From patchwork Fri Jan 25 09:51:29 2013 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Wanlong Gao X-Patchwork-Id: 215605 X-Patchwork-Delegate: davem@davemloft.net Return-Path: X-Original-To: patchwork-incoming@ozlabs.org Delivered-To: patchwork-incoming@ozlabs.org Received: from vger.kernel.org (vger.kernel.org [209.132.180.67]) by ozlabs.org (Postfix) with ESMTP id D94F92C008C for ; Fri, 25 Jan 2013 20:51:51 +1100 (EST) Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1755598Ab3AYJvs (ORCPT ); Fri, 25 Jan 2013 04:51:48 -0500 Received: from cn.fujitsu.com ([222.73.24.84]:3951 "EHLO song.cn.fujitsu.com" rhost-flags-OK-FAIL-OK-OK) by vger.kernel.org with ESMTP id S1752973Ab3AYJvo (ORCPT ); Fri, 25 Jan 2013 04:51:44 -0500 X-IronPort-AV: E=Sophos;i="4.84,537,1355068800"; d="scan'208";a="6646252" Received: from unknown (HELO tang.cn.fujitsu.com) ([10.167.250.3]) by song.cn.fujitsu.com with ESMTP; 25 Jan 2013 17:49:33 +0800 Received: from fnstmail02.fnst.cn.fujitsu.com (tang.cn.fujitsu.com [127.0.0.1]) by tang.cn.fujitsu.com (8.14.3/8.13.1) with ESMTP id r0P9peBU006663; Fri, 25 Jan 2013 17:51:40 +0800 Received: from gaowanlong.fnst.cn.fujitsu.com ([10.167.225.197]) by fnstmail02.fnst.cn.fujitsu.com (Lotus Domino Release 8.5.3) with ESMTP id 2013012517504189-61717 ; Fri, 25 Jan 2013 17:50:41 +0800 From: Wanlong Gao To: linux-kernel@vger.kernel.org Cc: Rusty Russell , "Michael S. Tsirkin" , Jason Wang , Eric Dumazet , "David S. Miller" , virtualization@lists.linux-foundation.org, netdev@vger.kernel.org, Wanlong Gao Subject: [PATCH V8 1/3] virtio-net: fix the set affinity bug when CPU IDs are not consecutive Date: Fri, 25 Jan 2013 17:51:29 +0800 Message-Id: <1359107491-6749-1-git-send-email-gaowanlong@cn.fujitsu.com> X-Mailer: git-send-email 1.8.1 X-MIMETrack: Itemize by SMTP Server on mailserver/fnst(Release 8.5.3|September 15, 2011) at 2013/01/25 17:50:42, Serialize by Router on mailserver/fnst(Release 8.5.3|September 15, 2011) at 2013/01/25 17:50:42, Serialize complete at 2013/01/25 17:50:42 Sender: netdev-owner@vger.kernel.org Precedence: bulk List-ID: X-Mailing-List: netdev@vger.kernel.org As Michael mentioned, set affinity and select queue will not work very well when CPU IDs are not consecutive, this can happen with hot unplug. Fix this bug by traversal the online CPUs, and create a per cpu variable to find the mapping from CPU to the preferable virtual-queue. Cc: Rusty Russell Cc: "Michael S. Tsirkin" Cc: Jason Wang Cc: Eric Dumazet Cc: "David S. Miller" Cc: virtualization@lists.linux-foundation.org Cc: netdev@vger.kernel.org Signed-off-by: Wanlong Gao Acked-by: Michael S. Tsirkin --- V7->V8: set vq index unconditionally, make bisect happy, make sure don't introduce bug. (Jason) V6->V7: serialize virtnet_set_queues to avoid a race with cpu hotplug (Jason) V5->V6: remove {get|put}_online_cpus from virtnet_del_vqs (Jason) V4->V5: Add get/put_online_cpus to avoid CPUs go up and down during operations (Rusty) V3->V4: move vq_index into virtnet_info (Jason) change the mapping value when not setting affinity (Jason) address the comments about select_queue (Rusty) drivers/net/virtio_net.c | 67 ++++++++++++++++++++++++++++++++++++++---------- 1 file changed, 54 insertions(+), 13 deletions(-) diff --git a/drivers/net/virtio_net.c b/drivers/net/virtio_net.c index a6fcf15..fda214a 100644 --- a/drivers/net/virtio_net.c +++ b/drivers/net/virtio_net.c @@ -123,6 +123,9 @@ struct virtnet_info { /* Does the affinity hint is set for virtqueues? */ bool affinity_hint_set; + + /* Per-cpu variable to show the mapping from CPU to virtqueue */ + int __percpu *vq_index; }; struct skb_vnet_hdr { @@ -1016,6 +1019,7 @@ static