net: use core MTU range checking in core net infra
authorJarod Wilson <jarod@redhat.com>
Thu, 20 Oct 2016 17:55:20 +0000 (13:55 -0400)
committerDavid S. Miller <davem@davemloft.net>
Thu, 20 Oct 2016 18:51:09 +0000 (14:51 -0400)
geneve:
- Merge __geneve_change_mtu back into geneve_change_mtu, set max_mtu
- This one isn't quite as straight-forward as others, could use some
  closer inspection and testing

macvlan:
- set min/max_mtu

tun:
- set min/max_mtu, remove tun_net_change_mtu

vxlan:
- Merge __vxlan_change_mtu back into vxlan_change_mtu
- Set max_mtu to IP_MAX_MTU and retain dynamic MTU range checks in
  change_mtu function
- This one is also not as straight-forward and could use closer inspection
  and testing from vxlan folks

bridge:
- set max_mtu of IP_MAX_MTU and retain dynamic MTU range checks in
  change_mtu function

openvswitch:
- set min/max_mtu, remove internal_dev_change_mtu
- note: max_mtu wasn't checked previously, it's been set to 65535, which
  is the largest possible size supported

sch_teql:
- set min/max_mtu (note: max_mtu previously unchecked, used max of 65535)

macsec:
- min_mtu = 0, max_mtu = 65535

macvlan:
- min_mtu = 0, max_mtu = 65535

ntb_netdev:
- min_mtu = 0, max_mtu = 65535

veth:
- min_mtu = 68, max_mtu = 65535

8021q:
- min_mtu = 0, max_mtu = 65535

CC: netdev@vger.kernel.org
CC: Nicolas Dichtel <nicolas.dichtel@6wind.com>
CC: Hannes Frederic Sowa <hannes@stressinduktion.org>
CC: Tom Herbert <tom@herbertland.com>
CC: Daniel Borkmann <daniel@iogearbox.net>
CC: Alexander Duyck <alexander.h.duyck@intel.com>
CC: Paolo Abeni <pabeni@redhat.com>
CC: Jiri Benc <jbenc@redhat.com>
CC: WANG Cong <xiyou.wangcong@gmail.com>
CC: Roopa Prabhu <roopa@cumulusnetworks.com>
CC: Pravin B Shelar <pshelar@ovn.org>
CC: Sabrina Dubroca <sd@queasysnail.net>
CC: Patrick McHardy <kaber@trash.net>
CC: Stephen Hemminger <stephen@networkplumber.org>
CC: Pravin Shelar <pshelar@nicira.com>
CC: Maxim Krasnyansky <maxk@qti.qualcomm.com>
Signed-off-by: Jarod Wilson <jarod@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
drivers/net/geneve.c
drivers/net/macsec.c
drivers/net/macvlan.c
drivers/net/ntb_netdev.c
drivers/net/tun.c
drivers/net/veth.c
drivers/net/vxlan.c
net/8021q/vlan_dev.c
net/bridge/br_device.c
net/openvswitch/vport-internal_dev.c
net/sched/sch_teql.c

index 3c20e87..752bcaa 100644 (file)
@@ -1034,39 +1034,18 @@ static netdev_tx_t geneve_xmit(struct sk_buff *skb, struct net_device *dev)
        return geneve_xmit_skb(skb, dev, info);
 }
 
-static int __geneve_change_mtu(struct net_device *dev, int new_mtu, bool strict)
+static int geneve_change_mtu(struct net_device *dev, int new_mtu)
 {
-       struct geneve_dev *geneve = netdev_priv(dev);
-       /* The max_mtu calculation does not take account of GENEVE
-        * options, to avoid excluding potentially valid
-        * configurations.
+       /* Only possible if called internally, ndo_change_mtu path's new_mtu
+        * is guaranteed to be between dev->min_mtu and dev->max_mtu.
         */
-       int max_mtu = IP_MAX_MTU - GENEVE_BASE_HLEN - dev->hard_header_len;
-
-       if (geneve->remote.sa.sa_family == AF_INET6)
-               max_mtu -= sizeof(struct ipv6hdr);
-       else
-               max_mtu -= sizeof(struct iphdr);
-
-       if (new_mtu < 68)
-               return -EINVAL;
-
-       if (new_mtu > max_mtu) {
-               if (strict)
-                       return -EINVAL;
-
-               new_mtu = max_mtu;
-       }
+       if (new_mtu > dev->max_mtu)
+               new_mtu = dev->max_mtu;
 
