Regarding bug 82471

Regarding bug 82471

[Laurent CARON]

Hi,

I've been biten by : https://bugzilla.kernel.org/show_bug.cgi?id=82471 
today.

I tried various kernels from 3.2.0 to 3.16.3

The last unaffected kernel according to my tests is 3.14 (didn't yet try 
3.15).

Please see logs when (on 3.16.2)
1/ using iptables
2/ having some traffic (even light) going through

http://pastebin.com/JWkFu2w2

My network config is the following:

a bond interface consisting of 2 10Gb ports (intel ixgbe / X520-SR2) on 
top of which I use 2 vlans.

Thanks for your help,

Laurent

Re: Regarding bug 82471

[Michal Kubecek]

On Fri, Sep 26, 2014 at 06:35:34PM +0200, Laurent CARON wrote:
> Hi,
> 
> I've been biten by :
> https://bugzilla.kernel.org/show_bug.cgi?id=82471 today.
> 
> I tried various kernels from 3.2.0 to 3.16.3
> 
> The last unaffected kernel according to my tests is 3.14 (didn't yet
> try 3.15).
> 
> Please see logs when (on 3.16.2)
> 1/ using iptables
> 2/ having some traffic (even light) going through
> 
> http://pastebin.com/JWkFu2w2
> 
> My network config is the following:
> 
> a bond interface consisting of 2 10Gb ports (intel ixgbe / X520-SR2)
> on top of which I use 2 vlans.

I don't see why it would affect ixgbe adapters which support Tx VLAN
offloading (and with which the original patch was tested) but the
symptoms look very much like those fixed by commit db115037bb57 ("net:
fix checksum features handling in netif_skb_features()"). Could you
please test with 3.17-rc5 or later?

                                                      Michal Kubeček

Re: Regarding bug 82471

[Laurent CARON]

On 29/09/2014 07:51, Michal Kubecek wrote:
> I don't see why it would affect ixgbe adapters which support Tx VLAN
> offloading (and with which the original patch was tested) but the
> symptoms look very much like those fixed by commit db115037bb57 ("net:
> fix checksum features handling in netif_skb_features()"). Could you
> please test with 3.17-rc5 or later?

Hi Michal,

Just compiled 3.17-rc7 and it seems it is fixed. Will report later on on 
this, but since the issue appears very shortly after server startup, we 
should be all set.

Thanks

Laurent