The situation is that I have a Fedora 12 and I get this messages:
dmesg
eth0: link down
eth0: link up, 100Mbps, full-duplex, lpa 0x45E1
eth0: link down
eth0: link up, 100Mbps, full-duplex, lpa 0x45E1
eth0: link down
that it i caused by the a buggy/bad driver (I read).
Digging into the system, I saw I have a MSI K9VGM-V motherboard who has a
Realtek 8201CL LAN card but the drive I have running is the via-rhine:
00:12.0 Ethernet controller: VIA Technologies, Inc. VT6102 [Rhine-II] (rev
7c)
So my question here is:
Am I using the right driver? Fedora picked it during installation. Is it
possible the manufacturer has changed the network card form the Realtek to
the VIA and Linux is detection properly?
Or autodetection is not working and I should change it to a Realtek driver
and then my problem would go away? How should I do that then?
Any hint would be appreciated. Thanks in advance.
--
Lord Eldritch
Lord Eldritch a ᅵcrit :
>
> Digging into the system, I saw I have a MSI K9VGM-V motherboard who has a
> Realtek 8201CL LAN card but the drive I have running is the via-rhine:
>
> 00:12.0 Ethernet controller: VIA Technologies, Inc. VT6102 [Rhine-II] (rev
> 7c)
This is only the device identification, not the running driver.
> So my question here is:
>
> Am I using the right driver?
Probably.
> Fedora picked it during installation. Is it
> possible the manufacturer has changed the network card form the Realtek to
> the VIA and Linux is detection properly?
Maybe the motherboard documentation is wrong, because in the driver list
the LAN driver is for a VIA VT6103/VT6105 controller. Can you see the
Realtek 8201CL chip on the motherboard ?
> Or autodetection is not working
This is highly unlikely, or the motherboard is horribly broken.
> and I should change it to a Realtek driver
> and then my problem would go away?
This won't work because a driver has a built-in list of identifiers it
can handle.
>> Fedora picked it during installation. Is it
>> possible the manufacturer has changed the network card form the Realtek
>> to the VIA and Linux is detection properly?
>
> Maybe the motherboard documentation is wrong, because in the driver list
> the LAN driver is for a VIA VT6103/VT6105 controller. Can you see the
> Realtek 8201CL chip on the motherboard ?
>
>> Or autodetection is not working
>
> This is highly unlikely, or the motherboard is horribly broken.
>
Well. I went for the brute force approach and I opened the case. The place
for the network chip shows a very small one (0.5�0.5cm maybe) showing
Realtek on it and RLT8201CL (I cannot swear on it because they are very
tiny). The chips around are all with "crab" and branded as Realtek. The only
Via chips are the (covered by dissipator) *bridges.
--
Lord Eldritch
> Hi, I've been surfing the net to find the solution to this, but the
> only clear thing I've got was that it was a driver problem.
>
> The situation is that I have a Fedora 12 and I get this messages:
> dmesg
> eth0: link down
> eth0: link up, 100Mbps, full-duplex, lpa 0x45E1
> eth0: link down
> eth0: link up, 100Mbps, full-duplex, lpa 0x45E1
> eth0: link down
>
> that it i caused by the a buggy/bad driver (I read).
>
Could be a bad driver, or a bug anyway, but you appear to be using the
right driver. Anyway, the above can be caused by a bad/loose
connection, a bad cable, bad port on a router/switch, etc. as well.
--
Not really a wanna-be, but I don't know everything.
I'm so stupid. I should have remembered that the RTL8201CL is not a
controller but only a PHYceiver between the link and the controller
which must be the VT6102 integrated in the VIA chipset.
I'd be very surprised if Fedora (not the distribution I use) got
something basic like that wrong. Have you thought of buying/borrowing
another card? You presumably have a PCI slot free. I'd deactivate
the card (Linux config or BIOS) and get something which uses a
different driver. Most cards have Linux drivers, I was really
surprised to find one recently which did not (an old Level One
FNC-0109TX).
Obviously, if the situation persists then something outside your PC is
wrong.
When I started checking things, I've got puzzled by the Realtek/Via duality
and I thought i could be a driver thing, but you all helped me to establish
that was ok and I should probably look for the cause somewhere outside.
I trust linux but it is always easier to blame the software before start
checking the hardware..
I changed cable now and connection port to the router and it remains. Now I
am checking the router configuration...
Any diagnostic tool I could use to locate the failure?
I guess I'll try to find a network card and check it too...
