/*
- * Copyright (c) 2008, 2009, 2010, 2011, 2012, 2013 Nicira, Inc.
+ * Copyright (c) 2008, 2009, 2010, 2011, 2012, 2013, 2014, 2016 Nicira, Inc.
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
#include <string.h>
#include <unistd.h>
+#ifndef _WIN32
+#include <ifaddrs.h>
+#include <net/if.h>
+#include <sys/ioctl.h>
+#include <sys/types.h>
+#endif
+
+#include "cmap.h"
#include "coverage.h"
-#include "dynamic-string.h"
+#include "dpif.h"
+#include "dp-packet.h"
+#include "openvswitch/dynamic-string.h"
#include "fatal-signal.h"
#include "hash.h"
-#include "list.h"
+#include "openvswitch/list.h"
+#include "netdev-dpdk.h"
#include "netdev-provider.h"
#include "netdev-vport.h"
-#include "ofpbuf.h"
+#include "odp-netlink.h"
#include "openflow/openflow.h"
#include "packets.h"
#include "poll-loop.h"
+#include "seq.h"
#include "shash.h"
#include "smap.h"
#include "sset.h"
#include "svec.h"
-#include "vlog.h"
+#include "openvswitch/vlog.h"
+#include "flow.h"
+#include "util.h"
VLOG_DEFINE_THIS_MODULE(netdev);
COVERAGE_DEFINE(netdev_add_router);
COVERAGE_DEFINE(netdev_get_stats);
-static struct shash netdev_classes = SHASH_INITIALIZER(&netdev_classes);
+struct netdev_saved_flags {
+ struct netdev *netdev;
+ struct ovs_list node; /* In struct netdev's saved_flags_list. */
+ enum netdev_flags saved_flags;
+ enum netdev_flags saved_values;
+};
+
+/* Protects 'netdev_shash' and the mutable members of struct netdev. */
+static struct ovs_mutex netdev_mutex = OVS_MUTEX_INITIALIZER;
/* All created network devices. */
-static struct shash netdev_dev_shash = SHASH_INITIALIZER(&netdev_dev_shash);
+static struct shash netdev_shash OVS_GUARDED_BY(netdev_mutex)
+ = SHASH_INITIALIZER(&netdev_shash);
+
+/* Mutual exclusion of */
+static struct ovs_mutex netdev_class_mutex OVS_ACQ_BEFORE(netdev_mutex)
+ = OVS_MUTEX_INITIALIZER;
+
+/* Contains 'struct netdev_registered_class'es. */
+static struct cmap netdev_classes = CMAP_INITIALIZER;
-/* All open network devices. */
-static struct list netdev_list = LIST_INITIALIZER(&netdev_list);
+struct netdev_registered_class {
+ struct cmap_node cmap_node; /* In 'netdev_classes', by class->type. */
+ const struct netdev_class *class;
+
+ /* Number of references: one for the class itself and one for every
+ * instance of the class. */
+ struct ovs_refcount refcnt;
+};
/* This is set pretty low because we probably won't learn anything from the
* additional log messages. */
static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(5, 20);
-static void close_all_netdevs(void *aux OVS_UNUSED);
-static int restore_flags(struct netdev *netdev);
-void update_device_args(struct netdev_dev *, const struct shash *args);
+static void restore_all_flags(void *aux OVS_UNUSED);
+void update_device_args(struct netdev *, const struct shash *args);
+
+int
+netdev_n_txq(const struct netdev *netdev)
+{
+ return netdev->n_txq;
+}
+
+int
+netdev_n_rxq(const struct netdev *netdev)
+{
+ return netdev->n_rxq;
+}
+
+bool
+netdev_is_pmd(const struct netdev *netdev)
+{
