X-Git-Url: http://git.onelab.eu/?a=blobdiff_plain;f=lib%2Fnetdev.c;h=1f0b764f99caa66927a503b3d3d8c39ec4125d20;hb=943e5afe0b450fc665a4a162fe1bacafd34d18e8;hp=0b0463b9070d2fb83e403faa7d00f52c8f998065;hpb=c83cdd307248d798eac0d85681b0abeaaff1bf61;p=sliver-openvswitch.git

diff --git a/lib/netdev.c b/lib/netdev.c
index 0b0463b90..1f0b764f9 100644
--- a/lib/netdev.c
+++ b/lib/netdev.c
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2008, 2009, 2010 Nicira Networks.
+ * Copyright (c) 2008, 2009, 2010, 2011 Nicira Networks.
  *
  * Licensed under the Apache License, Version 2.0 (the "License");
  * you may not use this file except in compliance with the License.
@@ -31,24 +31,22 @@
 #include "hash.h"
 #include "list.h"
 #include "netdev-provider.h"
+#include "netdev-vport.h"
 #include "ofpbuf.h"
 #include "openflow/openflow.h"
 #include "packets.h"
 #include "poll-loop.h"
 #include "shash.h"
+#include "sset.h"
 #include "svec.h"
-
-#define THIS_MODULE VLM_netdev
 #include "vlog.h"
 
-static const struct netdev_class *base_netdev_classes[] = {
-#ifdef HAVE_NETLINK
-    &netdev_linux_class,
-    &netdev_tap_class,
-    &netdev_patch_class,
-    &netdev_gre_class,
-#endif
-};
+VLOG_DEFINE_THIS_MODULE(netdev);
+
+COVERAGE_DEFINE(netdev_received);
+COVERAGE_DEFINE(netdev_sent);
+COVERAGE_DEFINE(netdev_add_router);
+COVERAGE_DEFINE(netdev_get_stats);
 
 static struct shash netdev_classes = SHASH_INITIALIZER(&netdev_classes);
 
@@ -69,17 +67,19 @@ void update_device_args(struct netdev_dev *, const struct shash *args);
 static void
 netdev_initialize(void)
 {
-    static int status = -1;
+    static bool inited;
 
-    if (status < 0) {
-        int i;
+    if (!inited) {
+        inited = true;
 
         fatal_signal_add_hook(close_all_netdevs, NULL, NULL, true);
 
-        status = 0;
-        for (i = 0; i < ARRAY_SIZE(base_netdev_classes); i++) {
-            netdev_register_provider(base_netdev_classes[i]);
-        }
+#ifdef HAVE_NETLINK
+        netdev_register_provider(&netdev_linux_class);
+        netdev_register_provider(&netdev_internal_class);
+        netdev_register_provider(&netdev_tap_class);
+        netdev_vport_register();
+#endif
     }
 }
 
@@ -120,8 +120,6 @@ netdev_wait(void)
 int
 netdev_register_provider(const struct netdev_class *new_class)
 {
-    struct netdev_class *new_provider;
-
     if (shash_find(&netdev_classes, new_class->type)) {
         VLOG_WARN("attempted to register duplicate netdev provider: %s",
                    new_class->type);
@@ -137,10 +135,7 @@ netdev_register_provider(const struct netdev_class *new_class)
         }
     }
 
-    new_provider = xmalloc(sizeof *new_provider);
-    memcpy(new_provider, new_class, sizeof *new_provider);
-
-    shash_add(&netdev_classes, new_class->type, new_provider);
+    shash_add(&netdev_classes, new_class->type, new_class);
 
     return 0;
 }
@@ -170,121 +165,38 @@ netdev_unregister_provider(const char *type)
     }
 
     shash_delete(&netdev_classes, del_node);
-    free(del_node->data);
 
     return 0;
 }
 
+const struct netdev_class *
+netdev_lookup_provider(const char *type)
+{
+    netdev_initialize();
+    return shash_find_data(&netdev_classes, type && type[0] ? type : "system");
+}
+
 /* Clears 'types' and enumerates the types of all currently registered netdev
- * providers into it.  The caller must first initialize the svec. */
+ * providers into it.  The caller must first initialize the sset. */
 void
-netdev_enumerate_types(struct svec *types)
+netdev_enumerate_types(struct sset *types)
 {
     struct shash_node *node;
 
     netdev_initialize();
-    svec_clear(types);
+    sset_clear(types);
 
     SHASH_FOR_EACH(node, &netdev_classes) {
         const struct netdev_class *netdev_class = node->data;
-        svec_add(types, netdev_class->type);
-    }
-}
-
-/* Compares 'args' to those used to those used by 'dev'.  Returns true
- * if the arguments are the same, false otherwise.  Does not update the
- * values stored in 'dev'. */
-static bool
-compare_device_args(const struct netdev_dev *dev, const struct shash *args)
-{
-    const struct shash_node **new_args;
-    bool result = true;
-    int i;
-
-    if (shash_count(args) != dev->n_args) {
-        return false;
+        sset_add(types, netdev_class->type);
     }
-
-    new_args = shash_sort(args);
-    for (i = 0; i < dev->n_args; i++) {
-        if (strcmp(dev->args[i].key, new_args[i]->name) || 
-            strcmp(dev->args[i].value, new_args[i]->data)) {
-            result = false;
-            goto finish;
-        }
-    }
-
-finish:
-    free(new_args);
-    return result;
-}
-
-static int
-compare_args(const void *a_, const void *b_)
-{
-    const struct arg *a = a_;
-    const struct arg *b = b_;
-    return strcmp(a->key, b->key);
 }
 
