eth0 randomly diconnecting and reconnecting

[GreyGeek]

I switched to an eth0 connection from using my wireless connection because connecting via wifi wouldn't allow my grandson to log into my minecraft server.

My past experience with Ethernet connections are that they are rock solid stable. Not so any more.

It randomly disconnects and then, usually within 10 seconds and sometimes after 30 or more seconds, it will reconnect.

The ethernet network chip is AR8151.

sudo hwinfo --netcard shows:

.....(omitting wlan0 device info) ......

25: PCI e00.0: 0200 Ethernet controller
  [Created at pci.318]
  UDI: /org/freedesktop/Hal/devices/pci_1969_1083
  Unique ID:....
  Parent ID: ...
  SysFS ID: /devices/pci0000:00/0000:00:1c.5/0000:0e:00.0
  SysFS BusID: 0000:0e:00.0
  Hardware Class: network
  Model: "Attansic Ethernet controller"
  Vendor: pci 0x1969 "Attansic Technology Corp."
  Device: pci 0x1083 
  SubVendor: pci 0x1025 "Acer Incorporated [ALI]"
  SubDevice: pci 0x0686 
  Revision: 0xc0
  Driver: "atl1c"
  Driver Modules: "atl1c"
  Device File: eth0
  Memory Range: 0xd3400000-0xd343ffff (rw,non-prefetchable)
  I/O Ports: 0x2000-0x2fff (rw)
  IRQ: 47 (11838 events)
  HW Address: 4c:72:b9:2d:8d:99
  Link detected: yes
  Module Alias: "pci:v00001969d00001083sv00001025sd00000686bc02sc00i00"
  Driver Info #0:
    Driver Status: atl1c is active
    Driver Activation Cmd: "modprobe atl1c"
  Config Status: cfg=new, avail=yes, need=no, active=unknown
  Attached to: #17 (PCI bridge)

and :~$ sudo hwinfo --netcard | grep "Module\|File" shows:

Driver Modules: "ath9k", "ath9k"
Device File: wlan0
Module Alias: "pci:v0000168Cd00000034sv000011ADsd00006621bc02sc8 0i00"
Driver Modules: "atl1c"
Device File: eth0
Module Alias: "pci:v00001969d00001083sv00001025sd00000686bc02sc00i00"

The modinfo command showed no adjustable PARAMS but shows support for
alias: pci:v00001969d00001083sv*sd*bc*sc*i*



It seems that the atl1c driver is "experimental" so I don't expect a bug report would do any good.
http://cateee.net/lkddb/web-lkddb/ATL1C.html


Ksystemlog shows

....kernel [ 7725.138784] atl1c 0000:0e:00.0: vpd r/w failed. This is likely a firmware bug on this device. Contact the card vendor for a firmware update.

Launchpad bug reports show no usable information. Other posts relating to the alt1 driver show three cases, 1) everything works fine, 2) atl1 works intermittently, or 3) atl1 doesn't work at all. 1&3 are not helpful. Advice in type 2 posts usually suggest recompling the atl1 driver. I doubt that would be useful since my atl1c driver is specifically for my 1969 1803 chip.

My grandson just logged into my Minecraft 162 server. When my atl1c disconnects he is thrown out but he can reconnect about as quick as he can get to the "join" button.

In searching around this problem with the atl1 driver has been around for more than 4 years, so I suspect that I will just have to put upwith it until a kernel comes down the pike which fixes it for my machine.

[GreyGeek]

After some experimenting around I discovered that the disconnect/reconnect occurs with both the eth0 and the wlan0 connections, which means the ath9k driver may be guilty of the same crime. So, perhaps my atl1c driver is not the flaky culprit I thought it was.

Thinking it could be kernel related I loaded up the 3.4 and the 3.2 kernels. Both gave me the same problem on both the eth0 and wlan0 connections, so I'll stick with the 3.8 kernel.

As usual, the system logs shows no evidence of the problem.

I know it's not the windstream DSL router. My Win7 side runs fine, as does my wife's Acer Aspire 0D521 running Kubuntu 12.04, as has every other computer that's connected.

