From mboxrd@z Thu Jan 1 00:00:00 1970 From: Eric Dumazet Subject: Re: [PATCH][net-next] gianfar: Fix reported sent bytes to BQL for Tx timestamping Date: Mon, 29 Apr 2013 06:53:13 -0700 Message-ID: <1367243593.8964.306.camel@edumazet-glaptop> References: <1367240268-32609-1-git-send-email-claudiu.manoil@freescale.com> Mime-Version: 1.0 Content-Type: text/plain; charset="UTF-8" Content-Transfer-Encoding: 7bit Cc: netdev@vger.kernel.org, Paul Gortmaker , "David S. Miller" To: Claudiu Manoil Return-path: Received: from mail-pa0-f42.google.com ([209.85.220.42]:41603 "EHLO mail-pa0-f42.google.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1752038Ab3D2NxQ (ORCPT ); Mon, 29 Apr 2013 09:53:16 -0400 Received: by mail-pa0-f42.google.com with SMTP id kl13so3685466pab.1 for ; Mon, 29 Apr 2013 06:53:16 -0700 (PDT) In-Reply-To: <1367240268-32609-1-git-send-email-claudiu.manoil@freescale.com> Sender: netdev-owner@vger.kernel.org List-ID: On Mon, 2013-04-29 at 15:57 +0300, Claudiu Manoil wrote: > When Tx timestamping is enabled the number of sent bytes reported > to BQL via tx_completed_queue() falls short by GMAC_FCB_LEN + > GMAC_TXPAL_LEN than the number of bytes reported via tx_sent_queue() > on xmit. This leads to BQL stopping transmission errorneously followed > by tx timeout firing. > > This fixes the amount of sent bytes reported to BQL on clean_tx_ring > to match the amount reported on xmit, when Tx timestamping enabled. > > Signed-off-by: Claudiu Manoil > --- > drivers/net/ethernet/freescale/gianfar.c | 1 + > 1 file changed, 1 insertion(+) > > diff --git a/drivers/net/ethernet/freescale/gianfar.c b/drivers/net/ethernet/freescale/gianfar.c > index 2375a01..0d26a8b 100644 > --- a/drivers/net/ethernet/freescale/gianfar.c > +++ b/drivers/net/ethernet/freescale/gianfar.c > @@ -2539,6 +2539,7 @@ static void gfar_clean_tx_ring(struct gfar_priv_tx_q *tx_queue) > skb_tstamp_tx(skb, &shhwtstamps); > bdp->lstatus &= BD_LFLAG(TXBD_WRAP); > bdp = next; > + bytes_sent += GMAC_FCB_LEN + GMAC_TXPAL_LEN; > } > > bdp->lstatus &= BD_LFLAG(TXBD_WRAP); Technically speaking these bytes are not sent to the wire. I would rather fix gfar_start_xmit() to give to netdev_tx_sent_queue() call the correct amount of bytes on wire, to be consistent with other drivers.