From mboxrd@z Thu Jan 1 00:00:00 1970 From: Jakub Kicinski Subject: [RFC 00/12] BPF hardware offload via cls_bpf Date: Wed, 1 Jun 2016 17:50:02 +0100 Message-ID: <1464799814-4453-1-git-send-email-jakub.kicinski@netronome.com> Cc: ast@kernel.org, daniel@iogearbox.net, dinan.gunawardena@netronome.com, Jakub Kicinski To: netdev@vger.kernel.org Return-path: Received: from mail-wm0-f42.google.com ([74.125.82.42]:34421 "EHLO mail-wm0-f42.google.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1752157AbcFAQvn (ORCPT ); Wed, 1 Jun 2016 12:51:43 -0400 Received: by mail-wm0-f42.google.com with SMTP id z87so29624003wmh.1 for ; Wed, 01 Jun 2016 09:51:42 -0700 (PDT) Sender: netdev-owner@vger.kernel.org List-ID: Hi! In recent months a lot of progress have been made on offloading simpler TC classifiers. There is also growing interest in using BPF for generic high-speed packet processing in the kernel. It seems beneficial to tie those two trends together and think about hardware offloads of BPF programs. This patch set presents such offload to Netronome smart NICs. cls_bpf is extended with hardware offload capabilities and NFP driver gets a JIT translator which in presence of capable firmware can be used to offload the BPF program onto the card. BPF JIT implementation is not 100% complete (e.g. missing instructions) but it is functional - in fact I implemented only bits which clang/cBPF used in my test programs. Encouragingly it should be possible to offload most (if not all) advanced BPF features onto the NIC - including packet modification, maps, tunnel encap/decap etc. Examples of tests I used: Basic eBPF/C/clang apps: __section_cls_entry int cls_entry(struct __sk_buff *skb) { if (load_byte(skb, 0) != 0x0) return 0; if (load_byte(skb, 4) != 0x1) return 0; skb->mark = 0xcafe; if (load_byte(skb, 50) != 0xff) return 0; return ~0U; } tcpdump generated filters, for instance: dst 10.1.255.255 and \ tcp and \ (port 90 or port 91) and \ tcp[tcpflags] & tcp-syn != 0 First patch is not really related but others depend on it, I hope to post it separately soon. Jakub Kicinski (12): add basic register-field manipulation macros net: cls_bpf: add hardware offload net: cls_bpf: limit hardware offload by software-only flag net: cls_bpf: add support for marking filters as hardware-only nfp: add BPF to NFP code translator nfp: add hardware cls_bpf offload nfp: add skb mark support to the bpf offload net: cls_bpf: allow offloaded filters to update stats nfp: report statistics of offloaded filters nfp: bpf: optimize register init nfp: bpf: add register rename nfp: bpf: add denser mode of execution drivers/net/ethernet/netronome/nfp/Makefile | 4 +- drivers/net/ethernet/netronome/nfp/nfp_asm.h | 191 ++++ drivers/net/ethernet/netronome/nfp/nfp_bpf.h | 136 +++ drivers/net/ethernet/netronome/nfp/nfp_bpf_jit.c | 1027 ++++++++++++++++++++ drivers/net/ethernet/netronome/nfp/nfp_net.h | 34 +- .../net/ethernet/netronome/nfp/nfp_net_common.c | 41 +- drivers/net/ethernet/netronome/nfp/nfp_net_ctrl.h | 21 +- .../net/ethernet/netronome/nfp/nfp_net_ethtool.c | 3 + .../net/ethernet/netronome/nfp/nfp_net_offload.c | 256 +++++ include/linux/bitfield.h | 89 ++ include/linux/netdevice.h | 2 + include/net/pkt_cls.h | 16 + include/uapi/linux/pkt_cls.h | 1 + net/sched/cls_bpf.c | 115 ++- 14 files changed, 1928 insertions(+), 8 deletions(-) create mode 100644 drivers/net/ethernet/netronome/nfp/nfp_asm.h create mode 100644 drivers/net/ethernet/netronome/nfp/nfp_bpf.h create mode 100644 drivers/net/ethernet/netronome/nfp/nfp_bpf_jit.c create mode 100644 drivers/net/ethernet/netronome/nfp/nfp_net_offload.c create mode 100644 include/linux/bitfield.h -- 1.9.1