Hi Ed, On Tue, Oct 29, 2019 at 06:02:57PM +0000, Edward Cree wrote: > Hi Pablo, > > I've been trying to figure out how the kernel API is going to look for > offloading conntrack entries into hardware (for ct_state matches in flower > rules to match on), as we're developing some hardware that intends to > support that. AFAICT such an API does not yet exist in nf-next. > > Back in 2017 you added some relevant-looking nf code [1] "Flow offload > infrastructure", including an RFC patch [2] which you said you would "keep > back until there's a driver". Is the API added in that patch (i.e. > ndo_flow_offload()) still the current plan, or has the design changed at > all over the course of the last two years? > > If you have some more up-to-date strawman/RFC code, that'd be really useful > for me to develop a prototype against. The most up to date code is available here: https://git.kernel.org/pub/scm/linux/kernel/git/pablo/nf.git/log/?h=flow-offload-hw-v3 > I am assuming that the first actual upstream driver supporting conntrack > offload will be something from Mellanox (they seem to have been driving > this process) and that the offload API will be submitted along with that; > is that assumption accurate? I don't work for Mellanox, I have absolutely no idea what the plans are. I received a few emails about how they might use the infrastructure I made, to consolidate common infrastructure. So I'm very much looking for to seeing patches like you are. > The information contained in this message is confidential and is > intended for the addressee(s) only. If you have received this > message in error, please notify the sender immediately and delete > the message. Unless you are an addressee (or authorized to receive > for an addressee), you may not use, copy or disclose to anyone this > message or any information contained in this message. The > unauthorized use, disclosure, copying or alteration of this message > is strictly prohibited. Don't worry, I'll keep this confidential :-)
