3c905c)
>
> Could somebody help me to find out what's append ?
>
> Many thanks in advance
> Kindest regards
>
> Bertrand Habib
From bogdan.costescu@iwr.uni-heidelberg.de Mon, 23 Apr 2001 13:10:10 +0200 (CEST)
Date: Mon, 23 Apr 2001 13:10:10 +0200 (CEST)
From: Bogdan Costescu bogdan.costescu@iwr.uni-heidelberg.de
Subject: [vortex] ERRATA - Cannot get 3c905c-tx running
On Mon, 23 Apr 2001, Bertrand Habib wrote:
> Sorry for mistake in product name.
> The problem which i describe belong to 3c905CX-TXM adapter!!
After reading your first message, that was my guess too!
For proper support for 905CX, you need the driver from 2.2.19.
Sincerely,
Bogdan Costescu
IWR - Interdisziplinaeres Zentrum fuer Wissenschaftliches Rechnen
Universitaet Heidelberg, INF 368, D-69120 Heidelberg, GERMANY
Telephone: +49 6221 54 8869, Telefax: +49 6221 54 8868
E-mail: Bogdan.Costescu@IWR.Uni-Heidelberg.De
From skunk@cg.tuwien.ac.at Thu, 26 Apr 2001 07:30:12 +0200
Date: Thu, 26 Apr 2001 07:30:12 +0200
From: Stephan Plepelits skunk@cg.tuwien.ac.at
Subject: [vortex] Wake-On-Lan Problem with 3c905C-TX
Hi!
I'm installing a nice cluster, the machines all have a 3c905C (the new one,
with a smaller layout).
I want to use the wake on lan feature. First try (after pluging the power
to it): I'm sending this magic packet, and it really starts.
After poweroff, the network card also is off (So the machine can not be
waked). so I searched the Web, and found 'pci-config'. After a patch to the
program (the ID of the NIC, and the ID the program requires, is not the
same. pci-config reports 920010b7, it requires 905510b7 for the 3com 905B),
the card seems to go to wake-on-lan-mode (the network is down, after
poweroff the card keeps flashing.
Problem: I can't wake the machine, although it seems to be right.
Is it possible, that the command for the 3c905b and -c is different?
Shall I provide you with more information? Output of pci-config, when
setting the card to wake-on-lan-mode?
greetings and thanks,
Stephan
--
** ** * * ** ** * * * ** ** * * * * ** ** * * ** *
Stephan Plepelits
Vienna University of Technology
Institute of Computergraphics and Algorithms
plepelits@cg.tuwien.ac.at http://www.cg.tuwien.ac.at
Tel.: +43 1 58801x18634 Fax: +43 1 58801x18698
** * * * * * * ** ** * * * * * * ** * * ** * * * *
From becker@scyld.com Thu, 26 Apr 2001 18:55:52 -0400 (EDT)
Date: Thu, 26 Apr 2001 18:55:52 -0400 (EDT)
From: Donald Becker becker@scyld.com
Subject: [vortex] Wake-On-Lan Problem with 3c905C-TX
On Thu, 26 Apr 2001, Stephan Plepelits wrote:
> I'm installing a nice cluster, the machines all have a 3c905C (the new one,
> with a smaller layout).
Have you tried the Scyld cluster system? We have W-O-L support in the
beosetup tool, and future versions will have wake-up rules in the node
scheduler. That will means you may write scheduler rules that describe
when additional compute nodes should be powered up or powered down.
> I want to use the wake on lan feature. First try (after pluging the power
> to it): I'm sending this magic packet, and it really starts.
>
> After poweroff, the network card also is off (So the machine can not be
> waked). so I searched the Web, and found 'pci-config'. After a patch to the
> program (the ID of the NIC, and the ID the program requires, is not the
> same. pci-config reports 920010b7, it requires 905510b7 for the 3com 905B),
> the card seems to go to wake-on-lan-mode (the network is down, after
> poweroff the card keeps flashing.
That code exists in 'pci-config' from when I was experimenting to
understand how the 3Com hardware worked. That knowledge was put into
my driver so that 'pci-config' is no longer needed.
I'm guessing that your problem is because are running a variant driver
that does not leave the device configured for W-O-L mode and in D3
state.
> Problem: I can't wake the machine, although it seems to be right.
> Is it possible, that the command for the 3c905b and -c is different?
The 'B' and 'C' cards are significantly different. The 'B' card was an
early implementation of WOL. It required that the driver or
'pci-config' be run before WOL was enabled. Yes, the machine had to be
powered up before it could be ...powered up.
The 'C' card fixed logical oversight by adding a bunch of information to
the EEPROM that the hardware loads to configure the chip when stand-by
power is first applied.
Donald Becker becker@scyld.com
Scyld Computing Corporation http://www.scyld.com
410 Severn Ave. Suite 210 Second Generation Beowulf Clusters
Annapolis MD 21403 410-990-9993