+ return netdev->netdev_class->is_pmd;
+}
static void
netdev_initialize(void)
+ OVS_EXCLUDED(netdev_mutex)
{
- static bool inited;
+ static struct ovsthread_once once = OVSTHREAD_ONCE_INITIALIZER;
- if (!inited) {
- inited = true;
+ if (ovsthread_once_start(&once)) {
+ fatal_signal_add_hook(restore_all_flags, NULL, NULL, true);
- fatal_signal_add_hook(close_all_netdevs, NULL, NULL, true);
netdev_vport_patch_register();
-#ifdef LINUX_DATAPATH
+#ifdef __linux__
netdev_register_provider(&netdev_linux_class);
netdev_register_provider(&netdev_internal_class);
netdev_register_provider(&netdev_tap_class);
netdev_vport_tunnel_register();
#endif
-#ifdef __FreeBSD__
+#if defined(__FreeBSD__) || defined(__NetBSD__)
netdev_register_provider(&netdev_tap_class);
netdev_register_provider(&netdev_bsd_class);
#endif
+#ifdef _WIN32
+ netdev_register_provider(&netdev_windows_class);
+ netdev_register_provider(&netdev_internal_class);
+ netdev_vport_tunnel_register();
+#endif
+ ovsthread_once_done(&once);
}
}
* main poll loop. */
void
netdev_run(void)
+ OVS_EXCLUDED(netdev_mutex)
{
- struct shash_node *node;
- SHASH_FOR_EACH(node, &netdev_classes) {
- const struct netdev_class *netdev_class = node->data;
- if (netdev_class->run) {
- netdev_class->run();
+ netdev_initialize();
+
+ struct netdev_registered_class *rc;
+ CMAP_FOR_EACH (rc, cmap_node, &netdev_classes) {
+ if (rc->class->run) {
+ rc->class->run();
}
}
}
* main poll loop. */
void
netdev_wait(void)
+ OVS_EXCLUDED(netdev_mutex)
{
- struct shash_node *node;
- SHASH_FOR_EACH(node, &netdev_classes) {
- const struct netdev_class *netdev_class = node->data;
- if (netdev_class->wait) {
- netdev_class->wait();
+ netdev_initialize();
+
+ struct netdev_registered_class *rc;
+ CMAP_FOR_EACH (rc, cmap_node, &netdev_classes) {
+ if (rc->class->wait) {
+ rc->class->wait();
+ }
+ }
+}
+
+static struct netdev_registered_class *
+netdev_lookup_class(const char *type)
+{
+ struct netdev_registered_class *rc;
+ CMAP_FOR_EACH_WITH_HASH (rc, cmap_node, hash_string(type, 0),
+ &netdev_classes) {
+ if (!strcmp(type, rc->class->type)) {
+ return rc;
}
}
+ return NULL;
}
/* Initializes and registers a new netdev provider. After successful
* registration, new netdevs of that type can be opened using netdev_open(). */
int
netdev_register_provider(const struct netdev_class *new_class)
+ OVS_EXCLUDED(netdev_class_mutex, netdev_mutex)
{
- if (shash_find(&netdev_classes, new_class->type)) {
+ int error;
+
+ ovs_mutex_lock(&netdev_class_mutex);
+ if (netdev_lookup_class(new_class->type)) {
VLOG_WARN("attempted to register duplicate netdev provider: %s",
new_class->type);
- return EEXIST;
- }
-
- if (new_class->init) {
- int error = new_class->init();
- if (error) {
+ error = EEXIST;
+ } else {
+ error = new_class->init ? new_class->init() : 0;
+ if (!error) {
+ struct netdev_registered_class *rc;
+
+ rc = xmalloc(sizeof *rc);
+ cmap_insert(&netdev_classes, &rc->cmap_node,
+ hash_string(new_class->type, 0));
+ rc->class = new_class;
+ ovs_refcount_init(&rc->refcnt);
