Re: [PATCH net-next 7/7] nfp: flower vxlan neighbour keep-alive

2017-09-26 Thread Or Gerlitz 
On Tue, Sep 26, 2017 at 12:37 PM, John Hurley  wrote:
> [ Reposting in plantext only]
> On Mon, Sep 25, 2017 at 7:32 PM, Or Gerlitz  wrote:
>>
>> On Mon, Sep 25, 2017 at 1:23 PM, Simon Horman
>>  wrote:
>> > From: John Hurley 
>> >
>> > Periodically receive messages containing the destination IPs of tunnels
>> > that have recently forwarded traffic. Update the neighbour entries 'used'
>> > value for these IPs next hop.
>>
>> Are you proactively sending keep alive messages from the driver or the
>> fw? what's wrong with the probes sent by the kernel NUD subsystem?

> The messages are sent from the FW to the driver. They indicate which
> offloaded tunnels are currently active.

Do you support flow counters for offloaded TC rules? do you support last-use?

If Y && Y and you cache someone the prev counter value, you can use
this for the "used" feedback. I don't see why add keep-alive and not piggy back
on the flow counters logic.

>> In our driver we also update the used value for neighs of offloaded
>> tunnels, we do it based on flow counters for the offloaded tunnels
>> which is an evidence for activity. Any reason for you not to apply a
>> similar practice?

> Yes, this would provide the same outcome. Because our firmware already
> offered these messages, we chose to support this approach.

Re: [PATCH net-next 7/7] nfp: flower vxlan neighbour keep-alive

2017-09-26 Thread John Hurley 
[ Reposting in plantext only]



On Mon, Sep 25, 2017 at 7:32 PM, Or Gerlitz  wrote:
>
> On Mon, Sep 25, 2017 at 1:23 PM, Simon Horman
>  wrote:
> > From: John Hurley 
> >
> > Periodically receive messages containing the destination IPs of tunnels
> > that have recently forwarded traffic. Update the neighbour entries 'used'
> > value for these IPs next hop.
>
> Are you proactively sending keep alive messages from the driver or the
> fw? what's wrong with the probes sent by the kernel NUD subsystem?
>

Hi Or,

The messages are sent from the FW to the driver. They indicate which
offloaded tunnels are currently active.

>
> In our driver we also update the used value for neighs of offloaded
> tunnels, we do it based on flow counters for the offloaded tunnels
> which is an evidence for activity. Any reason for you not to apply a
> similar practice?


Yes, this would provide the same outcome. Because our firmware already
offered these messages, we chose to support this approach.

>
>
> Or.

Re: [PATCH net-next 7/7] nfp: flower vxlan neighbour keep-alive

2017-09-25 Thread Or Gerlitz 
On Mon, Sep 25, 2017 at 1:23 PM, Simon Horman
 wrote:
> From: John Hurley 
>
> Periodically receive messages containing the destination IPs of tunnels
> that have recently forwarded traffic. Update the neighbour entries 'used'
> value for these IPs next hop.

Are you proactively sending keep alive messages from the driver or the
fw? what's wrong with the probes sent by the kernel NUD subsystem?

In our driver we also update the used value for neighs of offloaded
tunnels, we do it based on flow counters for the offloaded tunnels
which is an evidence for activity. Any reason for you not to apply a
similar practice?

Or.