From: Florian Westphal Subject: Re: [PATCH net-next] tg3: Add skb->xmit_more support Date: Mon, 13 Oct 2014 19:38:06 +0200 Message-ID: <20141013173806.GB26105@breakpoint.cc> References: <1413217302-15396-1-git-send-email-prashant@broadcom.com> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Cc: davem@davemloft.net, netdev@vger.kernel.org, dborkman@redhat.com, mchan@broadcom.com To: Prashant Sreedharan Return-path: Received: from Chamillionaire.breakpoint.cc ([80.244.247.6]:59054 "EHLO Chamillionaire.breakpoint.cc" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1752634AbaJMRiL (ORCPT ); Mon, 13 Oct 2014 13:38:11 -0400 Content-Disposition: inline In-Reply-To: <1413217302-15396-1-git-send-email-prashant@broadcom.com> Sender: netdev-owner@vger.kernel.org List-ID: Prashant Sreedharan wrote: > Ring TX doorbell only if xmit_more is not set or the queue is stopped. > > Suggested-by: Daniel Borkmann > Signed-off-by: Prashant Sreedharan > Signed-off-by: Michael Chan > --- > drivers/net/ethernet/broadcom/tg3.c | 10 ++++++---- > 1 files changed, 6 insertions(+), 4 deletions(-) > > diff --git a/drivers/net/ethernet/broadcom/tg3.c b/drivers/net/ethernet/broadcom/tg3.c > index ba49948..dbb41c1 100644 > --- a/drivers/net/ethernet/broadcom/tg3.c > +++ b/drivers/net/ethernet/broadcom/tg3.c > @@ -8099,9 +8099,6 @@ static netdev_tx_t tg3_start_xmit(struct sk_buff *skb, struct net_device *dev) > /* Sync BD data before updating mailbox */ > wmb(); > > - /* Packets are ready, update Tx producer idx local and on card. */ > - tw32_tx_mbox(tnapi->prodmbox, entry); > - > tnapi->tx_prod = entry; > if (unlikely(tg3_tx_avail(tnapi) <= (MAX_SKB_FRAGS + 1))) { > netif_tx_stop_queue(txq); > @@ -8116,7 +8113,12 @@ static netdev_tx_t tg3_start_xmit(struct sk_buff *skb, struct net_device *dev) > netif_tx_wake_queue(txq); > } > > - mmiowb(); > + if (!skb->xmit_more || netif_xmit_stopped(txq)) { > + /* Packets are ready, update Tx producer idx on card. */ > + tw32_tx_mbox(tnapi->prodmbox, entry); I think you need to swap the test, i.e. netif_xmit_stopped() || xmit_more Otherwise, hw may not be aware of further packets wainting and queue might not be restarted?