+ } else {
VLOG_ERR("failed to initialize %s network device class: %s",
- new_class->type, strerror(error));
- return error;
+ new_class->type, ovs_strerror(error));
}
}
+ ovs_mutex_unlock(&netdev_class_mutex);
- shash_add(&netdev_classes, new_class->type, new_class);
-
- return 0;
+ return error;
}
-/* Unregisters a netdev provider. 'type' must have been previously
- * registered and not currently be in use by any netdevs. After unregistration
- * new netdevs of that type cannot be opened using netdev_open(). */
+/* Unregisters a netdev provider. 'type' must have been previously registered
+ * and not currently be in use by any netdevs. After unregistration new
+ * netdevs of that type cannot be opened using netdev_open(). (However, the
+ * provider may still be accessible from other threads until the next RCU grace
+ * period, so the caller must not free or re-register the same netdev_class
+ * until that has passed.) */
int
netdev_unregister_provider(const char *type)
+ OVS_EXCLUDED(netdev_class_mutex, netdev_mutex)
{
- struct shash_node *del_node, *netdev_dev_node;
+ struct netdev_registered_class *rc;
+ int error;
+
+ netdev_initialize();
- del_node = shash_find(&netdev_classes, type);
- if (!del_node) {
+ ovs_mutex_lock(&netdev_class_mutex);
+ rc = netdev_lookup_class(type);
+ if (!rc) {
VLOG_WARN("attempted to unregister a netdev provider that is not "
"registered: %s", type);
- return EAFNOSUPPORT;
- }
-
- SHASH_FOR_EACH(netdev_dev_node, &netdev_dev_shash) {
- struct netdev_dev *netdev_dev = netdev_dev_node->data;
- if (!strcmp(netdev_dev->netdev_class->type, type)) {
- VLOG_WARN("attempted to unregister in use netdev provider: %s",
- type);
- return EBUSY;
- }
+ error = EAFNOSUPPORT;
+ } else if (ovs_refcount_unref(&rc->refcnt) != 1) {
+ ovs_refcount_ref(&rc->refcnt);
+ VLOG_WARN("attempted to unregister in use netdev provider: %s",
+ type);
+ error = EBUSY;
+ } else {
+ cmap_remove(&netdev_classes, &rc->cmap_node,
+ hash_string(rc->class->type, 0));
+ ovsrcu_postpone(free, rc);
+ error = 0;
}
+ ovs_mutex_unlock(&netdev_class_mutex);
- shash_delete(&netdev_classes, del_node);
-
- 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");
+ return error;
}
/* Clears 'types' and enumerates the types of all currently registered netdev
* providers into it. The caller must first initialize the sset. */
void
netdev_enumerate_types(struct sset *types)
+ OVS_EXCLUDED(netdev_mutex)
{
- struct shash_node *node;
-
netdev_initialize();
sset_clear(types);
- SHASH_FOR_EACH(node, &netdev_classes) {
- const struct netdev_class *netdev_class = node->data;
- sset_add(types, netdev_class->type);
+ struct netdev_registered_class *rc;
+ CMAP_FOR_EACH (rc, cmap_node, &netdev_classes) {
+ sset_add(types, rc->class->type);
+ }
+}
+
+/* Check that the network device name is not the same as any of the registered
+ * vport providers' dpif_port name (dpif_port is NULL if the vport provider
+ * does not define it) or the datapath internal port name (e.g. ovs-system).
