Re: DHCP/network issues

Am Sat, 29 Apr 2017 19:07:06 -0400 (EDT)
AN <andy_at_neu.net> schrieb:

> Hi OH:
> 
> On Sat, 18 Mar 2017, O. Hartmann wrote:
> 
> > Date: Sat, 18 Mar 2017 19:11:42 +0100
> > From: O. Hartmann <ohartmann_at_walstatt.org>
> > To: AN <andy_at_neu.net>
> > Cc: O. Hartmann <ohartmann_at_walstatt.org>
> > Subject: Re: DHCP/network issues
> > 
> > Am Sat, 18 Mar 2017 10:25:58 -0400 (EDT)
> > AN <andy_at_neu.net> schrieb:
> >  
> >> Hi:
> >>
> >> On Sat, 18 Mar 2017, O. Hartmann wrote:
> >>  
> >>> Date: Sat, 18 Mar 2017 15:20:13 +0100
> >>> From: O. Hartmann <ohartmann_at_walstatt.org>
> >>> To: AN <andy_at_neu.net>
> >>> Cc: freebsd-current_at_freebsd.org
> >>> Subject: Re: DHCP/network issues
> >>>
> >>> Am Sat, 18 Mar 2017 09:45:31 -0400 (EDT)
> >>> AN <andy_at_neu.net> schrieb:
> >>>  
> >>>> Is anyone seeing network or DHCP issues with a recent update to
> >>>> 12-current?
> >>>>
> >>>> On new hardware
> >>>> Dual Core Celeron J1800 Bay Trail 2.4GHz, 2MB L2 Cache
> >>>> 2 Gigabit Ethernet Intel NIC ports
> >>>>
> >>>> I'm seeing the following issues:
> >>>>
> >>>> - install latest 12-current snapshot
> >>>> FreeBSD-12.0-CURRENT-amd64-20170316-r315413-memstick.img, try DHCP during
> >>>> install and it fails.  Setup networking manually and try to ping default gateway
> >>>> results in the response
> >>>> ping: sendto: Host is down  
> >>>
> >>> Yes. Since IFLIB has been introduced, Intel NICs seem to be suffering from not
> >>> working properly anymore. i217-LM, i350 and sibblings are affected for which I can
> >>> speak because we suffer the same problems here. Even if the NIC works, it dies
> >>> after some time under heavy I/O.
> >>>  
> >>
> >> Thank you for the reply, I remember seeing your messages on the list about
> >> this now.  Is anyone trying to fix it, did you file a PR?  
> >
> > I think, as you may already have read, some of the developers are aware of this
> > problem. I haven't file a PR so far.
> >
> > Kind regards,
> >
> > oh
> >
> >  
> 
> 
> Do you know if the problem with Intel Nic in 12-current has been fixed?
> 
> Thanks,
> 
> AN

Hello.

The problem still persists!! I have a Fujitsu Celsius M740, which uses a i217 NIC. This
NIC freezes up while under heavy I/O.
In most cases, I can get it back online by bringing it first down and then up again with
ifconfig em0 down/up. But sometimes I have to reboot. I get some strange errors on the
console about "RX/TX no buffer left" - I'm sorry to be not more specific, I'm not in the
lab at the moment where I have a more specific record of that problem. The i217 NIC
suffers from this problem after IFLIB has been introduced and I can trigger the problem
easily whily rsyncing poudriere repository packages with the NFSv4 connected package
server.

I haven't checked on igb NICs (we also use i350 based dual-port NICs as I do with some
private servers). But last time I pushed them very hard with I/O, I was able to freeze
them the same way.

I do not know whether someone is working on this, I recall having filed a PR on this, but
I was very unspecific, because I do not have further details and running CURRENT with no
DEBUG features.

As of your question: it hasn't been fixed so far.

Kind regards,

Oliver

-- 
O. Hartmann

Ich widerspreche der Nutzung oder Übermittlung meiner Daten für
Werbezwecke oder für die Markt- oder Meinungsforschung (§ 28 Abs. 4 BDSG).