 void
 update_device_args(struct netdev_dev *dev, const struct shash *args)
 {
-    struct shash_node *node;
-    int i;
-
-    if (dev->n_args) {
-        for (i = 0; i < dev->n_args; i++) {
-            free(dev->args[i].key);
-            free(dev->args[i].value);
-        }
-
-        free(dev->args);
-        dev->n_args = 0;
-    }
-
-    if (!args || shash_is_empty(args)) {
-        return;
-    }
-
-    dev->n_args = shash_count(args);
-    dev->args = xmalloc(dev->n_args * sizeof *dev->args);
-
-    i = 0;
-    SHASH_FOR_EACH(node, args) {
-        dev->args[i].key = xstrdup(node->name);
-        dev->args[i].value = xstrdup(node->data);
-        i++;
-    }
-
-    qsort(dev->args, dev->n_args, sizeof *dev->args, compare_args);
-}
-
-static int
-create_device(struct netdev_options *options, struct netdev_dev **netdev_devp)
-{
-    struct netdev_class *netdev_class;
-
-    if (!options->may_create) {
-        VLOG_WARN("attempted to create a device that may not be created: %s",
-                  options->name);
-        return ENODEV;
-    }
-
-    if (!options->type || strlen(options->type) == 0) {
-        /* Default to system. */
-        options->type = "system";
-    }
-
-    netdev_class = shash_find_data(&netdev_classes, options->type);
-    if (!netdev_class) {
-        VLOG_WARN("could not create netdev %s of unknown type %s",
-                  options->name, options->type);
-        return EAFNOSUPPORT;
-    }
-
-    return netdev_class->create(options->name, options->type, options->args,
-                                netdev_devp);
+    smap_destroy(&dev->args);
+    smap_clone(&dev->args, args);
 }
 
 /* Opens the network device named 'name' (e.g. "eth0") and returns zero if
@@ -292,21 +204,12 @@ create_device(struct netdev_options *options, struct netdev_dev **netdev_devp)
  * to the new network device, otherwise to null.
  *
  * If this is the first time the device has been opened, then create is called
- * before opening.  The device is  created using the given type and arguments.
+ * before opening.  The device is created using the given type and arguments.
  *
  * 'ethertype' may be a 16-bit Ethernet protocol value in host byte order to
  * capture frames of that type received on the device.  It may also be one of
  * the 'enum netdev_pseudo_ethertype' values to receive frames in one of those
- * categories.
- *
- * If the 'may_create' flag is set then this is allowed to be the first time
- * the device is opened (i.e. the refcount will be 1 after this call).  It
- * may be set to false if the device should have already been created.
- *
- * If the 'may_open' flag is set then the call will succeed even if another
- * caller has already opened it.  It may be to false if the device should not
- * currently be open. */
-
+ * categories. */
 int
 netdev_open(struct netdev_options *options, struct netdev **netdevp)
 {
@@ -324,27 +227,30 @@ netdev_open(struct netdev_options *options, struct netdev **netdevp)
     netdev_dev = shash_find_data(&netdev_dev_shash, options->name);
 
     if (!netdev_dev) {
-        error = create_device(options, &netdev_dev);
+        const struct netdev_class *class;
+
+        class = netdev_lookup_provider(options->type);
+        if (!class) {
+            VLOG_WARN("could not create netdev %s of unknown type %s",
+                      options->name, options->type);
+            return EAFNOSUPPORT;
+        }
+        error = class->create(class, options->name, options->args,
+                              &netdev_dev);
         if (error) {
             return error;
         }
-        update_device_args(netdev_dev, options->args);
+        assert(netdev_dev->netdev_class == class);
 
-    } else if (options->may_open) {
-        if (!shash_is_empty(options->args) &&
-            !compare_device_args(netdev_dev, options->args)) {
+    } else if (!shash_is_empty(options->args) &&
+               !smap_equal(&netdev_dev->args, options->args)) {
 
-            VLOG_WARN("%s: attempted to open already created netdev with "
-                      "different arguments", options->name);
-            return EINVAL;
-        }
-    } else {
-        VLOG_WARN("%s: attempted to create a netdev device with bound name",
-                  options->name);
-        return EEXIST;
+        VLOG_WARN("%s: attempted to open already open netdev with "
+                  "different arguments", options->name);
+        return EINVAL;
     }
 
-    error = netdev_dev->netdev_class->open(netdev_dev, options->ethertype, 
+    error = netdev_dev->netdev_class->open(netdev_dev, options->ethertype,
                 netdevp);
 
     if (!error) {
@@ -364,11 +270,8 @@ netdev_open_default(const char *name, struct netdev **netdevp)
     struct netdev_options options;
 
     memset(&options, 0, sizeof options);
-
     options.name = name;
     options.ethertype = NETDEV_ETH_TYPE_NONE;
-    options.may_create = true;
-    options.may_open = true;
 
     return netdev_open(&options, netdevp);
 }
@@ -376,7 +279,7 @@ netdev_open_default(const char *name, struct netdev **netdevp)
 /* Reconfigures the device 'netdev' with 'args'.  'args' may be empty
  * or NULL if none are needed. */
 int
-netdev_reconfigure(struct netdev *netdev, const struct shash *args)
+netdev_set_config(struct netdev *netdev, const struct shash *args)
 {
     struct shash empty_args = SHASH_INITIALIZER(&empty_args);
     struct netdev_dev *netdev_dev = netdev_get_dev(netdev);
@@ -385,19 +288,31 @@ netdev_reconfigure(struct netdev *netdev, const struct shash *args)
         args = &empty_args;
     }
 