+ *
+ * Returns true if there is a name conflict, false otherwise. */
+bool
+netdev_is_reserved_name(const char *name)
+ OVS_EXCLUDED(netdev_mutex)
+{
+ netdev_initialize();
+
+ struct netdev_registered_class *rc;
+ CMAP_FOR_EACH (rc, cmap_node, &netdev_classes) {
+ const char *dpif_port = netdev_vport_class_get_dpif_port(rc->class);
+ if (dpif_port && !strncmp(name, dpif_port, strlen(dpif_port))) {
+ return true;
+ }
+ }
+
+ if (!strncmp(name, "ovs-", 4)) {
+ struct sset types;
+ const char *type;
+
+ sset_init(&types);
+ dp_enumerate_types(&types);
+ SSET_FOR_EACH (type, &types) {
+ if (!strcmp(name+4, type)) {
+ sset_destroy(&types);
+ return true;
+ }
+ }
+ sset_destroy(&types);
}
+
+ return false;
}
/* Opens the network device named 'name' (e.g. "eth0") of the specified 'type'
* before they can be used. */
int
netdev_open(const char *name, const char *type, struct netdev **netdevp)
+ OVS_EXCLUDED(netdev_mutex)
{
- struct netdev_dev *netdev_dev;
+ struct netdev *netdev;
int error;
- *netdevp = NULL;
netdev_initialize();
- netdev_dev = shash_find_data(&netdev_dev_shash, name);
-
- if (!netdev_dev) {
- const struct netdev_class *class;
-
- class = netdev_lookup_provider(type);
- if (!class) {
+ ovs_mutex_lock(&netdev_mutex);
+ netdev = shash_find_data(&netdev_shash, name);
+ if (!netdev) {
+ struct netdev_registered_class *rc;
+
+ rc = netdev_lookup_class(type && type[0] ? type : "system");
+ if (rc && ovs_refcount_try_ref_rcu(&rc->refcnt)) {
+ netdev = rc->class->alloc();
+ if (netdev) {
+ memset(netdev, 0, sizeof *netdev);
+ netdev->netdev_class = rc->class;
+ netdev->name = xstrdup(name);
+ netdev->change_seq = 1;
+ netdev->reconfigure_seq = seq_create();
+ netdev->last_reconfigure_seq =
+ seq_read(netdev->reconfigure_seq);
+ netdev->node = shash_add(&netdev_shash, name, netdev);
+
+ /* By default enable one tx and rx queue per netdev. */
+ netdev->n_txq = netdev->netdev_class->send ? 1 : 0;
+ netdev->n_rxq = netdev->netdev_class->rxq_alloc ? 1 : 0;
+
+ ovs_list_init(&netdev->saved_flags_list);
+
+ error = rc->class->construct(netdev);
+ if (!error) {
+ netdev_change_seq_changed(netdev);
+ } else {
+ ovs_refcount_unref(&rc->refcnt);
+ free(netdev->name);
+ ovs_assert(ovs_list_is_empty(&netdev->saved_flags_list));
+ shash_delete(&netdev_shash, netdev->node);
+ rc->class->dealloc(netdev);
+ }
+ } else {
+ error = ENOMEM;
+ }
+ } else {
VLOG_WARN("could not create netdev %s of unknown type %s",
name, type);
- return EAFNOSUPPORT;
+ error = EAFNOSUPPORT;
}
- error = class->create(class, name, &netdev_dev);
- if (error) {
- return error;
- }
- ovs_assert(netdev_dev->netdev_class == class);
-
+ } else {
+ error = 0;
}
- error = netdev_dev->netdev_class->open(netdev_dev, netdevp);
-
if (!error) {
- netdev_dev->ref_cnt++;
+ netdev->ref_cnt++;
+ *netdevp = netdev;
} else {
- if (!netdev_dev->ref_cnt) {
- netdev_dev_uninit(netdev_dev, true);
- }
+ *netdevp = NULL;
}
+ ovs_mutex_unlock(&netdev_mutex);
return error;
}
+/* Returns a reference to 'netdev_' for the caller to own. Returns null if
+ * 'netdev_' is null. */
+struct netdev *
+netdev_ref(const struct netdev *netdev_)
+ OVS_EXCLUDED(netdev_mutex)
+{
+ struct netdev *netdev = CONST_CAST(struct netdev *, netdev_);
+
+ if (netdev) {
+ ovs_mutex_lock(&netdev_mutex);
+ ovs_assert(netdev->ref_cnt > 0);
+ netdev->ref_cnt++;
+ ovs_mutex_unlock(&netdev_mutex);
+ }
+ return netdev;
+}
+
/* Reconfigures the device 'netdev' with 'args'. 'args' may be empty
