From mboxrd@z Thu Jan 1 00:00:00 1970 From: Eric Dumazet Subject: Re: [PATCH] bnx2x: fix rx checksum validation for IPv6 Date: Fri, 14 Sep 2012 01:14:39 +0200 Message-ID: <1347578079.8555.141.camel@edumazet-glaptop> References: <1347577184-8417-1-git-send-email-mschmidt@redhat.com> Mime-Version: 1.0 Content-Type: text/plain; charset="UTF-8" Content-Transfer-Encoding: 7bit Cc: netdev@vger.kernel.org, Eilon Greenstein , Eric Dumazet , Yaniv Rosner , Yuval Mintz , Merav Sicron , Robert Evans , Tom Herbert , Willem de Bruijn , David Miller , Havard Skinnemoen To: Michal Schmidt Return-path: Received: from mail-we0-f174.google.com ([74.125.82.174]:52837 "EHLO mail-we0-f174.google.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1750958Ab2IMXOp (ORCPT ); Thu, 13 Sep 2012 19:14:45 -0400 Received: by weyx8 with SMTP id x8so2018775wey.19 for ; Thu, 13 Sep 2012 16:14:44 -0700 (PDT) In-Reply-To: <1347577184-8417-1-git-send-email-mschmidt@redhat.com> Sender: netdev-owner@vger.kernel.org List-ID: On Fri, 2012-09-14 at 00:59 +0200, Michal Schmidt wrote: > Commit d6cb3e41 "bnx2x: fix checksum validation" caused a performance > regression for IPv6. Rx checksum offload does not work. IPv6 packets > are passed to the stack with CHECKSUM_NONE. > > The hardware obviously cannot perform IP checksum validation for IPv6, > because there is no checksum in the IPv6 header. This should not prevent > us from setting CHECKSUM_UNNECESSARY. > > Tested on BCM57711. > > Signed-off-by: Michal Schmidt > --- > drivers/net/ethernet/broadcom/bnx2x/bnx2x_cmn.c | 12 +++++++----- > 1 file changed, 7 insertions(+), 5 deletions(-) > > diff --git a/drivers/net/ethernet/broadcom/bnx2x/bnx2x_cmn.c b/drivers/net/ethernet/broadcom/bnx2x/bnx2x_cmn.c > index af20c6e..e8e97a7 100644 > --- a/drivers/net/ethernet/broadcom/bnx2x/bnx2x_cmn.c > +++ b/drivers/net/ethernet/broadcom/bnx2x/bnx2x_cmn.c > @@ -662,14 +662,16 @@ void bnx2x_csum_validate(struct sk_buff *skb, union eth_rx_cqe *cqe, > struct bnx2x_fastpath *fp, > struct bnx2x_eth_q_stats *qstats) > { > - /* Do nothing if no IP/L4 csum validation was done */ > - > + /* Do nothing if no L4 csum validation was done. > + * We do not check whether IP csum was validated. For IPv4 we assume > + * that if the card got as far as validating the L4 csum, it also > + * validated the IP csum. IPv6 has no IP csum. > + */ > if (cqe->fast_path_cqe.status_flags & > - (ETH_FAST_PATH_RX_CQE_IP_XSUM_NO_VALIDATION_FLG | > - ETH_FAST_PATH_RX_CQE_L4_XSUM_NO_VALIDATION_FLG)) > + ETH_FAST_PATH_RX_CQE_L4_XSUM_NO_VALIDATION_FLG) > return; > > - /* If both IP/L4 validation were done, check if an error was found. */ > + /* If L4 validation was done, check if an error was found. */ > > if (cqe->fast_path_cqe.type_error_flags & > (ETH_FAST_PATH_RX_CQE_IP_BAD_XSUM_FLG | Thanks for fixing this bug ! Acked-by: Eric Dumazet