So, I am suspecting that it is a network component in this Acer Aspire V3-771G that is common to both the eth0 and the wlan0 circuitry.

[GreyGeek]

I downloaded Kubuntu 13.04 and tried it via a USB stick. It gives me the exact same problem with both the eth0 and the wlan0 connections.

[SteveRiley]

My HP Mini 2140, acting as my do-everything server, has started exhibiting similar behavior. It simply disappears from the network every four days or so. And when that happens, all hell breaks loose at home, for it's my DNS server and SMTP/IMAP sever, among others. I have to power cycle the thing. Logs are devoid of any clues. I'm stumped.

root@mini2140:~# [B]lspci -k[/B]
...
20:00.0 Ethernet controller: Marvell Technology Group Ltd. 88E8072 PCI-E Gigabit Ethernet Controller (rev 10)
        Subsystem: Hewlett-Packard Company Device 3056
        Kernel driver in use: sky2
        Kernel modules: sky2

[GreyGeek]

Steve, I found that executing
echo "1969 1083" > /sys/bus/pci/drivers/atl1c/new_id
and rebooting didn't help. The numbers are the vendor and product number of the eth chip in my machine.

The r&d affects both my wireless and wired connection. I had come to the conclusion that some piece of hardware in this laptop common to both had gone on the fritz.
Then you posted your msg. Now I am not so sure.

The problem persists with the all three kernels -- 3.2, 3.5 and 3.8.
My Atheros AR8151 chip triggers the atl1c driver. There are no PARMs to modify.
Your sky2 module shows
parm: debugebug level (0=none,...,16=all) (int)
parm: copybreak:Receive copy threshold (int)
parm: disable_msiisable Message Signaled Interrupt (MSI) (int)
parm: legacy_pme:Legacy power management (int)

Several people have tried recompiling the latest driver set: compat-wireless-2012-07-03-pc.tar.bz2
with mixed results. I'm not ready to go that route because of the likelihood that the automatic repository updates will overwrite them.

I'm retired so I will soldier on until this mystery is solved, or at least I know the reasons why.

[GreyGeek]

While googeling the problem I came across this "fix":
http://superuser.com/questions/40027...to-the-network
The problem was the router.
I may have Windstream check my router from their end. I looked at the router log and so no entries that could be associated with the disconnect episodes, but there sure are a lot of sites hitting on my router's firewall and generating "intrusion" msgs.

[SteveRiley]

I've disabled the wireless NIC in my Mini 2140 and am using only Ethernet. I've seen a few folks post complaints about the driver for the Marvell Ethernet NIC and even some modified source code. But what mystifies me is that I've had Precise running on the little guy since April 2012, while the problem I'm having just began to appear last month.

Anyway, I'm going to give up trying to fix it. I ordered a new ThinkStation M92p from NewEgg today. All Intel, as every good computer should be.

[GreyGeek]

For several months I have been connecting to my eth0 port via a short piece of cat5 cable, since my router is only two feet from my computer. Usually, however, both my wireless and wired made connections. In addition, my IPv6 tunnel also had a connection. Probably half my traffic is by IPv6. One site I found suggested that the problem was that the wireless signal strength was too great. Another suggested that channel interference was the problem. I checked my router and it reported severe interference on the channel I was on. I tried several other channels and all were severe. My wife's Acer Aspire One (also running Precise and just updated yesterday to KDE 4.11) always connects by wireless and doesn't have the d&r problem, and the severity of the channel congestion doesn't seem to affect her network performance. That leads me to the conclusion that there is nothing wrong with the wireless router.

So, yesterday I, too, decided to disable the wireless in NetworkManager on my PC. And, I use that echo command:
echo "1969 1083" > /sys/bus/pci/drivers/atl1c/new_id
as root (sudo su).

With the wireless disabled and the "new_id" set to indicate the vendor & model number of my Ethernet PCI device, I've noticed the following: I still get d&r'ss. Not nearly as frequent as with the wireless running and new_id not set. As I continue to use my PC the frequency of d&r's diminish from one every 3 or 4 minutes to one every hour or so. When it does disconnect it reconnects a lot faster. I also noticed that a disconnect can occur under a rapidly changing load. I can download a 1GB ISO without problems but if I click ahead on a 1 hour YouTube movie it often causes a disconnect.