        dev->mtu = new_mtu;
        return 0;
 }
 
-static int geneve_change_mtu(struct net_device *dev, int new_mtu)
-{
-       return __geneve_change_mtu(dev, new_mtu, true);
-}
-
 static int geneve_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb)
 {
        struct ip_tunnel_info *info = skb_tunnel_info(skb);
@@ -1170,6 +1149,14 @@ static void geneve_setup(struct net_device *dev)
        dev->hw_features |= NETIF_F_SG | NETIF_F_HW_CSUM | NETIF_F_RXCSUM;
        dev->hw_features |= NETIF_F_GSO_SOFTWARE;
 
+       /* MTU range: 68 - (something less than 65535) */
+       dev->min_mtu = ETH_MIN_MTU;
+       /* The max_mtu calculation does not take account of GENEVE
+        * options, to avoid excluding potentially valid
+        * configurations. This will be further reduced by IPvX hdr size.
+        */
+       dev->max_mtu = IP_MAX_MTU - GENEVE_BASE_HLEN - dev->hard_header_len;
+
        netif_keep_dst(dev);
        dev->priv_flags &= ~IFF_TX_SKB_SHARING;
        dev->priv_flags |= IFF_LIVE_ADDR_CHANGE | IFF_NO_QUEUE;
@@ -1285,10 +1272,13 @@ static int geneve_configure(struct net *net, struct net_device *dev,
 
        /* make enough headroom for basic scenario */
        encap_len = GENEVE_BASE_HLEN + ETH_HLEN;
-       if (remote->sa.sa_family == AF_INET)
+       if (remote->sa.sa_family == AF_INET) {
                encap_len += sizeof(struct iphdr);
-       else
+               dev->max_mtu -= sizeof(struct iphdr);
+       } else {
                encap_len += sizeof(struct ipv6hdr);
+               dev->max_mtu -= sizeof(struct ipv6hdr);
+       }
        dev->needed_headroom = encap_len + ETH_HLEN;
 
        if (metadata) {
@@ -1488,7 +1478,7 @@ struct net_device *geneve_dev_create_fb(struct net *net, const char *name,
        /* openvswitch users expect packet sizes to be unrestricted,
         * so set the largest MTU we can.
         */
-       err = __geneve_change_mtu(dev, IP_MAX_MTU, false);
+       err = geneve_change_mtu(dev, IP_MAX_MTU);
        if (err)
                goto err;
 
index 3ea47f2..1a134cb 100644 (file)
@@ -2970,6 +2970,8 @@ static void macsec_free_netdev(struct net_device *dev)
 static void macsec_setup(struct net_device *dev)
 {
        ether_setup(dev);
+       dev->min_mtu = 0;
+       dev->max_mtu = ETH_MAX_MTU;
        dev->priv_flags |= IFF_NO_QUEUE;
        dev->netdev_ops = &macsec_netdev_ops;
        dev->destructor = macsec_free_netdev;
index 3234fcd..a064415 100644 (file)
@@ -777,7 +777,7 @@ static int macvlan_change_mtu(struct net_device *dev, int new_mtu)
 {
        struct macvlan_dev *vlan = netdev_priv(dev);
 
-       if (new_mtu < 68 || vlan->lowerdev->mtu < new_mtu)
+       if (vlan->lowerdev->mtu < new_mtu)
                return -EINVAL;
        dev->mtu = new_mtu;
        return 0;
@@ -1085,6 +1085,8 @@ void macvlan_common_setup(struct net_device *dev)
 {
        ether_setup(dev);
 
+       dev->min_mtu            = 0;
+       dev->max_mtu            = ETH_MAX_MTU;
        dev->priv_flags        &= ~IFF_TX_SKB_SHARING;
        netif_keep_dst(dev);
        dev->priv_flags        |= IFF_UNICAST_FLT;
@@ -1297,6 +1299,10 @@ int macvlan_common_newlink(struct net *src_net, struct net_device *dev,
        else if (dev->mtu > lowerdev->mtu)
                return -EINVAL;
 