-    if (netdev_dev->netdev_class->reconfigure) {
-        if (!compare_device_args(netdev_dev, args)) {
+    if (netdev_dev->netdev_class->set_config) {
+        if (!smap_equal(&netdev_dev->args, args)) {
             update_device_args(netdev_dev, args);
-            return netdev_dev->netdev_class->reconfigure(netdev_dev, args);
+            return netdev_dev->netdev_class->set_config(netdev_dev, args);
         }
     } else if (!shash_is_empty(args)) {
-        VLOG_WARN("%s: arguments provided to device that does not have a "
-                  "reconfigure function", netdev_get_name(netdev));
+        VLOG_WARN("%s: arguments provided to device whose configuration "
+                  "cannot be changed", netdev_get_name(netdev));
     }
 
     return 0;
 }
 
+/* Returns the current configuration for 'netdev'.  This is either the
+ * configuration passed to netdev_open() or netdev_set_config(), or it is a
+ * configuration retrieved from the device itself if no configuration was
+ * passed to those functions.
+ *
+ * 'netdev' retains ownership of the returned configuration. */
+const struct shash *
+netdev_get_config(const struct netdev *netdev)
+{
+    return &netdev_get_dev(netdev)->args;
+}
+
 /* Closes and destroys 'netdev'. */
 void
 netdev_close(struct netdev *netdev)
@@ -445,20 +360,20 @@ netdev_is_open(const char *name)
     return !!shash_find_data(&netdev_dev_shash, name);
 }
 
-/*  Clears 'svec' and enumerates the names of all known network devices. */
+/*  Clears 'sset' and enumerates the names of all known network devices. */
 int
-netdev_enumerate(struct svec *svec)
+netdev_enumerate(struct sset *sset)
 {
     struct shash_node *node;
     int error = 0;
 
     netdev_initialize();
-    svec_clear(svec);
+    sset_clear(sset);
 
     SHASH_FOR_EACH(node, &netdev_classes) {
         const struct netdev_class *netdev_class = node->data;
         if (netdev_class->enumerate) {
-            int retval = netdev_class->enumerate(svec);
+            int retval = netdev_class->enumerate(sset);
             if (retval) {
                 VLOG_WARN("failed to enumerate %s network devices: %s",
                           netdev_class->type, strerror(retval));
@@ -485,8 +400,7 @@ netdev_enumerate(struct svec *svec)
  * be returned.
  *
  * Some network devices may not implement support for this function.  In such
- * cases this function will always return EOPNOTSUPP.
- */
+ * cases this function will always return EOPNOTSUPP. */
 int
 netdev_recv(struct netdev *netdev, struct ofpbuf *buffer)
 {
@@ -608,8 +522,9 @@ netdev_get_name(const struct netdev *netdev)
  * (and received) packets, in bytes, not including the hardware header; thus,
  * this is typically 1500 bytes for Ethernet devices.
  *
- * If successful, returns 0 and stores the MTU size in '*mtup'.  On failure,
- * returns a positive errno value and stores ETH_PAYLOAD_MAX (1500) in
+ * If successful, returns 0 and stores the MTU size in '*mtup'.  Stores INT_MAX
+ * in '*mtup' if 'netdev' does not have an MTU (as e.g. some tunnels do not).On
+ * failure, returns a positive errno value and stores ETH_PAYLOAD_MAX (1500) in
  * '*mtup'. */
 int
 netdev_get_mtu(const struct netdev *netdev, int *mtup)
@@ -652,14 +567,13 @@ netdev_get_ifindex(const struct netdev *netdev)
  * passed-in values are set to 0.
  *
  * Some network devices may not implement support for this function.  In such
- * cases this function will always return EOPNOTSUPP.
- */
+ * cases this function will always return EOPNOTSUPP. */
 int
-netdev_get_features(struct netdev *netdev,
+netdev_get_features(const struct netdev *netdev,
                     uint32_t *current, uint32_t *advertised,
                     uint32_t *supported, uint32_t *peer)
 {
-    int (*get_features)(struct netdev *netdev,
+    int (*get_features)(const struct netdev *netdev,
                         uint32_t *current, uint32_t *advertised,
                         uint32_t *supported, uint32_t *peer);
     uint32_t dummy[4];
@@ -738,8 +652,8 @@ netdev_set_advertisements(struct netdev *netdev, uint32_t advertise)
  *
  *   - EOPNOTSUPP: No IPv4 network stack attached to 'netdev'.
  *
- * 'address' or 'netmask' or both may be null, in which case the address or netmask
- * is not reported. */
+ * 'address' or 'netmask' or both may be null, in which case the address or 
+ * netmask is not reported. */
 int
 netdev_get_in4(const struct netdev *netdev,
                struct in_addr *address_, struct in_addr *netmask_)
@@ -749,7 +663,7 @@ netdev_get_in4(const struct netdev *netdev,
     int error;
 
     error = (netdev_get_dev(netdev)->netdev_class->get_in4
-             ? netdev_get_dev(netdev)->netdev_class->get_in4(netdev, 
+             ? netdev_get_dev(netdev)->netdev_class->get_in4(netdev,
                     &address, &netmask)
              : EOPNOTSUPP);
     if (address_) {
@@ -806,6 +720,21 @@ netdev_get_next_hop(const struct netdev *netdev,
     return error;
 }
 