int virtnet_vlan_rx_kill_vid(struct net_device *dev, u16 vid) static void virtnet_set_affinity(struct virtnet_info *vi, bool set) { int i; + int cpu; /* In multiqueue mode, when the number of cpu is equal to the number of * queue pairs, we let the queue pairs to be private to one cpu by @@ -1023,22 +1027,40 @@ static void virtnet_set_affinity(struct virtnet_info *vi, bool set) */ if ((vi->curr_queue_pairs == 1 || vi->max_queue_pairs != num_online_cpus()) && set) { - if (vi->affinity_hint_set) + if (vi->affinity_hint_set) { set = false; - else + } else { + i = 0; + for_each_online_cpu(cpu) + *per_cpu_ptr(vi->vq_index, cpu) = + ++i % vi->curr_queue_pairs; return; + } } - for (i = 0; i < vi->max_queue_pairs; i++) { - int cpu = set ? i : -1; - virtqueue_set_affinity(vi->rq[i].vq, cpu); - virtqueue_set_affinity(vi->sq[i].vq, cpu); - } + if (set) { + i = 0; + for_each_online_cpu(cpu) { + virtqueue_set_affinity(vi->rq[i].vq, cpu); + virtqueue_set_affinity(vi->sq[i].vq, cpu); + *per_cpu_ptr(vi->vq_index, cpu) = i; + i++; + } - if (set) vi->affinity_hint_set = true; - else + } else { + for(i = 0; i < vi->max_queue_pairs; i++) { + virtqueue_set_affinity(vi->rq[i].vq, -1); + virtqueue_set_affinity(vi->sq[i].vq, -1); + } + + i = 0; + for_each_online_cpu(cpu) + *per_cpu_ptr(vi->vq_index, cpu) = + ++i % vi->curr_queue_pairs; + vi->affinity_hint_set = false; + } } static void virtnet_get_ringparam(struct net_device *dev, @@ -1082,6 +1104,7 @@ static int virtnet_set_channels(struct net_device *dev, if (queue_pairs > vi->max_queue_pairs) return -EINVAL; + get_online_cpus(); err = virtnet_set_queues(vi, queue_pairs); if (!err) { netif_set_real_num_tx_queues(dev, queue_pairs); @@ -1089,6 +1112,7 @@ static int virtnet_set_channels(struct net_device *dev, virtnet_set_affinity(vi, true); } + put_online_cpus(); return err; } @@ -1127,12 +1151,19 @@ static int virtnet_change_mtu(struct net_device *dev, int new_mtu) /* To avoid contending a lock hold by a vcpu who would exit to host, select the * txq based on the processor id. - * TODO: handle cpu hotplug. */ static u16 virtnet_select_queue(struct net_device *dev, struct sk_buff *skb) { - int txq = skb_rx_queue_recorded(skb) ? skb_get_rx_queue(skb) : - smp_processor_id(); + int txq; + struct virtnet_info *vi = netdev_priv(dev); + + if (skb_rx_queue_recorded(skb)) { + txq = skb_get_rx_queue(skb); + } else { + txq = *__this_cpu_ptr(vi->vq_index); + if (txq == -1) + txq = 0; + } while (unlikely(txq >= dev->real_num_tx_queues)) txq -= dev->real_num_tx_queues; @@ -1371,7 +1402,10 @@ static int init_vqs(struct virtnet_info *vi) if (ret) goto err_free; + get_online_cpus(); virtnet_set_affinity(vi, true); + put_online_cpus(); + return 0; err_free: @@ -1453,6 +1487,10 @@ static int virtnet_probe(struct virtio_device *vdev) if (vi->stats == NULL) goto free; + vi->vq_index = alloc_percpu(int); + if (vi->vq_index == NULL) + goto free_stats; + mutex_init(&vi->config_lock); vi->config_enable = true; INIT_WORK(&vi->config_work, virtnet_config_changed_work); @@ -1476,7 +1514,7 @@ static int virtnet_probe(struct virtio_device *vdev) /* Allocate/initialize the rx/tx queues, and invoke find_vqs */ err = init_vqs(vi); if (err) - goto free_stats; + goto free_index; netif_set_real_num_tx_queues(dev, 1); netif_set_real_num_rx_queues(dev, 1); @@ -1520,6 +1558,8 @@ free_recv_bufs: free_vqs: cancel_delayed_work_sync(&vi->refill); virtnet_del_vqs(vi); +free_index: + free_percpu(vi->vq_index); free_stats: free_percpu(vi->stats); free: @@ -1554,6 +1594,7 @@ static void virtnet_remove(struct virtio_device *vdev) flush_work(&vi->config_work); + free_percpu(vi->vq_index); free_percpu(vi->stats); free_netdev(vi->dev); }