+       /* MTU range: 68 - lowerdev->max_mtu */
+       dev->min_mtu = ETH_MIN_MTU;
+       dev->max_mtu = lowerdev->max_mtu;
+
        if (!tb[IFLA_ADDRESS])
                eth_hw_addr_random(dev);
 
index a9acf71..36877ba 100644 (file)
@@ -433,6 +433,9 @@ static int ntb_netdev_probe(struct device *client_dev)
        ndev->netdev_ops = &ntb_netdev_ops;
        ndev->ethtool_ops = &ntb_ethtool_ops;
 
+       ndev->min_mtu = 0;
+       ndev->max_mtu = ETH_MAX_MTU;
+
        dev->qp = ntb_transport_create_queue(ndev, client_dev,
                                             &ntb_netdev_handlers);
        if (!dev->qp) {
index 8093e39..9328568 100644 (file)
@@ -925,18 +925,6 @@ static void tun_net_mclist(struct net_device *dev)
         */
 }
 
-#define MIN_MTU 68
-#define MAX_MTU 65535
-
-static int
-tun_net_change_mtu(struct net_device *dev, int new_mtu)
-{
-       if (new_mtu < MIN_MTU || new_mtu + dev->hard_header_len > MAX_MTU)
-               return -EINVAL;
-       dev->mtu = new_mtu;
-       return 0;
-}
-
 static netdev_features_t tun_net_fix_features(struct net_device *dev,
        netdev_features_t features)
 {
@@ -1014,7 +1002,6 @@ static const struct net_device_ops tun_netdev_ops = {
        .ndo_open               = tun_net_open,
        .ndo_stop               = tun_net_close,
        .ndo_start_xmit         = tun_net_xmit,
-       .ndo_change_mtu         = tun_net_change_mtu,
        .ndo_fix_features       = tun_net_fix_features,
        .ndo_select_queue       = tun_select_queue,
 #ifdef CONFIG_NET_POLL_CONTROLLER
@@ -1029,7 +1016,6 @@ static const struct net_device_ops tap_netdev_ops = {
        .ndo_open               = tun_net_open,
        .ndo_stop               = tun_net_close,
        .ndo_start_xmit         = tun_net_xmit,
-       .ndo_change_mtu         = tun_net_change_mtu,
        .ndo_fix_features       = tun_net_fix_features,
        .ndo_set_rx_mode        = tun_net_mclist,
        .ndo_set_mac_address    = eth_mac_addr,
@@ -1062,6 +1048,9 @@ static void tun_flow_uninit(struct tun_struct *tun)
        tun_flow_flush(tun);
 }
 
+#define MIN_MTU 68
+#define MAX_MTU 65535
+
 /* Initialize net device. */
 static void tun_net_init(struct net_device *dev)
 {
@@ -1092,6 +1081,9 @@ static void tun_net_init(struct net_device *dev)
 
                break;
        }
+
+       dev->min_mtu = MIN_MTU;
+       dev->max_mtu = MAX_MTU - dev->hard_header_len;
 }
 
 /* Character device part */
index fbc853e..0520952 100644 (file)
@@ -23,9 +23,6 @@
 #define DRV_NAME       "veth"
 #define DRV_VERSION    "1.0"
 
-#define MIN_MTU 68             /* Min L3 MTU */
-#define MAX_MTU 65535          /* Max L3 MTU (arbitrary) */
-
 struct pcpu_vstats {
        u64                     packets;
        u64                     bytes;
@@ -216,17 +213,9 @@ static int veth_close(struct net_device *dev)
        return 0;
 }
 