+/* Populates 'sh' with status information.
+ *
+ * Populates 'sh' with 'netdev' specific status information.  This information
+ * may be used to populate the status column of the Interface table as defined
+ * in ovs-vswitchd.conf.db(5). */
+int
+netdev_get_status(const struct netdev *netdev, struct shash *sh)
+{
+    struct netdev_dev *dev = netdev_get_dev(netdev);
+
+    return (dev->netdev_class->get_status
+            ? dev->netdev_class->get_status(netdev, sh)
+            : EOPNOTSUPP);
+}
+
 /* If 'netdev' has an assigned IPv6 address, sets '*in6' to that address and
  * returns 0.  Otherwise, returns a positive errno value and sets '*in6' to
  * all-zero-bits (in6addr_any).
@@ -824,7 +753,7 @@ netdev_get_in6(const struct netdev *netdev, struct in6_addr *in6)
     int error;
 
     error = (netdev_get_dev(netdev)->netdev_class->get_in6
-             ? netdev_get_dev(netdev)->netdev_class->get_in6(netdev, 
+             ? netdev_get_dev(netdev)->netdev_class->get_in6(netdev,
                     in6 ? in6 : &dummy)
              : EOPNOTSUPP);
     if (error && in6) {
@@ -845,7 +774,7 @@ do_update_flags(struct netdev *netdev, enum netdev_flags off,
     enum netdev_flags old_flags;
     int error;
 
-    error = netdev_get_dev(netdev)->netdev_class->update_flags(netdev, 
+    error = netdev_get_dev(netdev)->netdev_class->update_flags(netdev,
                 off & ~on, on, &old_flags);
     if (error) {
         VLOG_WARN_RL(&rl, "failed to %s flags for network device %s: %s",
@@ -917,10 +846,10 @@ netdev_turn_flags_off(struct netdev *netdev, enum netdev_flags flags,
  * ENXIO indicates that there is no ARP table entry for 'ip' on 'netdev'. */
 int
 netdev_arp_lookup(const struct netdev *netdev,
-                  uint32_t ip, uint8_t mac[ETH_ADDR_LEN])
+                  ovs_be32 ip, uint8_t mac[ETH_ADDR_LEN])
 {
     int error = (netdev_get_dev(netdev)->netdev_class->arp_lookup
-                 ? netdev_get_dev(netdev)->netdev_class->arp_lookup(netdev, 
+                 ? netdev_get_dev(netdev)->netdev_class->arp_lookup(netdev,
                         ip, mac)
                  : EOPNOTSUPP);
     if (error) {
@@ -929,19 +858,49 @@ netdev_arp_lookup(const struct netdev *netdev,
     return error;
 }
 
-/* Sets 'carrier' to true if carrier is active (link light is on) on
- * 'netdev'. */
-int
-netdev_get_carrier(const struct netdev *netdev, bool *carrier)
+/* Returns true if carrier is active (link light is on) on 'netdev'. */
+bool
+netdev_get_carrier(const struct netdev *netdev)
 {
-    int error = (netdev_get_dev(netdev)->netdev_class->get_carrier
-                 ? netdev_get_dev(netdev)->netdev_class->get_carrier(netdev, 
-                        carrier)
-                 : EOPNOTSUPP);
+    int error;
+    enum netdev_flags flags;
+    bool carrier;
+
+    netdev_get_flags(netdev, &flags);
+    if (!(flags & NETDEV_UP)) {
+        return false;
+    }
+
+    if (!netdev_get_dev(netdev)->netdev_class->get_carrier) {
+        return true;
+    }
+
+    error = netdev_get_dev(netdev)->netdev_class->get_carrier(netdev,
+                                                              &carrier);
     if (error) {
-        *carrier = false;
+        VLOG_DBG("%s: failed to get network device carrier status, assuming "
+                 "down: %s", netdev_get_name(netdev), strerror(error));
+        carrier = false;
     }
-    return error;
+
+    return carrier;
+}
+
+/* Attempts to force netdev_get_carrier() to poll 'netdev''s MII registers for
+ * link status instead of checking 'netdev''s carrier.  'netdev''s MII
+ * registers will be polled once ever 'interval' milliseconds.  If 'netdev'
+ * does not support MII, another method may be used as a fallback.  If
+ * 'interval' is less than or equal to zero, reverts netdev_get_carrier() to
+ * its normal behavior.
+ *
+ * Returns 0 if successful, otherwise a positive errno value. */
+int
+netdev_set_miimon_interval(struct netdev *netdev, long long int interval)
+{
+    struct netdev_dev *netdev_dev = netdev_get_dev(netdev);
+    return (netdev_dev->netdev_class->set_miimon_interval
+            ? netdev_dev->netdev_class->set_miimon_interval(netdev, interval)
+            : EOPNOTSUPP);
 }
 
 /* Retrieves current device stats for 'netdev'. */
@@ -981,11 +940,302 @@ netdev_set_policing(struct netdev *netdev, uint32_t kbits_rate,
                     uint32_t kbits_burst)
 {
     return (netdev_get_dev(netdev)->netdev_class->set_policing
-            ? netdev_get_dev(netdev)->netdev_class->set_policing(netdev, 
+            ? netdev_get_dev(netdev)->netdev_class->set_policing(netdev,
                     kbits_rate, kbits_burst)
             : EOPNOTSUPP);
 }
 
