From mboxrd@z Thu Jan 1 00:00:00 1970 From: Rick Jones Subject: Re: [PATCH] cxgb3: Set vlan_feature on net_device Date: Wed, 18 Jul 2012 13:12:41 -0700 Message-ID: <500718B9.8000508@hp.com> References: <1342639748-16276-1-git-send-email-brenohl@br.ibm.com> Mime-Version: 1.0 Content-Type: text/plain; charset=ISO-8859-1; format=flowed Content-Transfer-Encoding: 7bit Cc: "divy@chelsio.com" , "netdev@vger.kernel.org" To: "brenohl@br.ibm.com" Return-path: Received: from g4t0014.houston.hp.com ([15.201.24.17]:27124 "EHLO g4t0014.houston.hp.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1753204Ab2GRUMo (ORCPT ); Wed, 18 Jul 2012 16:12:44 -0400 In-Reply-To: <1342639748-16276-1-git-send-email-brenohl@br.ibm.com> Sender: netdev-owner@vger.kernel.org List-ID: On 07/18/2012 12:29 PM, brenohl@br.ibm.com wrote: > cxgb3 interface has a bad performance when VLAN is set. On my current > setup, a PowerLinux 7R2, I am able to get around 7 Gbps on a TCP_STREAM > (8 instances, 4k message). > With this patch, I am able to reach 9.5 Gbps. Getting service demand out of an aggregate netperf test is a chore, but reporting the change in CPU utilization should be pretty straightforward. Since you ended-up being constrained by link-rate, showing the CPU utilization change (and calculating service demand manually if you feel up to it) may help show the change has an even greater effect then (9.5-7)/7 or 35%. What does the change do for latency and/or maximum, min-sized packets per second. rick jones there is more to the network than just bits/s :) > > Signed-off-by: Breno Leitao > > diff --git a/drivers/net/ethernet/chelsio/cxgb3/cxgb3_main.c b/drivers/net/ethernet/chelsio/cxgb3/cxgb3_main.c > index abb6ce7..fcf4b31 100644 > --- a/drivers/net/ethernet/chelsio/cxgb3/cxgb3_main.c > +++ b/drivers/net/ethernet/chelsio/cxgb3/cxgb3_main.c > @@ -3173,6 +3173,9 @@ static void __devinit cxgb3_init_iscsi_mac(struct net_device *dev) > pi->iscsic.mac_addr[3] |= 0x80; > } > > +#define TSO_FLAGS (NETIF_F_TSO | NETIF_F_TSO6 | NETIF_F_TSO_ECN) > +#define VLAN_FEAT (NETIF_F_SG | NETIF_F_IP_CSUM | TSO_FLAGS | \ > + NETIF_F_IPV6_CSUM | NETIF_F_HIGHDMA) > static int __devinit init_one(struct pci_dev *pdev, > const struct pci_device_id *ent) > { > @@ -3293,6 +3296,7 @@ static int __devinit init_one(struct pci_dev *pdev, > netdev->hw_features = NETIF_F_SG | NETIF_F_IP_CSUM | > NETIF_F_TSO | NETIF_F_RXCSUM | NETIF_F_HW_VLAN_RX; > netdev->features |= netdev->hw_features | NETIF_F_HW_VLAN_TX; > + netdev->vlan_features |= netdev->features & VLAN_FEAT; > if (pci_using_dac) > netdev->features |= NETIF_F_HIGHDMA; >