[ovs-dev] [PATCH] Fix openvswitch: netlink: Key attribute has unexpected length (type=62, length=4, expected=0) messages.

[Jesse Gross]

Yes, that's right. At the moment, the error is completely normal and
harmless - it's just trying to detect a feature that doesn't exist.

Jarno was talking about exposing kernel capabilities at one point
though, so this type of probing (and the error message) doesn't need
to happen.

On Wed, May 28, 2014 at 5:23 PM, Madhu Challa <challa at noironetworks.com> wrote:
> Oh ok. So I should not see this error once the mpls probe functionality is
> implemented in the kernel. Is my understanding correct ? Right now we always
> come out of the loop in check_max_mpls_depth() for n = 0.
>
> Thanks.
>
>
> On Wed, May 28, 2014 at 5:08 PM, Jesse Gross <jesse at nicira.com> wrote:
>>
>> On Wed, May 28, 2014 at 11:48 AM, Madhu Challa <challa at noironetworks.com>
>> wrote:
>> > These come from dpif_flow_put() in check_max_mpls_depth() when adding a
>> > datapath. The key length is variable depending on number of MPLS labels.
>> > By
>> > initializing to -1 kmod will not check for the key length.
>> > Signed-off-by: Madhu Challa <challa at noironetworks.com>
>>
>> Right but the purpose of this operation is to determine the number
>> labels that the kernel supports. Silently allowing all values is not
>> particularly useful.
>
>