-static int is_valid_veth_mtu(int new_mtu)
+static int is_valid_veth_mtu(int mtu)
 {
-       return new_mtu >= MIN_MTU && new_mtu <= MAX_MTU;
-}
-
-static int veth_change_mtu(struct net_device *dev, int new_mtu)
-{
-       if (!is_valid_veth_mtu(new_mtu))
-               return -EINVAL;
-       dev->mtu = new_mtu;
-       return 0;
+       return mtu >= ETH_MIN_MTU && mtu <= ETH_MAX_MTU;
 }
 
 static int veth_dev_init(struct net_device *dev)
@@ -300,7 +289,6 @@ static const struct net_device_ops veth_netdev_ops = {
        .ndo_open            = veth_open,
        .ndo_stop            = veth_close,
        .ndo_start_xmit      = veth_xmit,
-       .ndo_change_mtu      = veth_change_mtu,
        .ndo_get_stats64     = veth_get_stats64,
        .ndo_set_rx_mode     = veth_set_multicast_list,
        .ndo_set_mac_address = eth_mac_addr,
@@ -337,6 +325,7 @@ static void veth_setup(struct net_device *dev)
                               NETIF_F_HW_VLAN_CTAG_RX |
                               NETIF_F_HW_VLAN_STAG_RX);
        dev->destructor = veth_dev_free;
+       dev->max_mtu = ETH_MAX_MTU;
 
        dev->hw_features = VETH_FEATURES;
        dev->hw_enc_features = VETH_FEATURES;
index e7d1668..c0170b6 100644 (file)
@@ -2367,43 +2367,31 @@ static void vxlan_set_multicast_list(struct net_device *dev)
 {
 }
 
-static int __vxlan_change_mtu(struct net_device *dev,
-                             struct net_device *lowerdev,
-                             struct vxlan_rdst *dst, int new_mtu, bool strict)
+static int vxlan_change_mtu(struct net_device *dev, int new_mtu)
 {
-       int max_mtu = IP_MAX_MTU;
-
-       if (lowerdev)
-               max_mtu = lowerdev->mtu;
+       struct vxlan_dev *vxlan = netdev_priv(dev);
+       struct vxlan_rdst *dst = &vxlan->default_dst;
+       struct net_device *lowerdev = __dev_get_by_index(vxlan->net,
+                                                        dst->remote_ifindex);
+       bool use_ipv6 = false;
 
        if (dst->remote_ip.sa.sa_family == AF_INET6)
-               max_mtu -= VXLAN6_HEADROOM;
-       else
-               max_mtu -= VXLAN_HEADROOM;
-
-       if (new_mtu < 68)
-               return -EINVAL;
+               use_ipv6 = true;
 
-       if (new_mtu > max_mtu) {
-               if (strict)
+       /* This check is different than dev->max_mtu, because it looks at
+        * the lowerdev->mtu, rather than the static dev->max_mtu
+        */
+       if (lowerdev) {
+               int max_mtu = lowerdev->mtu -
+                             (use_ipv6 ? VXLAN6_HEADROOM : VXLAN_HEADROOM);
+               if (new_mtu > max_mtu)
                        return -EINVAL;
-
-               new_mtu = max_mtu;
        }
 
        dev->mtu = new_mtu;
        return 0;
 }
 
-static int vxlan_change_mtu(struct net_device *dev, int new_mtu)
-{
-       struct vxlan_dev *vxlan = netdev_priv(dev);
-       struct vxlan_rdst *dst = &vxlan->default_dst;
-       struct net_device *lowerdev = __dev_get_by_index(vxlan->net,
-                                                        dst->remote_ifindex);
-       return __vxlan_change_mtu(dev, lowerdev, dst, new_mtu, true);
-}
-
 static int vxlan_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb)
 {
        struct vxlan_dev *vxlan = netdev_priv(dev);
@@ -2795,6 +2783,10 @@ static int vxlan_dev_configure(struct net *src_net, struct net_device *dev,
                vxlan_ether_setup(dev);
        }
 
+       /* MTU range: 68 - 65535 */
+       dev->min_mtu = ETH_MIN_MTU;
+       dev->max_mtu = ETH_MAX_MTU;
+
        vxlan->net = src_net;
 
        dst->remote_vni = conf->vni;
@@ -2838,7 +2830,8 @@ static int vxlan_dev_configure(struct net *src_net, struct net_device *dev,
 #endif
 
                if (!conf->mtu)
-                       dev->mtu = lowerdev->mtu - (use_ipv6 ? VXLAN6_HEADROOM : VXLAN_HEADROOM);
+                       dev->mtu = lowerdev->mtu -
+                                  (use_ipv6 ? VXLAN6_HEADROOM : VXLAN_HEADROOM);
 