+/* Adds to 'types' all of the forms of QoS supported by 'netdev', or leaves it
+ * empty if 'netdev' does not support QoS.  Any names added to 'types' should
+ * be documented as valid for the "type" column in the "QoS" table in
+ * vswitchd/vswitch.xml (which is built as ovs-vswitchd.conf.db(8)).
+ *
+ * Every network device supports disabling QoS with a type of "", but this type
+ * will not be added to 'types'.
+ *
+ * The caller must initialize 'types' (e.g. with sset_init()) before calling
+ * this function.  The caller is responsible for destroying 'types' (e.g. with
+ * sset_destroy()) when it is no longer needed.
+ *
+ * Returns 0 if successful, otherwise a positive errno value. */
+int
+netdev_get_qos_types(const struct netdev *netdev, struct sset *types)
+{
+    const struct netdev_class *class = netdev_get_dev(netdev)->netdev_class;
+    return (class->get_qos_types
+            ? class->get_qos_types(netdev, types)
+            : 0);
+}
+
+/* Queries 'netdev' for its capabilities regarding the specified 'type' of QoS,
+ * which should be "" or one of the types returned by netdev_get_qos_types()
+ * for 'netdev'.  Returns 0 if successful, otherwise a positive errno value.
+ * On success, initializes 'caps' with the QoS capabilities; on failure, clears
+ * 'caps' to all zeros. */
+int
+netdev_get_qos_capabilities(const struct netdev *netdev, const char *type,
+                            struct netdev_qos_capabilities *caps)
+{
+    const struct netdev_class *class = netdev_get_dev(netdev)->netdev_class;
+
+    if (*type) {
+        int retval = (class->get_qos_capabilities
+                      ? class->get_qos_capabilities(netdev, type, caps)
+                      : EOPNOTSUPP);
+        if (retval) {
+            memset(caps, 0, sizeof *caps);
+        }
+        return retval;
+    } else {
+        /* Every netdev supports turning off QoS. */
+        memset(caps, 0, sizeof *caps);
+        return 0;
+    }
+}
+
+/* Obtains the number of queues supported by 'netdev' for the specified 'type'
+ * of QoS.  Returns 0 if successful, otherwise a positive errno value.  Stores
+ * the number of queues (zero on failure) in '*n_queuesp'.
+ *
+ * This is just a simple wrapper around netdev_get_qos_capabilities(). */
+int
+netdev_get_n_queues(const struct netdev *netdev,
+                    const char *type, unsigned int *n_queuesp)
+{
+    struct netdev_qos_capabilities caps;
+    int retval;
+
+    retval = netdev_get_qos_capabilities(netdev, type, &caps);
+    *n_queuesp = caps.n_queues;
+    return retval;
+}
+
+/* Queries 'netdev' about its currently configured form of QoS.  If successful,
+ * stores the name of the current form of QoS into '*typep', stores any details
+ * of configuration as string key-value pairs in 'details', and returns 0.  On
+ * failure, sets '*typep' to NULL and returns a positive errno value.
+ *
+ * A '*typep' of "" indicates that QoS is currently disabled on 'netdev'.
+ *
+ * The caller must initialize 'details' as an empty shash (e.g. with
+ * shash_init()) before calling this function.  The caller must free 'details',
+ * including 'data' members, when it is no longer needed (e.g. with
+ * shash_destroy_free_data()).
+ *
+ * The caller must not modify or free '*typep'.
+ *
+ * '*typep' will be one of the types returned by netdev_get_qos_types() for
+ * 'netdev'.  The contents of 'details' should be documented as valid for
+ * '*typep' in the "other_config" column in the "QoS" table in
+ * vswitchd/vswitch.xml (which is built as ovs-vswitchd.conf.db(8)). */
+int
+netdev_get_qos(const struct netdev *netdev,
+               const char **typep, struct shash *details)
+{
+    const struct netdev_class *class = netdev_get_dev(netdev)->netdev_class;
+    int retval;
+
+    if (class->get_qos) {
+        retval = class->get_qos(netdev, typep, details);
+        if (retval) {
+            *typep = NULL;
+            shash_clear_free_data(details);
+        }
+        return retval;
+    } else {
+        /* 'netdev' doesn't support QoS, so report that QoS is disabled. */
+        *typep = "";
+        return 0;
+    }
+}
+
+/* Attempts to reconfigure QoS on 'netdev', changing the form of QoS to 'type'
+ * with details of configuration from 'details'.  Returns 0 if successful,
+ * otherwise a positive errno value.  On error, the previous QoS configuration
+ * is retained.
+ *
+ * When this function changes the type of QoS (not just 'details'), this also
+ * resets all queue configuration for 'netdev' to their defaults (which depend
+ * on the specific type of QoS).  Otherwise, the queue configuration for
+ * 'netdev' is unchanged.
+ *
+ * 'type' should be "" (to disable QoS) or one of the types returned by
+ * netdev_get_qos_types() for 'netdev'.  The contents of 'details' should be
+ * documented as valid for the given 'type' in the "other_config" column in the
+ * "QoS" table in vswitchd/vswitch.xml (which is built as
+ * ovs-vswitchd.conf.db(8)).
+ *
+ * NULL may be specified for 'details' if there are no configuration
+ * details. */
+int
+netdev_set_qos(struct netdev *netdev,
+               const char *type, const struct shash *details)
+{
+    const struct netdev_class *class = netdev_get_dev(netdev)->netdev_class;
+
+    if (!type) {
+        type = "";
+    }
+
+    if (class->set_qos) {
+        if (!details) {
+            static struct shash empty = SHASH_INITIALIZER(&empty);
+            details = &empty;
+        }
+        return class->set_qos(netdev, type, details);
+    } else {
+        return *type ? EOPNOTSUPP : 0;
+    }
+}
+
+/* Queries 'netdev' for information about the queue numbered 'queue_id'.  If
+ * successful, adds that information as string key-value pairs to 'details'.
+ * Returns 0 if successful, otherwise a positive errno value.
+ *
+ * 'queue_id' must be less than the number of queues supported by 'netdev' for
+ * the current form of QoS (e.g. as returned by netdev_get_n_queues(netdev)).
+ *
+ * The returned contents of 'details' should be documented as valid for the
+ * given 'type' in the "other_config" column in the "Queue" table in
+ * vswitchd/vswitch.xml (which is built as ovs-vswitchd.conf.db(8)).
+ *
+ * The caller must initialize 'details' (e.g. with shash_init()) before calling
+ * this function.  The caller must free 'details', including 'data' members,
