> > > On 6 Oct 2021, at 11:32, Lorenzo Bianconi wrote: > > >> On Fri, 1 Oct 2021 11:03:58 +0200 Lorenzo Bianconi wrote: > >>> Can you please check if the code above is aligned to current requirements or if > >>> it is missing something? > >>> If this code it is fine, I guess we have two option here: > >>> - integrate the commits above in xdp multi-buff series (posting v15) and work on > >>> the verfier code in parallel (if xdp_mb_pointer helper is not required from day0) > >>> - integrate verfier changes in xdp multi-buff series, drop bpf_xdp_load_bytes > >>> helper (probably we will still need bpf_xdp_store_bytes) and introduce > >>> bpf_xdp_pointer as new ebpf helper. > >> > >> It wasn't clear to me that we wanted bpf_xdp_load_bytes() to exist. > >> But FWIW no preference here. > >> > > > > ack, same here. Any other opinion about it? > > I was under the impression getting a pointer might be enough. But playing with the bpf ring buffers for a bit, it still might be handy to extract some data to be sent to userspace. So I would not mind keeping it. > ok, so it seems we have a use-case for bpf_xdp_load_bytes(). If everybody agree, I will post v15 with them included and we then we can work in parallel for a bpf_xdp_pointer ebpf helper. Regards, Lorenzo
Attachment:
signature.asc
Description: PGP signature