* or NULL if none are needed. */
int
-netdev_set_config(struct netdev *netdev, const struct smap *args)
+netdev_set_config(struct netdev *netdev, const struct smap *args, char **errp)
+ OVS_EXCLUDED(netdev_mutex)
{
- struct netdev_dev *netdev_dev = netdev_get_dev(netdev);
+ if (netdev->netdev_class->set_config) {
+ const struct smap no_args = SMAP_INITIALIZER(&no_args);
+ int error;
- if (netdev_dev->netdev_class->set_config) {
- struct smap no_args = SMAP_INITIALIZER(&no_args);
- return netdev_dev->netdev_class->set_config(netdev_dev,
- args ? args : &no_args);
+ error = netdev->netdev_class->set_config(netdev,
+ args ? args : &no_args);
+ if (error) {
+ VLOG_WARN_BUF(errp, "%s: could not set configuration (%s)",
+ netdev_get_name(netdev), ovs_strerror(error));
+ }
+ return error;
} else if (args && !smap_is_empty(args)) {
- VLOG_WARN("%s: arguments provided to device that is not configurable",
- netdev_get_name(netdev));
+ VLOG_WARN_BUF(errp, "%s: arguments provided to device that is not configurable",
+ netdev_get_name(netdev));
}
-
return 0;
}
* smap_destroy(). */
int
netdev_get_config(const struct netdev *netdev, struct smap *args)
+ OVS_EXCLUDED(netdev_mutex)
{
- struct netdev_dev *netdev_dev = netdev_get_dev(netdev);
int error;
smap_clear(args);
- if (netdev_dev->netdev_class->get_config) {
- error = netdev_dev->netdev_class->get_config(netdev_dev, args);
+ if (netdev->netdev_class->get_config) {
+ error = netdev->netdev_class->get_config(netdev, args);
if (error) {
smap_clear(args);
}
const struct netdev_tunnel_config *
netdev_get_tunnel_config(const struct netdev *netdev)
+ OVS_EXCLUDED(netdev_mutex)
{
- struct netdev_dev *netdev_dev = netdev_get_dev(netdev);
-
- if (netdev_dev->netdev_class->get_tunnel_config) {
- return netdev_dev->netdev_class->get_tunnel_config(netdev_dev);
+ if (netdev->netdev_class->get_tunnel_config) {
+ return netdev->netdev_class->get_tunnel_config(netdev);
} else {
return NULL;
}
}
+/* Returns the id of the numa node the 'netdev' is on. If the function
+ * is not implemented, returns NETDEV_NUMA_UNSPEC. */
+int
+netdev_get_numa_id(const struct netdev *netdev)
+{
+ if (netdev->netdev_class->get_numa_id) {
+ return netdev->netdev_class->get_numa_id(netdev);
+ } else {
+ return NETDEV_NUMA_UNSPEC;
+ }
+}
+
+static void
+netdev_unref(struct netdev *dev)
+ OVS_RELEASES(netdev_mutex)
+{
+ ovs_assert(dev->ref_cnt);
+ if (!--dev->ref_cnt) {
+ const struct netdev_class *class = dev->netdev_class;
+ struct netdev_registered_class *rc;
+
+ dev->netdev_class->destruct(dev);
+
+ if (dev->node) {
+ shash_delete(&netdev_shash, dev->node);
+ }
+ free(dev->name);
+ seq_destroy(dev->reconfigure_seq);
+ dev->netdev_class->dealloc(dev);
+ ovs_mutex_unlock(&netdev_mutex);
+
+ rc = netdev_lookup_class(class->type);
+ ovs_refcount_unref(&rc->refcnt);
+ } else {
+ ovs_mutex_unlock(&netdev_mutex);
+ }
+}
+
/* Closes and destroys 'netdev'. */
void
netdev_close(struct netdev *netdev)
+ OVS_EXCLUDED(netdev_mutex)
{
if (netdev) {
- struct netdev_dev *netdev_dev = netdev_get_dev(netdev);
-
- ovs_assert(netdev_dev->ref_cnt);
- netdev_dev->ref_cnt--;
- netdev_uninit(netdev, true);
+ ovs_mutex_lock(&netdev_mutex);
+ netdev_unref(netdev);
+ }
+}
- /* If the reference count for the netdev device is zero, destroy it. */
- if (!netdev_dev->ref_cnt) {
- netdev_dev_uninit(netdev_dev, true);
+/* Removes 'netdev' from the global shash and unrefs 'netdev'.
