X-Git-Url: https://pintos-os.org/cgi-bin/gitweb.cgi?a=blobdiff_plain;f=lib%2Fnetdev-linux.c;h=e9fabc5b8083b067a23ae94b7405fba5e8c4fe50;hb=9dca7bd50a93eeb26fb7e47ddb896de22f782e7b;hp=e86a160c8e17da10d506ab277950e75e0faf1448;hpb=8722022c0c0d29d3f998dc26c50944c456e56646;p=openvswitch

diff --git a/lib/netdev-linux.c b/lib/netdev-linux.c
index e86a160c..e9fabc5b 100644
--- a/lib/netdev-linux.c
+++ b/lib/netdev-linux.c
@@ -24,6 +24,7 @@
 #include <linux/ip.h>
 #include <linux/types.h>
 #include <linux/ethtool.h>
+#include <linux/pkt_sched.h>
 #include <linux/rtnetlink.h>
 #include <linux/sockios.h>
 #include <linux/version.h>
@@ -81,7 +82,8 @@ enum {
     VALID_IN6 = 1 << 3,
     VALID_MTU = 1 << 4,
     VALID_CARRIER = 1 << 5,
-    VALID_IS_PSEUDO = 1 << 6       /* Represents is_internal and is_tap. */
+    VALID_IS_PSEUDO = 1 << 6,       /* Represents is_internal and is_tap. */
+    VALID_POLICING = 1 << 7
 };
 
 struct tap_state {
@@ -108,6 +110,8 @@ struct netdev_dev_linux {
     int carrier;
     bool is_internal;           /* Is this an openvswitch internal device? */
     bool is_tap;                /* Is this a tuntap device? */
+    uint32_t kbits_rate;        /* Policing data. */
+    uint32_t kbits_burst;
 
     union {
         struct tap_state tap;
@@ -156,6 +160,7 @@ static int set_etheraddr(const char *netdev_name, int hwaddr_family,
                          const uint8_t[ETH_ADDR_LEN]);
 static int get_stats_via_netlink(int ifindex, struct netdev_stats *stats);
 static int get_stats_via_proc(const char *netdev_name, struct netdev_stats *stats);
+static int get_rtnl_sock(struct nl_sock **);
 
 static bool
 is_netdev_linux_class(const struct netdev_class *netdev_class)
@@ -1272,35 +1277,88 @@ done:
 
 #define POLICE_ADD_CMD "/sbin/tc qdisc add dev %s handle ffff: ingress"
 #define POLICE_CONFIG_CMD "/sbin/tc filter add dev %s parent ffff: protocol ip prio 50 u32 match ip src 0.0.0.0/0 police rate %dkbit burst %dk mtu 65535 drop flowid :1"
-/* We redirect stderr to /dev/null because we often want to remove all
- * traffic control configuration on a port so its in a known state.  If
- * this done when there is no such configuration, tc complains, so we just
- * always ignore it.
+
+/* Remove ingress policing from 'netdev'.  Returns 0 if successful, otherwise a
+ * positive errno value.
+ *
+ * This function is equivalent to running
+ *     /sbin/tc qdisc del dev %s handle ffff: ingress
+ * but it is much, much faster.
  */
-#define POLICE_DEL_CMD "/sbin/tc qdisc del dev %s handle ffff: ingress 2>/dev/null"
+static int
+netdev_linux_remove_policing(struct netdev *netdev)
+{
+    struct netdev_dev_linux *netdev_dev =
+        netdev_dev_linux_cast(netdev_get_dev(netdev));
+    const char *netdev_name = netdev_get_name(netdev);
+
+    struct ofpbuf request;
+    struct ofpbuf *reply;
+    struct tcmsg *tcmsg;
+    struct nl_sock *rtnl_sock;
+    int ifindex;
+    int error;
+
+    error = get_ifindex(netdev, &ifindex);
+    if (error) {
+        return error;
+    }
+
+    error = get_rtnl_sock(&rtnl_sock);
+    if (error) {
+        return error;
+    }
+
+    ofpbuf_init(&request, 0);
+    nl_msg_put_nlmsghdr(&request, rtnl_sock, sizeof *tcmsg,
+                        RTM_DELQDISC, NLM_F_REQUEST);
+    tcmsg = ofpbuf_put_zeros(&request, sizeof *tcmsg);
+    tcmsg->tcm_family = AF_UNSPEC;
+    tcmsg->tcm_ifindex = ifindex;
+    tcmsg->tcm_handle = 0xffff0000;
+    tcmsg->tcm_parent = TC_H_INGRESS;
+    nl_msg_put_string(&request, TCA_KIND, "ingress");
+    nl_msg_put_unspec(&request, TCA_OPTIONS, NULL, 0);
+    error = nl_sock_transact(rtnl_sock, &request, &reply);
+    ofpbuf_uninit(&request);
+    ofpbuf_delete(reply);
+    if (error && error != ENOENT) {
+        VLOG_WARN_RL(&rl, "%s: removing policing failed: %s",
+                     netdev_name, strerror(error));
+        return error;
+    }
+
+    netdev_dev->kbits_rate = 0;
+    netdev_dev->kbits_burst = 0;
+    netdev_dev->cache_valid |= VALID_POLICING;
+    return 0;
+}
 
