2016-05-12 14:10 GMT-07:00 Roopa Prabhu <ro...@cumulusnetworks.com>: > On 5/12/16, 4:48 AM, Jiri Pirko wrote: >> From: Nogah Frankel <nog...@mellanox.com> >> >> Till now we had a ndo statistics function that returned SW statistics. >> We want to change the "basic" statistics to return HW statistics if >> available. >> In this case we need to expose a new ndo to return the SW statistics. >> Add a new ndo declaration to get SW statistics >> Add a function that gets SW statistics if a competible ndo exist >> >> Signed-off-by: Nogah Frankel <nog...@mellanox.com> >> Reviewed-by: Ido Schimmel <ido...@mellanox.com> >> Signed-off-by: Jiri Pirko <j...@mellanox.com> >> --- >> > To me netdev stats is combined 'SW + HW' stats for that netdev. > ndo_get_stats64 callback into the drivers does the magic of adding HW stats > to SW (netdev) stats and returning (see enic_get_stats). HW stats is > available for netdevs > that are offloaded or are backed by hardware. SW stats is the stats that the > driver maintains > (logical or physical). HW stats is queried and added to the SW stats. > In fact, we represent all our offloaded netdev stats following this model. > This fits well for logical devices like vlan/bond/vxlan which are hardware > offloaded too.
Agreed, if you take a look at what the DSA slave network devices do in net/dsa/slave.c they basically overload the switch-mainainted statistics with some per-port net_device software stats. We would probably want to do the same thing > > There is HW extended stats which today are represented by ethtool > and I had an example of a new nested attribute like > IFLA_STATS_LINK_HW_EXTENDED > to provide ethtool like extensible stats. If people want pure hardware stats, > they > get it via this. In which case I am not seeing the point in adding a new ndo > and a new > stats attribute just for this. If you are looking for pure hardware stats, > any reason we cant > just use the IFLA_STATS_LINK_HW_EXTENDED and add ethtool like hw stats ?. > This will be one place to look for all hardware stats and will be extensible > too. Agreed, worste case, you format your statistics such that they reflect they come from the HW (e.g: hw_TxOctets...) -- Florian