+ * when it is no longer needed (e.g. with shash_destroy_free_data()). */
+int
+netdev_get_queue(const struct netdev *netdev,
+                 unsigned int queue_id, struct shash *details)
+{
+    const struct netdev_class *class = netdev_get_dev(netdev)->netdev_class;
+    int retval;
+
+    retval = (class->get_queue
+              ? class->get_queue(netdev, queue_id, details)
+              : EOPNOTSUPP);
+    if (retval) {
+        shash_clear_free_data(details);
+    }
+    return retval;
+}
+
+/* Configures the queue numbered 'queue_id' on 'netdev' with the key-value
+ * string pairs in 'details'.  The contents of 'details' should be documented
+ * as valid for the given 'type' in the "other_config" column in the "Queue"
+ * table in vswitchd/vswitch.xml (which is built as ovs-vswitchd.conf.db(8)).
+ * Returns 0 if successful, otherwise a positive errno value.  On failure, the
+ * given queue's configuration should be unmodified.
+ *
+ * 'queue_id' must be less than the number of queues supported by 'netdev' for
+ * the current form of QoS (e.g. as returned by netdev_get_n_queues(netdev)).
+ *
+ * This function does not modify 'details', and the caller retains ownership of
+ * it. */
+int
+netdev_set_queue(struct netdev *netdev,
+                 unsigned int queue_id, const struct shash *details)
+{
+    const struct netdev_class *class = netdev_get_dev(netdev)->netdev_class;
+    return (class->set_queue
+            ? class->set_queue(netdev, queue_id, details)
+            : EOPNOTSUPP);
+}
+
+/* Attempts to delete the queue numbered 'queue_id' from 'netdev'.  Some kinds
+ * of QoS may have a fixed set of queues, in which case attempts to delete them
+ * will fail with EOPNOTSUPP.
+ *
+ * Returns 0 if successful, otherwise a positive errno value.  On failure, the
+ * given queue will be unmodified.
+ *
+ * 'queue_id' must be less than the number of queues supported by 'netdev' for
+ * the current form of QoS (e.g. as returned by
+ * netdev_get_n_queues(netdev)). */
+int
+netdev_delete_queue(struct netdev *netdev, unsigned int queue_id)
+{
+    const struct netdev_class *class = netdev_get_dev(netdev)->netdev_class;
+    return (class->delete_queue
+            ? class->delete_queue(netdev, queue_id)
+            : EOPNOTSUPP);
+}
+
+/* Obtains statistics about 'queue_id' on 'netdev'.  On success, returns 0 and
+ * fills 'stats' with the queue's statistics; individual members of 'stats' may
+ * be set to all-1-bits if the statistic is unavailable.  On failure, returns a
+ * positive errno value and fills 'stats' with all-1-bits. */
+int
+netdev_get_queue_stats(const struct netdev *netdev, unsigned int queue_id,
+                       struct netdev_queue_stats *stats)
+{
+    const struct netdev_class *class = netdev_get_dev(netdev)->netdev_class;
+    int retval;
+
+    retval = (class->get_queue_stats
+              ? class->get_queue_stats(netdev, queue_id, stats)
+              : EOPNOTSUPP);
+    if (retval) {
+        memset(stats, 0xff, sizeof *stats);
+    }
+    return retval;
+}
+
+/* Iterates over all of 'netdev''s queues, calling 'cb' with the queue's ID,
+ * its configuration, and the 'aux' specified by the caller.  The order of
+ * iteration is unspecified, but (when successful) each queue is visited
+ * exactly once.
+ *
+ * Calling this function may be more efficient than calling netdev_get_queue()
+ * for every queue.
+ *
+ * 'cb' must not modify or free the 'details' argument passed in.
+ *
+ * Returns 0 if successful, otherwise a positive errno value.  On error, some
+ * configured queues may not have been included in the iteration. */
+int
+netdev_dump_queues(const struct netdev *netdev,
+                   netdev_dump_queues_cb *cb, void *aux)
+{
+    const struct netdev_class *class = netdev_get_dev(netdev)->netdev_class;
+    return (class->dump_queues
+            ? class->dump_queues(netdev, cb, aux)
+            : EOPNOTSUPP);
+}
+
+/* Iterates over all of 'netdev''s queues, calling 'cb' with the queue's ID,
+ * its statistics, and the 'aux' specified by the caller.  The order of
+ * iteration is unspecified, but (when successful) each queue is visited
+ * exactly once.
+ *
+ * Calling this function may be more efficient than calling
+ * netdev_get_queue_stats() for every queue.
+ *
+ * 'cb' must not modify or free the statistics passed in.
+ *
+ * Returns 0 if successful, otherwise a positive errno value.  On error, some
+ * configured queues may not have been included in the iteration. */
+int
+netdev_dump_queue_stats(const struct netdev *netdev,
+                        netdev_dump_queue_stats_cb *cb, void *aux)
+{
+    const struct netdev_class *class = netdev_get_dev(netdev)->netdev_class;
+    return (class->dump_queue_stats
+            ? class->dump_queue_stats(netdev, cb, aux)
+            : EOPNOTSUPP);
+}
+
+/* Returns a sequence number which indicates changes in one of 'netdev''s
+ * properties.  The returned sequence will be nonzero so that callers have a
+ * value which they may use as a reset when tracking 'netdev'.
+ *
+ * The returned sequence number will change whenever 'netdev''s flags,
+ * features, ethernet address, or carrier changes.  It may change for other
+ * reasons as well, or no reason at all. */
+unsigned int
+netdev_change_seq(const struct netdev *netdev)
+{
+    return netdev_get_dev(netdev)->netdev_class->change_seq(netdev);
+}
+
 /* If 'netdev' is a VLAN network device (e.g. one created with vconfig(8)),
  * sets '*vlan_vid' to the VLAN VID associated with that device and returns 0.
  * Otherwise returns a errno value (specifically ENOENT if 'netdev_name' is the
@@ -995,7 +1245,7 @@ int
 netdev_get_vlan_vid(const struct netdev *netdev, int *vlan_vid)
 {
     int error = (netdev_get_dev(netdev)->netdev_class->get_vlan_vid
-                 ? netdev_get_dev(netdev)->netdev_class->get_vlan_vid(netdev, 
+                 ? netdev_get_dev(netdev)->netdev_class->get_vlan_vid(netdev,
                         vlan_vid)
                  : ENOENT);
     if (error) {
@@ -1010,12 +1260,11 @@ struct netdev *
 netdev_find_dev_by_in4(const struct in_addr *in4)
 {
     struct netdev *netdev;
-    struct svec dev_list = SVEC_EMPTY_INITIALIZER;
-    size_t i;
+    struct sset dev_list = SSET_INITIALIZER(&dev_list);
+    const char *name;
 