+ *
+ * This allows handler and revalidator threads to still retain references
+ * to this netdev while the main thread changes interface configuration.
+ *
+ * This function should only be called by the main thread when closing
+ * netdevs during user configuration changes. Otherwise, netdev_close should be
+ * used to close netdevs. */
+void
+netdev_remove(struct netdev *netdev)
+{
+ if (netdev) {
+ ovs_mutex_lock(&netdev_mutex);
+ if (netdev->node) {
+ shash_delete(&netdev_shash, netdev->node);
+ netdev->node = NULL;
+ netdev_change_seq_changed(netdev);
}
+ netdev_unref(netdev);
}
}
}
}
-/* Attempts to set up 'netdev' for receiving packets with netdev_recv().
- * Returns 0 if successful, otherwise a positive errno value. EOPNOTSUPP
- * indicates that the network device does not implement packet reception
- * through this interface. */
+/* Attempts to open a netdev_rxq handle for obtaining packets received on
+ * 'netdev'. On success, returns 0 and stores a nonnull 'netdev_rxq *' into
+ * '*rxp'. On failure, returns a positive errno value and stores NULL into
+ * '*rxp'.
+ *
+ * Some kinds of network devices might not support receiving packets. This
+ * function returns EOPNOTSUPP in that case.*/
int
-netdev_listen(struct netdev *netdev)
+netdev_rxq_open(struct netdev *netdev, struct netdev_rxq **rxp, int id)
+ OVS_EXCLUDED(netdev_mutex)
{
- int (*listen)(struct netdev *);
+ int error;
+
+ if (netdev->netdev_class->rxq_alloc && id < netdev->n_rxq) {
+ struct netdev_rxq *rx = netdev->netdev_class->rxq_alloc();
+ if (rx) {
+ rx->netdev = netdev;
+ rx->queue_id = id;
+ error = netdev->netdev_class->rxq_construct(rx);
+ if (!error) {
+ netdev_ref(netdev);
+ *rxp = rx;
+ return 0;
+ }
+ netdev->netdev_class->rxq_dealloc(rx);
+ } else {
+ error = ENOMEM;
+ }
+ } else {
+ error = EOPNOTSUPP;
+ }
- listen = netdev_get_dev(netdev)->netdev_class->listen;
- return listen ? (listen)(netdev) : EOPNOTSUPP;
+ *rxp = NULL;
+ return error;
}
-/* Attempts to receive a packet from 'netdev' into 'buffer', which the caller
- * must have initialized with sufficient room for the packet. The space
- * required to receive any packet is ETH_HEADER_LEN bytes, plus VLAN_HEADER_LEN
- * bytes, plus the device's MTU (which may be retrieved via netdev_get_mtu()).
- * (Some devices do not allow for a VLAN header, in which case VLAN_HEADER_LEN
- * need not be included.)
- *
- * This function can only be expected to return a packet if ->listen() has
- * been called successfully.
+/* Closes 'rx'. */
+void
+netdev_rxq_close(struct netdev_rxq *rx)
+ OVS_EXCLUDED(netdev_mutex)
+{
+ if (rx) {
+ struct netdev *netdev = rx->netdev;
+ netdev->netdev_class->rxq_destruct(rx);
+ netdev->netdev_class->rxq_dealloc(rx);
+ netdev_close(netdev);
+ }
+}
+
+/* Attempts to receive a batch of packets from 'rx'. 'pkts' should point to
+ * the beginning of an array of MAX_RX_BATCH pointers to dp_packet. If
+ * successful, this function stores pointers to up to MAX_RX_BATCH dp_packets
+ * into the array, transferring ownership of the packets to the caller, stores
+ * the number of received packets into '*cnt', and returns 0.
