From: Willem de Bruijn Subject: [PATCH net-next 0/4] zerocopy refinements Date: Fri, 22 Dec 2017 19:00:16 -0500 Message-ID: <20171223000020.55509-1-willemdebruijn.kernel@gmail.com> Cc: davem@davemloft.net, Willem de Bruijn To: netdev@vger.kernel.org Return-path: Received: from mail-yw0-f194.google.com ([209.85.161.194]:45725 "EHLO mail-yw0-f194.google.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1756868AbdLWAAY (ORCPT ); Fri, 22 Dec 2017 19:00:24 -0500 Received: by mail-yw0-f194.google.com with SMTP id y187so4634578ywd.12 for ; Fri, 22 Dec 2017 16:00:24 -0800 (PST) Sender: netdev-owner@vger.kernel.org List-ID: From: Willem de Bruijn 1/4 is a small optimization follow-up to the earlier fix to skb_segment: check skb state once per skb, instead of once per frag. 2/4 makes behavior more consistent between standard and zerocopy send: set the PSH bit when hitting MAX_SKB_FRAGS. This helps GRO. 3/4 resolves a surprising inconsistency in notification: because small packets were not stored in frags, they would not set the copied error code over loopback. This change also optimizes the path by removing copying and making tso_fragment cheaper. 4/4 follows-up to 3/4 by no longer allocated now unused memory. this was actually already in RFC patches, but dropped as I pared down the patch set during revisions. Willem de Bruijn (4): skbuff: in skb_segment, call zerocopy functions once per nskb tcp: push full zerocopy packets tcp: place all zerocopy payload in frags tcp: do not allocate linear memory for zerocopy skbs net/core/skbuff.c | 14 +++++++++----- net/ipv4/tcp.c | 24 +++++++++++++++--------- 2 files changed, 24 insertions(+), 14 deletions(-) -- 2.15.1.620.gb9897f4670-goog