--
Lord Eldritch
I use ifconfig to check for TX/RX errors. Other hardware tools of interest
might be ethtool or mii-tool. Then there is wireshark for looking
through the packet headers.
If speed tests look good, then blame the network to/from the ftp
site or your isp.
In no order of importance or checked url lately
http://www.speakeasy.net/speedtest/
http://homepage.eircom.net/~leslie/testpage.htm
http://support.midco.net/
http://reviews.cnet.com/7004-7254_7-0.html
http://www.broadbandreports.com/stest
http://www.browsertune.com/bt2kfast/
http://www.cable-modem.net/
http://www.dslreports.com/stest
http://www.dslreports.com/tweaks
http://www.testmy.net
http://www.toast.net/performance/
It still might be a bug. I've not used Fedora in years, thankfully (my
own opinion). I just am glad you checked the simple things first
before taking more of your time. Others recommended swapping nic's and
see if that helps (onboard or another card, if you have either option)
and see if the issue remains. Also, you might try a change in the
settings:
ethtool -K eth0 tso off
This was a known issue that would cause it to drop (if tso is enabled).
See if that helps?
So far I've tried changing cable and changing the router port. Also
installed a new network card I had (Zyxel -Realtek) and it is detected and a
Realtek driver is loaded but as soon as I plug the cable on it, the same
message appears (this time on eth1, of course). I guess this clears the
drivers of having bugs.
I'll go though all the suggestions this weekend.
I was also suggested by a friend that could also be just a router failure
(Zyxel 660 HW) and just the ports circuits are giving up. Can it be that
way?
--
Lord Eldritch
> ethtool -K eth0 tso off
>
> This was a known issue that would cause it to drop (if tso is enabled).
> See if that helps?
The answer is:
[root@eagle2 /tmp]# ethtool -K eth0 tso off
Cannot set device tcp segmentation offload settings: Operation not supported
This is the network config:
eth0 Link encap:Ethernet HWaddr 00:19:DB:2F:C9:80
inet addr:192.168.1.2 Bcast:192.168.1.255 Mask:255.255.255.0
inet6 addr: fe80::219:dbff:fe2f:c980/64 Scope:Link
UP BROADCAST MULTICAST MTU:1500 Metric:1
RX packets:320006 errors:0 dropped:0 overruns:0 frame:8
TX packets:283084 errors:0 dropped:0 overruns:0 carrier:0
collisions:0 txqueuelen:1000
RX bytes:304801331 (290.6 MiB) TX bytes:45149552 (43.0 MiB)
Interrupt:23
It is alternated with
UP BROADCAST RUNNING MULTICAST MTU:1500 Metric:1
I guess that's the result of the up/down loop.
I has been able to plug a Windows (7) laptop wire and wireless-way and
everything looked fine from that. Tested speed and reached 2.5Mb/s (from a
declared 3Mb/s). So it looks again that problem is back is in the linux box.
I'll try to compile a kernel to see it I have the same problem with the new
drivers.
--
Lord Eldritch
Settings for eth0:
Supported ports: [ TP MII ]
Supported link modes: 10baseT/Half 10baseT/Full
100baseT/Half 100baseT/Full
Supports auto-negotiation: Yes
Advertised link modes: Not reported
Advertised auto-negotiation: No
Speed: 10Mb/s
Duplex: Full
Port: MII
PHYAD: 1
Transceiver: internal
Auto-negotiation: off
Supports Wake-on: pumbg
Wake-on: d
Current message level: 0x00000001 (1)
Link detected: yes
I've set the speed to 10Mb/s and autoneg off. I am don't understand why:
router and card are able of 100 and the cables I have (and some other I've
tried are new 1 Gb ones)...
You to see if this rings you a bell...:) but , in any case, thank you all.
At least, the network here is .. behaving...:)
--
Lord Eldritch
Just because the router is capable of 100 MB/s doesn't mean you have it
set to 100 Mb/s.
Note that if you disable auto-negotiation on your side, if the other
side has auto-negotiation enabled then it might revert to half duplex
because it does not know that your side is capable of full duplex. So
you may have to force half duplex on your side or full duplex on the
other side. Check the collision counter on the other side.
> Just because the router is capable of 100 MB/s doesn't mean you have it
> set to 100 Mb/s.
>
Yes but it is recognized by it. When I change speed to 100, the LED changes
indicating the new speed, but it just, there is no way to ping out the
network.
--
Lord Eldritch