After seeing the "intrusion" attempt msgs on the router log I decided to install Snort and watch.

As usual, logs show nothing.

[GreyGeek]

mmm... Well, the problem seems to have faded away. I have plasma widget which indicates what my ISP IP address is. When my connection disconnects it switches from displaying the IP address to showing the German phrase for "Not Connected". However, it usually reconnects so fast that my movies or downloads or page displays are seldom affected.

[GreyGeek]

Since 8-22-13 the problem had come back with a vengeance. Neither my wired nor my wireless connection would maintain a reliable connection for any useful length of time.
Running "sudo service network-manager restart" renewed the connection but within minutes the "ping google.com" packet times would jump from 30ms to 120ms to 1500ms and the band speed would drop from 12Mb/s to 50kb/s or even 0kb/s.

I booted a Kubuntu 13.04 LiveUSB to see if it had the same problem. IT did NOT. I checked the kernel. It was a 3.8-19. I was running 3.8-29. I checked and noticed that I didn't have the 3.8-19 kernel but I did have the 3.8-52 kernel, compiled on July 26, 2013. I gave it a try.

My system is now as fast as ever and has been so every since I booted up the 3.8.52 kernel 3.5 hours ago. IF it were gong to stall it would have done it by now.
I consider this problem sovled.

[Snowhog]

Not to sidetrack this thread, but "since I booted up the 3.8.52 kernel"

Your running 12.04; I'm on 13.04. Why isn't there a 3.8 kernel (above)?

[GreyGeek]

http://www.leaseweblabs.com/2013/07/...kernel-to-3-4/

[Buddlespit]

https://wiki.ubuntu.com/Kernel/MainlineBuilds
and
http://liquorix.net/

[GreyGeek]

I spoke too soon. The problem of the disconnecting internet was solved, but my cure introduced another problem.

My 3D desktop acceleration disappeared. I found this out when I missed certain accelerated activities, like running Minecraft. When I tried to activate the desktop I was thrown back to the login. However, the new kernel did leave more informative messages in the logs. In the 10 hours or so I ran my eth0 connection I had only one disconnect. The log said the eth0 device was put into a "sleep mode" because of nvidia IRQ conflict.

So, I purged all my bumblebee, primus, dmks and nvidia apps and installed the latest nvidia driver from the repository and reinstalled the rest. Then I booted back into my old kernel, the 29 one.

Both my wifi and eth connections have been rock solid stable and I got 2.1 Mb/s download speeds from my Windstream 12Mb/s connection, downloading ASCENT.ISO, which is a 1080HD video of various views of a space shuttle's launch and accent into orbit.

So, my 3D is working great -- glxspheres gives me

optirun -c yuv glxspheres
Polygons in scene: 62464
Visual ID of window: 0x20
Context is Direct
OpenGL Renderer: GeForce GT 650M/PCIe/SSE2
295.166080 frames/sec - 294.634781 Mpixels/sec
299.936641 frames/sec - 299.396755 Mpixels/sec
300.799768 frames/sec - 300.258329 Mpixels

NOW it is solved.

[GreyGeek]

RATS!

My system has been running fantastic since I did the purge and reinstall of nvidia, bumblebee, primus and dmks apps. I was getting 300 frames/sec and NO network disconnects. The Internet connection was the fastest its been in months.

At 9 AM an update notice came. I clicked on it and not 60 seconds after it was done my network connection broke. I ran
sudo service network-manager restart
and it started right back up, but broke within a minute. Rinse and repeat. (There isn't an animated icon which appropriately expresses my anger right now.)
Now I can't get any faster than 200 frames/sec. But, Internet speed is OK and the eth0 connection seems stable. Time will tell.

I started up muon and checked the history to see what was updated. Then I reverted as many as I could to the previous version. Top on the list was network-manager. I've attached a graphic showing what was upgraded and those files which I could downgrade. Muon wouldn't let me revert them all. I need to find the apt command that allows me to specifically revert to the previously installed package. Either that, or reinstall 12.04 from the latest ISO.