     netdev_enumerate(&dev_list);
-    for (i = 0; i < dev_list.n; i++) {
-        const char *name = dev_list.names[i];
+    SSET_FOR_EACH (name, &dev_list) {
         struct in_addr dev_in4;
 
         if (!netdev_open_default(name, &netdev)
@@ -1028,18 +1277,25 @@ netdev_find_dev_by_in4(const struct in_addr *in4)
     netdev = NULL;
 
 exit:
-    svec_destroy(&dev_list);
+    sset_destroy(&dev_list);
     return netdev;
 }
 
-/* Initializes 'netdev_dev' as a netdev device named 'name' of the
- * specified 'netdev_class'.
+/* Initializes 'netdev_dev' as a netdev device named 'name' of the specified
+ * 'netdev_class'.  This function is ordinarily called from a netdev provider's
+ * 'create' function.
+ *
+ * 'args' should be the arguments that were passed to the netdev provider's
+ * 'create'.  If an empty set of arguments was passed, and 'name' is the name
+ * of a network device that existed before the 'create' call, then 'args' may
+ * instead be the configuration for that existing device.
  *
  * This function adds 'netdev_dev' to a netdev-owned shash, so it is
  * very important that 'netdev_dev' only be freed after calling
  * the refcount drops to zero.  */
 void
 netdev_dev_init(struct netdev_dev *netdev_dev, const char *name,
+                const struct shash *args,
                 const struct netdev_class *netdev_class)
 {
     assert(!shash_find(&netdev_dev_shash, name));
@@ -1048,6 +1304,7 @@ netdev_dev_init(struct netdev_dev *netdev_dev, const char *name,
     netdev_dev->netdev_class = netdev_class;
     netdev_dev->name = xstrdup(name);
     netdev_dev->node = shash_add(&netdev_dev_shash, name, netdev_dev);
+    smap_clone(&netdev_dev->args, args);
 }
 
 /* Undoes the results of initialization.
@@ -1065,7 +1322,7 @@ netdev_dev_uninit(struct netdev_dev *netdev_dev, bool destroy)
     assert(!netdev_dev->ref_cnt);
 
     shash_delete(&netdev_dev_shash, netdev_dev->node);
-    update_device_args(netdev_dev, NULL);
+    smap_destroy(&netdev_dev->args);
 
     if (destroy) {
         netdev_dev->netdev_class->destroy(netdev_dev);
@@ -1160,7 +1417,7 @@ netdev_uninit(struct netdev *netdev, bool close)
 }
 