*
- * If a packet is successfully retrieved, returns 0. In this case 'buffer' is
- * guaranteed to contain at least ETH_TOTAL_MIN bytes. Otherwise, returns a
- * positive errno value. Returns EAGAIN immediately if no packet is ready to
- * be returned.
+ * The implementation does not necessarily initialize any non-data members of
+ * 'pkts'. That is, the caller must initialize layer pointers and metadata
+ * itself, if desired, e.g. with pkt_metadata_init() and miniflow_extract().
*
- * Some network devices may not implement support for this function. In such
- * cases this function will always return EOPNOTSUPP. */
+ * Returns EAGAIN immediately if no packet is ready to be received or another
+ * positive errno value if an error was encountered. */
int
-netdev_recv(struct netdev *netdev, struct ofpbuf *buffer)
+netdev_rxq_recv(struct netdev_rxq *rx, struct dp_packet_batch *batch)
{
- int (*recv)(struct netdev *, void *, size_t);
int retval;
- ovs_assert(buffer->size == 0);
- ovs_assert(ofpbuf_tailroom(buffer) >= ETH_TOTAL_MIN);
-
- recv = netdev_get_dev(netdev)->netdev_class->recv;
- retval = (recv
- ? (recv)(netdev, buffer->data, ofpbuf_tailroom(buffer))
- : -EOPNOTSUPP);
- if (retval >= 0) {
+ retval = rx->netdev->netdev_class->rxq_recv(rx, batch);
+ if (!retval) {
COVERAGE_INC(netdev_received);
- buffer->size += retval;
- if (buffer->size < ETH_TOTAL_MIN) {
- ofpbuf_put_zeros(buffer, ETH_TOTAL_MIN - buffer->size);
- }
- return 0;
} else {
- return -retval;
+ batch->count = 0;
}
+ return retval;
}
-/* Registers with the poll loop to wake up from the next call to poll_block()
- * when a packet is ready to be received with netdev_recv() on 'netdev'. */
+/* Arranges for poll_block() to wake up when a packet is ready to be received
+ * on 'rx'. */
void
-netdev_recv_wait(struct netdev *netdev)
+netdev_rxq_wait(struct netdev_rxq *rx)
{
- void (*recv_wait)(struct netdev *);
+ rx->netdev->netdev_class->rxq_wait(rx);
+}
- recv_wait = netdev_get_dev(netdev)->netdev_class->recv_wait;
- if (recv_wait) {
- recv_wait(netdev);
- }
+/* Discards any packets ready to be received on 'rx'. */
+int
+netdev_rxq_drain(struct netdev_rxq *rx)
+{
+ return (rx->netdev->netdev_class->rxq_drain
+ ? rx->netdev->netdev_class->rxq_drain(rx)
+ : 0);
}
-/* Discards all packets waiting to be received from 'netdev'. */
+/* Configures the number of tx queues of 'netdev'. Returns 0 if successful,
+ * otherwise a positive errno value.
+ *
+ * 'n_txq' specifies the exact number of transmission queues to create.
+ * If this function returns successfully, the caller can make 'n_txq'
+ * concurrent calls to netdev_send() (each one with a different 'qid' in the
+ * range [0..'n_txq'-1]).
+ *
+ * The change might not effective immediately. The caller must check if a
+ * reconfiguration is required with netdev_is_reconf_required() and eventually
+ * call netdev_reconfigure() before using the new queues.
+ *
+ * On error, the tx queue configuration is unchanged */
int
-netdev_drain(struct netdev *netdev)
+netdev_set_tx_multiq(struct netdev *netdev, unsigned int n_txq)
{
- int (*drain)(struct netdev *);
+ int error;
+
+ error = (netdev->netdev_class->set_tx_multiq
+ ? netdev->netdev_class->set_tx_multiq(netdev, MAX(n_txq, 1))
+ : EOPNOTSUPP);
+
+ if (error && error != EOPNOTSUPP) {
+ VLOG_DBG_RL(&rl, "failed to set tx queue for network device %s:"
+ "%s", netdev_get_name(netdev), ovs_strerror(error));
+ }
- drain = netdev_get_dev(netdev)->netdev_class->drain;
- return drain ? drain(netdev) : 0;
+ return error;
}
-/* Sends 'buffer' on 'netdev'. Returns 0 if successful, otherwise a positive
- * errno value. Returns EAGAIN without blocking if the packet cannot be queued
- * immediately. Returns EMSGSIZE if a partial packet was transmitted or if
- * the packet is too big or too small to transmit on the device.
