[ofa-general] [PATCH RFCv2 1/2] RDMA: Add protocol statistics attributes to the RDMA sysfs.

[Steve Wise]

Caitlin Bestler wrote:
> On Thu, Jun 19, 2008 at 8:16 AM, Roland Dreier <rdreier at cisco.com> wrote:
>   
>>  > The "Retransmission count" is per netdev and we support multiple netdev
>>  > per physical port.  We have designed our silicon to support stats at
>>  > netdev level and it doesn't make sense for us to aggregate stats from
>>  > all netdevs and report at a device level.  Perhaps we can compromise,
>>  > Chelsio can give device level stats at /sys/class/infiniband/$device and
>>  > for NetEffect, we can provide netdev level stats below it?
>>
>> Hmm, kind of a mess if we can't achieve some uniformity...
>>
>>     
>
> There's an even tougher potential issue on the "uniformity" issue than
> whether statistics are talled per port or per device -- and that is whether
> they are gathered separately for offload services and the default stack
> or not.
>
> I can envision designs where the statistics for the offload services
> would have to be gathered separately, and those where they would
> have to be gathered for all the traffic (no matter which L4/L5 service
> was being used).
>
> One way to deal with that might be to define the "iwarp statistics"
> as being those for the interface that were *not* reported to the
> general purpose stack. Or it might be a flag indicating whether
> the two sets of statistics were intermingled or  not.
>   

The L3 stats for cxgb3 are for both IP packets going to the native stack 
-and- the TCP offload engine.  The L4 stats are just for the TCP offload 
engine.

But I'm not sure how much we care about this.  As long as the consumer 
knows what they mean.  And I don't believe the consumer needs an API to 
understand this (vs just release notes per device).


Steve.