 
-/* Returns the class type of 'netdev'.  
+/* Returns the class type of 'netdev'.
  *
  * The caller must not free the returned value. */
 const char *
@@ -1174,147 +1431,6 @@ netdev_get_dev(const struct netdev *netdev)
 {
     return netdev->netdev_dev;
 }
-
-/* Initializes 'notifier' as a netdev notifier for 'netdev', for which
- * notification will consist of calling 'cb', with auxiliary data 'aux'. */
-void
-netdev_notifier_init(struct netdev_notifier *notifier, struct netdev *netdev,
-                     void (*cb)(struct netdev_notifier *), void *aux)
-{
-    notifier->netdev = netdev;
-    notifier->cb = cb;
-    notifier->aux = aux;
-}
-
-/* Tracks changes in the status of a set of network devices. */
-struct netdev_monitor {
-    struct shash polled_netdevs;
-    struct shash changed_netdevs;
-};
-
-/* Creates and returns a new structure for monitor changes in the status of
- * network devices. */
-struct netdev_monitor *
-netdev_monitor_create(void)
-{
-    struct netdev_monitor *monitor = xmalloc(sizeof *monitor);
-    shash_init(&monitor->polled_netdevs);
-    shash_init(&monitor->changed_netdevs);
-    return monitor;
-}
-
-/* Destroys 'monitor'. */
-void
-netdev_monitor_destroy(struct netdev_monitor *monitor)
-{
-    if (monitor) {
-        struct shash_node *node;
-
-        SHASH_FOR_EACH (node, &monitor->polled_netdevs) {
-            struct netdev_notifier *notifier = node->data;
-            netdev_get_dev(notifier->netdev)->netdev_class->poll_remove(
-                    notifier);
-        }
-
-        shash_destroy(&monitor->polled_netdevs);
-        shash_destroy(&monitor->changed_netdevs);
-        free(monitor);
-    }
-}
-
-static void
-netdev_monitor_cb(struct netdev_notifier *notifier)
-{
-    struct netdev_monitor *monitor = notifier->aux;
-    const char *name = netdev_get_name(notifier->netdev);
-    if (!shash_find(&monitor->changed_netdevs, name)) {
-        shash_add(&monitor->changed_netdevs, name, NULL);
-    }
-}
-
-/* Attempts to add 'netdev' as a netdev monitored by 'monitor'.  Returns 0 if
- * successful, otherwise a positive errno value.
- *
- * Adding a given 'netdev' to a monitor multiple times is equivalent to adding
- * it once. */
-int
-netdev_monitor_add(struct netdev_monitor *monitor, struct netdev *netdev)
-{
-    const char *netdev_name = netdev_get_name(netdev);
-    int error = 0;
-    if (!shash_find(&monitor->polled_netdevs, netdev_name)
-            && netdev_get_dev(netdev)->netdev_class->poll_add)
-    {
-        struct netdev_notifier *notifier;
-        error = netdev_get_dev(netdev)->netdev_class->poll_add(netdev,
-                    netdev_monitor_cb, monitor, &notifier);
-        if (!error) {
-            assert(notifier->netdev == netdev);
-            shash_add(&monitor->polled_netdevs, netdev_name, notifier);
-        }
-    }
-    return error;
-}
-
-/* Removes 'netdev' from the set of netdevs monitored by 'monitor'.  (This has
- * no effect if 'netdev' is not in the set of devices monitored by
- * 'monitor'.) */
-void
-netdev_monitor_remove(struct netdev_monitor *monitor, struct netdev *netdev)
-{
-    const char *netdev_name = netdev_get_name(netdev);
-    struct shash_node *node;
-
-    node = shash_find(&monitor->polled_netdevs, netdev_name);
-    if (node) {
-        /* Cancel future notifications. */
-        struct netdev_notifier *notifier = node->data;
-        netdev_get_dev(netdev)->netdev_class->poll_remove(notifier);
-        shash_delete(&monitor->polled_netdevs, node);
-
-        /* Drop any pending notification. */
-        node = shash_find(&monitor->changed_netdevs, netdev_name);
-        if (node) {
-            shash_delete(&monitor->changed_netdevs, node);
-        }
-    }
-}
-
-/* Checks for changes to netdevs in the set monitored by 'monitor'.  If any of
- * the attributes (Ethernet address, carrier status, speed or peer-advertised
- * speed, flags, etc.) of a network device monitored by 'monitor' has changed,
- * sets '*devnamep' to the name of a device that has changed and returns 0.
- * The caller is responsible for freeing '*devnamep' (with free()).
- *
- * If no devices have changed, sets '*devnamep' to NULL and returns EAGAIN.
- */
-int
-netdev_monitor_poll(struct netdev_monitor *monitor, char **devnamep)
-{
-    struct shash_node *node = shash_first(&monitor->changed_netdevs);
-    if (!node) {
-        *devnamep = NULL;
-        return EAGAIN;
-    } else {
-        *devnamep = xstrdup(node->name);
-        shash_delete(&monitor->changed_netdevs, node);
-        return 0;
-    }
-}
-
-/* Registers with the poll loop to wake up from the next call to poll_block()
- * when netdev_monitor_poll(monitor) would indicate that a device has
- * changed. */
-void
-netdev_monitor_poll_wait(const struct netdev_monitor *monitor)
-{
-    if (!shash_is_empty(&monitor->changed_netdevs)) {
-        poll_immediate_wake();
-    } else {
-        /* XXX Nothing needed here for netdev_linux, but maybe other netdev
-         * classes need help. */
-    }
-}
 
 /* Restore the network device flags on 'netdev' to those that were active
  * before we changed them.  Returns 0 if successful, otherwise a positive
@@ -1340,7 +1456,7 @@ static void
 close_all_netdevs(void *aux OVS_UNUSED)
 {
     struct netdev *netdev, *next;
-    LIST_FOR_EACH_SAFE(netdev, next, struct netdev, node, &netdev_list) {
+    LIST_FOR_EACH_SAFE(netdev, next, node, &netdev_list) {
         netdev_close(netdev);
     }
 }