                needed_headroom = lowerdev->hard_header_len;
        } else if (vxlan_addr_multicast(&dst->remote_ip)) {
@@ -2847,9 +2840,20 @@ static int vxlan_dev_configure(struct net *src_net, struct net_device *dev,
        }
 
        if (conf->mtu) {
-               err = __vxlan_change_mtu(dev, lowerdev, dst, conf->mtu, false);
-               if (err)
-                       return err;
+               int max_mtu = ETH_MAX_MTU;
+
+               if (lowerdev)
+                       max_mtu = lowerdev->mtu;
+
+               max_mtu -= (use_ipv6 ? VXLAN6_HEADROOM : VXLAN_HEADROOM);
+
+               if (conf->mtu < dev->min_mtu || conf->mtu > dev->max_mtu)
+                       return -EINVAL;
+
+               dev->mtu = conf->mtu;
+
+               if (conf->mtu > max_mtu)
+                       dev->mtu = max_mtu;
        }
 
        if (use_ipv6 || conf->flags & VXLAN_F_COLLECT_METADATA)
index fbfacd5..10da6c5 100644 (file)
@@ -826,5 +826,8 @@ void vlan_setup(struct net_device *dev)
        dev->destructor         = vlan_dev_free;
        dev->ethtool_ops        = &vlan_ethtool_ops;
 
+       dev->min_mtu            = 0;
+       dev->max_mtu            = ETH_MAX_MTU;
+
        eth_zero_addr(dev->broadcast);
 }
index 89a687f..c08e02b 100644 (file)
@@ -185,7 +185,7 @@ static struct rtnl_link_stats64 *br_get_stats64(struct net_device *dev,
 static int br_change_mtu(struct net_device *dev, int new_mtu)
 {
        struct net_bridge *br = netdev_priv(dev);
-       if (new_mtu < 68 || new_mtu > br_min_mtu(br))
+       if (new_mtu > br_min_mtu(br))
                return -EINVAL;
 
        dev->mtu = new_mtu;
@@ -410,6 +410,7 @@ void br_dev_setup(struct net_device *dev)
        br->bridge_hello_time = br->hello_time = 2 * HZ;
        br->bridge_forward_delay = br->forward_delay = 15 * HZ;
        br->ageing_time = BR_DEFAULT_AGEING_TIME;
+       dev->max_mtu = ETH_MAX_MTU;
 
        br_netfilter_rtable_init(br);
        br_stp_timer_init(br);
index e7da290..d5d6cae 100644 (file)
@@ -89,15 +89,6 @@ static const struct ethtool_ops internal_dev_ethtool_ops = {
        .get_link       = ethtool_op_get_link,
 };
 
-static int internal_dev_change_mtu(struct net_device *netdev, int new_mtu)
-{
-       if (new_mtu < 68)
-               return -EINVAL;
-
-       netdev->mtu = new_mtu;
-       return 0;
-}
-
 static void internal_dev_destructor(struct net_device *dev)
 {
        struct vport *vport = ovs_internal_dev_get_vport(dev);
@@ -148,7 +139,6 @@ static const struct net_device_ops internal_dev_netdev_ops = {
        .ndo_stop = internal_dev_stop,
        .ndo_start_xmit = internal_dev_xmit,
        .ndo_set_mac_address = eth_mac_addr,
-       .ndo_change_mtu = internal_dev_change_mtu,
        .ndo_get_stats64 = internal_get_stats,
        .ndo_set_rx_headroom = internal_set_rx_headroom,
 };
index 2cd9b44..b019636 100644 (file)
@@ -418,9 +418,6 @@ static int teql_master_mtu(struct net_device *dev, int new_mtu)
        struct teql_master *m = netdev_priv(dev);
        struct Qdisc *q;
 
-       if (new_mtu < 68)
-               return -EINVAL;
-
        q = m->slaves;
        if (q) {
                do {
@@ -460,6 +457,8 @@ static __init void teql_master_setup(struct net_device *dev)
        dev->netdev_ops =       &teql_netdev_ops;
        dev->type               = ARPHRD_VOID;
        dev->mtu                = 1500;
+       dev->min_mtu            = 68;
+       dev->max_mtu            = 65535;
        dev->tx_queue_len       = 100;
        dev->flags              = IFF_NOARP;
        dev->hard_header_len    = LL_MAX_HEADER;