From patchwork Wed Apr 13 08:52:20 2016 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Paolo Abeni X-Patchwork-Id: 609902 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 3qlHck1pMcz9t5C for ; Wed, 13 Apr 2016 18:52:38 +1000 (AEST) Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S934590AbcDMIwf (ORCPT ); Wed, 13 Apr 2016 04:52:35 -0400 Received: from mx1.redhat.com ([209.132.183.28]:43483 "EHLO mx1.redhat.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1758852AbcDMIwc (ORCPT ); Wed, 13 Apr 2016 04:52:32 -0400 Received: from int-mx11.intmail.prod.int.phx2.redhat.com (int-mx11.intmail.prod.int.phx2.redhat.com [10.5.11.24]) (using TLSv1.2 with cipher ECDHE-RSA-AES256-GCM-SHA384 (256/256 bits)) (No client certificate requested) by mx1.redhat.com (Postfix) with ESMTPS id 14B697F080; Wed, 13 Apr 2016 08:52:32 +0000 (UTC) Received: from dhcp-176-88.mxp.redhat.com (dhcp-176-88.mxp.redhat.com [10.32.176.88]) by int-mx11.intmail.prod.int.phx2.redhat.com (8.14.4/8.14.4) with ESMTP id u3D8qTeA000954; Wed, 13 Apr 2016 04:52:30 -0400 From: Paolo Abeni To: netdev@vger.kernel.org Cc: "David S. Miller" , "Michael S. Tsirkin" , Hannes Frederic Sowa , "Eric W. Biederman" , Greg Kurz , Jason Wang Subject: [PATCH net-next] tun: use per cpu variables for stats accounting Date: Wed, 13 Apr 2016 10:52:20 +0200 Message-Id: <597a2e5c2f664d48d3e4aac2f7d43e61940d4037.1460537477.git.pabeni@redhat.com> X-Scanned-By: MIMEDefang 2.68 on 10.5.11.24 Sender: netdev-owner@vger.kernel.org Precedence: bulk List-ID: X-Mailing-List: netdev@vger.kernel.org Currently the tun device accounting uses dev->stats without applying any kind of protection, regardless that accounting happens in preemptible process context. This patch move the tun stats to a per cpu data structure, and protect the updates with u64_stats_update_begin()/u64_stats_update_end() or this_cpu_inc according to the stat type. The per cpu stats are aggregated by the newly added ndo_get_stats64 ops. Signed-off-by: Paolo Abeni --- drivers/net/tun.c | 95 ++++++++++++++++++++++++++++++++++++++++++++++++------- 1 file changed, 83 insertions(+), 12 deletions(-) diff --git a/drivers/net/tun.c b/drivers/net/tun.c index a746616..faf9297 100644 --- a/drivers/net/tun.c +++ b/drivers/net/tun.c @@ -131,6 +131,17 @@ struct tap_filter { #define TUN_FLOW_EXPIRE (3 * HZ) +struct tun_pcpu_stats { + u64 rx_packets; + u64 rx_bytes; + u64 tx_packets; + u64 tx_bytes; + struct u64_stats_sync syncp; + u32 rx_dropped; + u32 tx_dropped; + u32 rx_frame_errors; +}; + /* A tun_file connects an open character device to a tuntap netdevice. It * also contains all socket related structures (except sock_fprog and tap_filter) * to serve as one transmit queue for tuntap device. The sock_fprog and @@ -205,6 +216,7 @@ struct tun_struct { struct list_head disabled; void *security; u32 flow_count; + struct tun_pcpu_stats __percpu *pcpu_stats; }; #ifdef CONFIG_TUN_VNET_CROSS_LE @@ -886,7 +898,7 @@ static netdev_tx_t tun_net_xmit(struct sk_buff *skb, struct net_device *dev) return NETDEV_TX_OK; drop: - dev->stats.tx_dropped++; + this_cpu_inc(tun->pcpu_stats->tx_dropped); skb_tx_error(skb); kfree_skb(skb); rcu_read_unlock(); @@ -949,6 +961,43 @@ static void tun_set_headroom(struct net_device *dev, int new_hr) tun->align = new_hr; } +static struct rtnl_link_stats64 * +tun_net_get_stats64(struct net_device *dev, struct rtnl_link_stats64 *stats) +{ + u32 rx_dropped = 0, tx_dropped = 0, rx_frame_errors = 0; + struct tun_struct *tun = netdev_priv(dev); + struct tun_pcpu_stats *p; + int i; + + for_each_possible_cpu(i) { + u64 rxpackets, rxbytes, txpackets, txbytes; + unsigned int start; + + p = per_cpu_ptr(tun->pcpu_stats, i); + do { + start = u64_stats_fetch_begin(&p->syncp); + rxpackets = p->rx_packets; + rxbytes = p->rx_bytes; + txpackets = p->tx_packets; + txbytes = p->tx_bytes; + } while (u64_stats_fetch_retry(&p->syncp, start)); + + stats->rx_packets += rxpackets; + stats->rx_bytes += rxbytes; + stats->tx_packets += txpackets; + stats->tx_bytes += txbytes; + + /* u32 counters */ + rx_dropped += p->rx_dropped; + rx_frame_errors += p->rx_frame_errors; + tx_dropped += p->tx_dropped; + } + stats->rx_dropped = rx_dropped; + stats->rx_frame_errors = rx_frame_errors; + stats->tx_dropped = tx_dropped; + return stats; +} + static const struct net_device_ops tun_netdev_ops = { .ndo_uninit = tun_net_uninit, .ndo_open = tun_net_open, @@ -961,6 +1010,7 @@ static const struct net_device_ops tun_netdev_ops = { .ndo_poll_controller = tun_poll_controller, #endif .ndo_set_rx_headroom = tun_set_headroom, + .ndo_get_stats64 = tun_net_get_stats64, }; static const struct net_device_ops tap_netdev_ops = { @@ -979,6 +1029,7 @@ static const struct net_device_ops tap_netdev_ops = { #endif .ndo_features_check = passthru_features_check, .ndo_set_rx_headroom = tun_set_headroom, + .ndo_get_stats64 = tun_net_get_stats64, }; static void tun_flow_init(struct tun_struct *tun) @@ -1103,6 +1154,7 @@ static ssize_t tun_get_user(struct tun_struct *tun, struct tun_file *tfile, size_t total_len = iov_iter_count(from); size_t len = total_len, align = tun->align, linear; struct virtio_net_hdr gso = { 0 }; + struct tun_pcpu_stats *stats; int good_linear; int copylen; bool zerocopy = false; @@ -1177,7 +1229,7 @@ static ssize_t tun_get_user(struct tun_struct *tun, struct tun_file *tfile, skb = tun_alloc_skb(tfile, align, copylen, linear, noblock); if (IS_ERR(skb)) { if (PTR_ERR(skb) != -EAGAIN) - tun->dev->stats.rx_dropped++; + this_cpu_inc(tun->pcpu_stats->rx_dropped); return PTR_ERR(skb); } @@ -1192,7 +1244,7 @@ static ssize_t tun_get_user(struct tun_struct *tun, struct tun_file *tfile, } if (err) { - tun->dev->stats.rx_dropped++; + this_cpu_inc(tun->pcpu_stats->rx_dropped); kfree_skb(skb); return -EFAULT; } @@ -1200,7 +1252,7 @@ static ssize_t tun_get_user(struct tun_struct *tun, struct tun_file *tfile, if (gso.flags & VIRTIO_NET_HDR_F_NEEDS_CSUM) { if (!skb_partial_csum_set(skb, tun16_to_cpu(tun, gso.csum_start), tun16_to_cpu(tun, gso.csum_offset))) { - tun->dev->stats.rx_frame_errors++; + this_cpu_inc(tun->pcpu_stats->rx_frame_errors); kfree_skb(skb); return -EINVAL; } @@ -1217,7 +1269,7 @@ static ssize_t tun_get_user(struct tun_struct *tun, struct tun_file *tfile, pi.proto = htons(ETH_P_IPV6); break; default: - tun->dev->stats.rx_dropped++; + this_cpu_inc(tun->pcpu_stats->rx_dropped); kfree_skb(skb); return -EINVAL; } @@ -1245,7 +1297,7 @@ static ssize_t tun_get_user(struct tun_struct *tun, struct tun_file *tfile, skb_shinfo(skb)->gso_type = SKB_GSO_UDP; break; default: - tun->dev->stats.rx_frame_errors++; + this_cpu_inc(tun->pcpu_stats->rx_frame_errors); kfree_skb(skb); return -EINVAL; } @@ -1255,7 +1307,7 @@ static ssize_t tun_get_user(struct tun_struct *tun, struct tun_file *tfile, skb_shinfo(skb)->gso_size = tun16_to_cpu(tun, gso.gso_size); if (skb_shinfo(skb)->gso_size == 0) { - tun->dev->stats.rx_frame_errors++; + this_cpu_inc(tun->pcpu_stats->rx_frame_errors); kfree_skb(skb); return -EINVAL; } @@ -1278,8 +1330,12 @@ static ssize_t tun_get_user(struct tun_struct *tun, struct tun_file *tfile, rxhash = skb_get_hash(skb); netif_rx_ni(skb); - tun->dev->stats.rx_packets++; - tun->dev->stats.rx_bytes += len; + stats = get_cpu_ptr(tun->pcpu_stats); + u64_stats_update_begin(&stats->syncp); + stats->rx_packets++; + stats->rx_bytes += len; + u64_stats_update_end(&stats->syncp); + put_cpu_ptr(stats); tun_flow_update(tun, rxhash, tfile); return total_len; @@ -1308,6 +1364,7 @@ static ssize_t tun_put_user(struct tun_struct *tun, struct iov_iter *iter) { struct tun_pi pi = { 0, skb->protocol }; + struct tun_pcpu_stats *stats; ssize_t total; int vlan_offset = 0; int vlan_hlen = 0; @@ -1408,8 +1465,13 @@ static ssize_t tun_put_user(struct tun_struct *tun, skb_copy_datagram_iter(skb, vlan_offset, iter, skb->len - vlan_offset); done: - tun->dev->stats.tx_packets++; - tun->dev->stats.tx_bytes += skb->len + vlan_hlen; + /* caller is in process context, */ + stats = get_cpu_ptr(tun->pcpu_stats); + u64_stats_update_begin(&stats->syncp); + stats->tx_packets++; + stats->tx_bytes += skb->len + vlan_hlen; + u64_stats_update_end(&stats->syncp); + put_cpu_ptr(tun->pcpu_stats); return total; } @@ -1467,6 +1529,7 @@ static void tun_free_netdev(struct net_device *dev) struct tun_struct *tun = netdev_priv(dev); BUG_ON(!(list_empty(&tun->disabled))); + free_percpu(tun->pcpu_stats); tun_flow_uninit(tun); security_tun_dev_free_security(tun->security); free_netdev(dev); @@ -1715,11 +1778,17 @@ static int tun_set_iff(struct net *net, struct file *file, struct ifreq *ifr) tun->filter_attached = false; tun->sndbuf = tfile->socket.sk->sk_sndbuf; + tun->pcpu_stats = netdev_alloc_pcpu_stats(struct tun_pcpu_stats); + if (!tun->pcpu_stats) { + err = -ENOMEM; + goto err_free_dev; + } + spin_lock_init(&tun->lock); err = security_tun_dev_alloc_security(&tun->security); if (err < 0) - goto err_free_dev; + goto err_free_stat; tun_net_init(dev); tun_flow_init(tun); @@ -1763,6 +1832,8 @@ err_detach: err_free_flow: tun_flow_uninit(tun); security_tun_dev_free_security(tun->security); +err_free_stat: + free_percpu(tun->pcpu_stats); err_free_dev: free_netdev(dev); return err;