[Nut-upsuser] stopping ups services resets my network

[Antoine Gatineau]

 Hello,

I'm back again on this issue.
I have contacted Red Hat support but they won't do anything to help since it
is not an official Red Hat package. Unless I prove the conflict come from
RH.

So I am trying to figure out what is going on.

Adding -DDDDD will not add any more log at the services stop.

When I look at the sources, I couldn't find the piece of code that release
properly the usb UPS device. There is just a normal SIGTERM handling. Is
this normal?

I also noticed that there are debug logs but I don't know how to activate
them... Can somebody tell me how to do it?

Do you have any advice on how to debug this? I have not seen anything
particuliar about IRQs and there is no message in syslog when stopping ups.

The fact is that I don't know at all what is blocking my network card and
even less why... :(

Thanks,
Antoine



-----Message d'origine-----
De : Charles Lepple [mailto:clepple at gmail.com] 
Envoyé : vendredi 2 octobre 2009 01:05
À : Antoine Gatineau
Cc : nut-upsuser at lists.alioth.debian.org
Objet : Re: [Nut-upsuser] stopping ups services resets my network

On Mon, Sep 28, 2009 at 6:40 AM, Antoine Gatineau
<antoine.gatineau at alcatel-lucent.com> wrote:
> Hello,
>
>
> I have this issue: ups is working fine for a time but when I stop the 
> services or if I unplug the ups device (over usb) I get the following 
> in /var/log/messages
>
> I am running nut-2.4.1 on RHEL4.
>
> Sep 28 14:44:57 mabolas upsd[24151]: Signal 15: exiting Sep 28 
> 14:45:06 mabolas upsmon[24156]: Signal 15: exiting Sep 28 15:04:10 
> mabolas sshd(pam_unix)[29190]: session closed for user root Sep 28 
> 15:04:36 mabolas kernel: NETDEV WATCHDOG: eth0: transmit timed out Sep 
> 28 15:04:36 mabolas kernel: tg3: eth0: transmit timed out, resetting 
> Sep 28 15:04:36 mabolas kernel: tg3: DEBUG: MAC_TX_STATUS[00000008] 
> MAC_RX_STATUS[00000000] Sep 28 15:04:36 mabolas kernel: tg3: DEBUG: 
> RDMAC_STATUS[00000000] WDMAC_STATUS[00000000] Sep 28 15:04:36 mabolas 
> kernel: tg3: eth0: Link is down.
> Sep 28 15:04:38 mabolas kernel: tg3: eth0: Link is up at 1000 Mbps, 
> full duplex.
> Sep 28 15:04:38 mabolas kernel: tg3: eth0: Flow control is off for TX 
> and off for RX.

This sounds like a bug in the tg3 driver in that kernel version. Have you
checked the RedHat bugzilla to see if anyone else sees this?

Depending on how the motherboard is laid out, moving the UPS to a different
USB port might avoid this problem. It could have something to do with
interrupt handling (for example, the tg3 driver expects an interrupt that is
stolen by the USB controller).

--
- Charles Lepple