Problems with network speed setup in F12

Tue Jan 26 17:24:38 UTC 2010

On 01/26/2010 02:13 AM, Gianluca Cecchi wrote:
> On Fri Jan 22 17:48:03 UTC 2010 Rick Stevens wrote:
>> It's NetworkManager doing that to you.
> Ok, as I supposed.
>
>> Is there a reason they've set that up as fixed?  Is there an autoneg
>> issue with the switch (some don't do autonegotiation correctly).
>
> There was a problem on older gigabit switches and for historical
> reasons was set up this way yet. Now they are on new ones and I asked
> them to put the switch port to autoneg.
> OK now for the speed part. I get 100/full after autoneg at boot.
>
> But when I start for example an scp session of big files (around some
> tenths of megabytes) from/to my pc, I get many link down messages.....
> The transfer begins at 9.7MB/s and then continuously decreases at
> about 7 and also 5 MB/s.....
>
> See below my /var/log/messages after reboot at about 09.30 this morning.
> For example at 10:23 and 11:05 I started scp sessions (single file of
> about 2.5Gb)
> Jan 26 09:34:54 tekkaman kernel: tg3: eth0: Link is down.
> Jan 26 09:34:54 tekkaman NetworkManager:<info>   (eth0): carrier now
> OFF (device state 8, deferring action for 4 seconds)
> Jan 26 09:34:55 tekkaman kernel: tg3: eth0: Link is up at 100 Mbps, full duplex.
> Jan 26 09:34:55 tekkaman NetworkManager:<info>   (eth0): carrier now
> ON (device state 8)
> Jan 26 09:34:55 tekkaman kernel: tg3: eth0: Flow control is off for TX
> and off for RX.
> Jan 26 09:38:56 tekkaman ntpd[1264]: synchronized to 10.4.1.46, stratum 3
> Jan 26 09:38:56 tekkaman ntpd[1264]: time reset +0.391977 s
> Jan 26 09:38:56 tekkaman ntpd[1264]: kernel time sync status change 2001
> Jan 26 09:39:10 tekkaman dnsmasq[1451]: reading /etc/resolv.conf
> Jan 26 09:39:10 tekkaman dnsmasq[1451]: using nameserver 10.4.1.121#53
> Jan 26 09:39:10 tekkaman dnsmasq[1451]: using nameserver 10.4.1.120#53
> Jan 26 09:46:05 tekkaman ntpd[1264]: synchronized to 10.4.1.46, stratum 3
> Jan 26 09:51:02 tekkaman NetworkManager:<info>   (wlan0): device state
> change: 3 ->  2 (reason 0)
> Jan 26 09:51:02 tekkaman NetworkManager:<info>   (wlan0): deactivating
> device (reason: 0).
> Jan 26 09:51:02 tekkaman NetworkManager:<info>   Policy set 'Poli'
> (eth0) as default for routing and DNS.
> Jan 26 09:51:02 tekkaman NetworkManager:<info>   (wlan0): taking down device.
> Jan 26 09:51:02 tekkaman dbus-daemon: Rejected send message, 2 matched
> rules; type="method_return", sender=":1.9" (uid=0 pid=1226
> comm="NetworkManager) interface="(unset)" member="(unset)" error
> name="(unset)" requested_reply=0 destination=":1.45" (uid=500 pid=1773
> comm="nm-applet))
> Jan 26 10:08:18 tekkaman dnsmasq[1451]: reading /etc/resolv.conf
> Jan 26 10:08:18 tekkaman dnsmasq[1451]: using nameserver 10.4.1.121#53
> Jan 26 10:08:18 tekkaman dnsmasq[1451]: using nameserver 10.4.1.120#53
> Jan 26 10:22:37 tekkaman kernel: tg3: eth0: Link is down.
> Jan 26 10:22:37 tekkaman NetworkManager:<info>   (eth0): carrier now
> OFF (device state 8, deferring action for 4 seconds)
> Jan 26 10:22:39 tekkaman kernel: tg3: eth0: Link is up at 100 Mbps, full duplex.
> Jan 26 10:22:39 tekkaman kernel: tg3: eth0: Flow control is off for TX
> and off for RX.
> Jan 26 10:22:39 tekkaman NetworkManager:<info>   (eth0): carrier now
> ON (device state 8)
> Jan 26 10:23:37 tekkaman kernel: tg3: eth0: Link is down.
> Jan 26 10:23:37 tekkaman NetworkManager:<info>   (eth0): carrier now
> OFF (device state 8, deferring action for 4 seconds)
> Jan 26 10:23:39 tekkaman kernel: tg3: eth0: Link is up at 100 Mbps, full duplex.
> Jan 26 10:23:39 tekkaman NetworkManager:<info>   (eth0): carrier now
> ON (device state 8)
> Jan 26 10:23:39 tekkaman kernel: tg3: eth0: Flow control is off for TX
> and off for RX.
> Jan 26 10:40:53 tekkaman kernel: CE: hpet increasing min_delta_ns to 15000 nsec
> Jan 26 11:04:00 tekkaman kernel: tg3: eth0: Link is down.
> Jan 26 11:04:00 tekkaman NetworkManager:<info>   (eth0): carrier now
> OFF (device state 8, deferring action for 4 seconds)
> Jan 26 11:04:02 tekkaman NetworkManager:<info>   (eth0): carrier now
> ON (device state 8)
> Jan 26 11:04:02 tekkaman kernel: tg3: eth0: Link is up at 100 Mbps, full duplex.
> Jan 26 11:04:02 tekkaman kernel: tg3: eth0: Flow control is off for TX
> and off for RX.
> Jan 26 11:05:50 tekkaman kernel: tg3: eth0: Link is down.
> Jan 26 11:05:50 tekkaman NetworkManager:<info>   (eth0): carrier now
> OFF (device state 8, deferring action for 4 seconds)
> Jan 26 11:05:52 tekkaman kernel: tg3: eth0: Link is up at 100 Mbps, full duplex.
> Jan 26 11:05:52 tekkaman NetworkManager:<info>   (eth0): carrier now
> ON (device state 8)
> Jan 26 11:05:52 tekkaman kernel: tg3: eth0: Flow control is off for TX
> and off for RX.

Link down/up sequences are pretty indicative of a bad switch or a bad
network cable.  NetworkManager won't cause those, but it will respond
to them.  Note that the "NetworkManager: <info>" messages always follow
"kernel: tg3" messages (which are coming from the driver).

I'd first try a new cable from your system to your wallplate (if you're
wired like that).  If there's no improvement, I'd have your IT people
check the wiring in the wall and on your wallplate.  If it's a typical
66-type punchdown wallplate socket, they sometimes get grungy with age
and a repunch is necessary.
----------------------------------------------------------------------
- Rick Stevens, Systems Engineer                      ricks at nerd.com -
- AIM/Skype: therps2        ICQ: 22643734            Yahoo: origrps2 -
-                                                                    -
-   "Do you suffer from long-term memory loss?"  "I don't remember"  -
-                            -- Chumbawumba, "Amnesia" (TubThumping) -
----------------------------------------------------------------------