> > Hi Wei, > > > > > > > > Hi, Konstantin > > > I see your point. > > > I think your method can work. > > > But I think your method is a bit complex. As you method need to add more > > codes. > > > Anyway this is a simple bug. > > > How do you think now ? > > > > I still think it is better for all apps to handle this issue in a uniform > > way. > > Again in that case for NICs that do support PTYPE offloads the performance > > should be unaffected. > > Konstantin > > > > I have just had a quick look through the l3fwd and didn't find any codes to > check what ptypes capabilities ae provided by stuff below DPDK PMD & its base > driver. > L3fwd only check an input argument "--parse-ptype" to enable ptype check > implemented in > a Rx callback function.
As an example for lpm: http://dpdk.org/browse/dpdk/tree/examples/l3fwd/l3fwd_lpm.c#n279 Konstantin > In this l3fwd rx callback function, it has done the same thing as my code. > Anyway, I'd like to provide a v2 patch to deal with this issue in a uniform > way. > > Thanks & Best Regards > -Wei
