> BTW, are we considering some offload where the eBPF would return a set of > changes [based on some pre-set capabilities set by driver] done on buffer and > let > the HW offload those? > > I understand end goal is eBPF hw-offloading, but seems like there are a lot of > existing offload facilities that might be leveraged.
On second thought, given that the program is likely to change only a small set of fields [as opposed to building whole new buffers], probably fixing the csum isn't going to be much costlier [if at all] then the logic that would be required to determine whether the device would be able to offload it.