Parsing is part of the classification functions of ODP. Exactly what parts of that will be part of ODP v1.0 will be discussed along with the rest of the v1.0 final API list during next Tuesday's ODP call. Anything not in v1.0 will be deferred for consideration for inclusion in a point release as 2015 progresses.
For most implementations, packet parsing and classification is provided in HW, so there is no performance penalty as these functions are done inline by HW at line rate as packets are received. Obviously SW is free to ignore the fields it is not interested in, and the purpose of the classification pattern matching rules (PMRs) is to specify what fields and terms SW is in interested in to the underlying implementation. So the PMR calls that SW makes in setting up classification rules is how applications configure this. See the classification API doc for details. Bill On Fri, Nov 14, 2014 at 3:51 PM, Shmulik Ladkani <[email protected]> wrote: > Hi, > > I noticed linux-generic does some parsing upon ingress, and updates > 'input_flags' with various L2/L3/L4 details. > > 1. > Is this part of ODP API? > Meaning, can I expect to have same parsing done in other implementations? > > 2. > What if the application does NOT require deep parsing? > For example, app is interested only in L2 details, thus wishes to avoid > the penalty associated with L3/L4 parsing. Is it user configurable? > > Regards, > Shmulik > > _______________________________________________ > lng-odp mailing list > [email protected] > http://lists.linaro.org/mailman/listinfo/lng-odp >
_______________________________________________ lng-odp mailing list [email protected] http://lists.linaro.org/mailman/listinfo/lng-odp