 /* Attempts to set input rate limiting (policing) policy. */
 static int
 netdev_linux_set_policing(struct netdev *netdev,
                           uint32_t kbits_rate, uint32_t kbits_burst)
 {
+    struct netdev_dev_linux *netdev_dev =
+        netdev_dev_linux_cast(netdev_get_dev(netdev));
     const char *netdev_name = netdev_get_name(netdev);
     char command[1024];
 
     COVERAGE_INC(netdev_set_policing);
-    if (kbits_rate) {
-        if (!kbits_burst) {
-            /* Default to 1000 kilobits if not specified. */
-            kbits_burst = 1000;
-        }
 
-        /* xxx This should be more careful about only adding if it
-         * xxx actually exists, as opposed to always deleting it. */
-        snprintf(command, sizeof(command), POLICE_DEL_CMD, netdev_name);
-        if (system(command) == -1) {
-            VLOG_WARN_RL(&rl, "%s: problem removing policing", netdev_name);
-        }
+    kbits_burst = (!kbits_rate ? 0       /* Force to 0 if no rate specified. */
+                   : !kbits_burst ? 1000 /* Default to 1000 kbits if 0. */
+                   : kbits_burst);       /* Stick with user-specified value. */
 
+    if (netdev_dev->cache_valid & VALID_POLICING
+        && netdev_dev->kbits_rate == kbits_rate
+        && netdev_dev->kbits_burst == kbits_burst) {
+        /* Assume that settings haven't changed since we last set them. */
+        return 0;
+    }
+
+    netdev_linux_remove_policing(netdev);
+    if (kbits_rate) {
         snprintf(command, sizeof(command), POLICE_ADD_CMD, netdev_name);
         if (system(command) != 0) {
             VLOG_WARN_RL(&rl, "%s: problem adding policing", netdev_name);
@@ -1314,11 +1372,10 @@ netdev_linux_set_policing(struct netdev *netdev,
                     netdev_name);
             return -1;
         }
-    } else {
-        snprintf(command, sizeof(command), POLICE_DEL_CMD, netdev_name);
-        if (system(command) == -1) {
-            VLOG_WARN_RL(&rl, "%s: problem removing policing", netdev_name);
-        }
+
+        netdev_dev->kbits_rate = kbits_rate;
+        netdev_dev->kbits_burst = kbits_burst;
+        netdev_dev->cache_valid |= VALID_POLICING;
     }
 
     return 0;
@@ -1850,8 +1907,7 @@ get_stats_via_netlink(int ifindex, struct netdev_stats *stats)
                          .min_len = sizeof(struct rtnl_link_stats) },
     };
 
-
-    static struct nl_sock *rtnl_sock;
+    struct nl_sock *rtnl_sock;
     struct ofpbuf request;
     struct ofpbuf *reply;
     struct ifinfomsg *ifi;
@@ -1859,13 +1915,9 @@ get_stats_via_netlink(int ifindex, struct netdev_stats *stats)
     struct nlattr *attrs[ARRAY_SIZE(rtnlgrp_link_policy)];
     int error;
 
-    if (!rtnl_sock) {
-        error = nl_sock_create(NETLINK_ROUTE, 0, 0, 0, &rtnl_sock);
-        if (error) {
-            VLOG_ERR_RL(&rl, "failed to create rtnetlink socket: %s",
-                        strerror(error));
-            return error;
-        }
+    error = get_rtnl_sock(&rtnl_sock);
+    if (error) {
+        return error;
     }
 
     ofpbuf_init(&request, 0);
@@ -2129,3 +2181,26 @@ netdev_linux_get_ipv4(const struct netdev *netdev, struct in_addr *ip,
     }
     return error;
 }
+
+/* Obtains a Netlink routing socket that is not subscribed to any multicast
+ * groups.  Returns 0 if successful, otherwise a positive errno value.  Stores
+ * the socket in '*rtnl_sockp' if successful, otherwise a null pointer. */
+static int
+get_rtnl_sock(struct nl_sock **rtnl_sockp)
+{
+    static struct nl_sock *sock;
+    int error;
+
+    if (!sock) {
+        error = nl_sock_create(NETLINK_ROUTE, 0, 0, 0, &sock);
+        if (error) {
+            VLOG_ERR_RL(&rl, "failed to create rtnetlink socket: %s",
+                        strerror(error));
+        }
+    } else {
+        error = 0;
+    }
+
+    *rtnl_sockp = sock;
+    return error;
+}