+/* Sends 'batch' on 'netdev'. Returns 0 if successful (for every packet),
+ * otherwise a positive errno value. Returns EAGAIN without blocking if
+ * at least one the packets cannot be queued immediately. Returns EMSGSIZE
+ * if a partial packet was transmitted or if a packet is too big or too small
+ * to transmit on the device.
+ *
+ * If the function returns a non-zero value, some of the packets might have
+ * been sent anyway.
*
- * The caller retains ownership of 'buffer' in all cases.
+ * If 'may_steal' is false, the caller retains ownership of all the packets.
+ * If 'may_steal' is true, the caller transfers ownership of all the packets
+ * to the network device, regardless of success.
*
- * The kernel maintains a packet transmission queue, so the caller is not
- * expected to do additional queuing of packets.
+ * The network device is expected to maintain one or more packet
+ * transmission queues, so that the caller does not ordinarily have to
+ * do additional queuing of packets. 'qid' specifies the queue to use
+ * and can be ignored if the implementation does not support multiple
+ * queues.
*
* Some network devices may not implement support for this function. In such
* cases this function will always return EOPNOTSUPP. */
int
-netdev_send(struct netdev *netdev, const struct ofpbuf *buffer)
+netdev_send(struct netdev *netdev, int qid, struct dp_packet_batch *batch,
+ bool may_steal)
{
- int (*send)(struct netdev *, const void *, size_t);
- int error;
+ if (!netdev->netdev_class->send) {
+ dp_packet_delete_batch(batch, may_steal);
+ return EOPNOTSUPP;
+ }
- send = netdev_get_dev(netdev)->netdev_class->send;
- error = send ? (send)(netdev, buffer->data, buffer->size) : EOPNOTSUPP;
+ int error = netdev->netdev_class->send(netdev, qid, batch, may_steal);
if (!error) {
COVERAGE_INC(netdev_sent);
+ if (!may_steal) {
+ dp_packet_batch_reset_cutlen(batch);
+ }
}
return error;
}
+void
+netdev_pop_header(struct netdev *netdev, struct dp_packet_batch *batch)
+{
+ int i, n_cnt = 0;
+ struct dp_packet **buffers = batch->packets;
+
+ if (!netdev->netdev_class->pop_header) {
+ dp_packet_delete_batch(batch, true);
+ batch->count = 0;
+ return;
+ }
+
+ for (i = 0; i < batch->count; i++) {
+ buffers[i] = netdev->netdev_class->pop_header(buffers[i]);
+ if (buffers[i]) {
+ buffers[n_cnt++] = buffers[i];
+ }
+ }
+ batch->count = n_cnt;
+}
+
+void
+netdev_init_tnl_build_header_params(struct netdev_tnl_build_header_params *params,
+ const struct flow *tnl_flow,
+ const struct in6_addr *src,
+ struct eth_addr dmac,
+ struct eth_addr smac)
+{
+ params->flow = tnl_flow;
+ params->dmac = dmac;
+ params->smac = smac;
+ params->s_ip = src;
+ params->is_ipv6 = !IN6_IS_ADDR_V4MAPPED(src);
+}
+
+int netdev_build_header(const struct netdev *netdev,
+ struct ovs_action_push_tnl *data,
+ const struct netdev_tnl_build_header_params *params)
+{
+ if (netdev->netdev_class->build_header) {
+ return netdev->netdev_class->build_header(netdev, data, params);
+ }
+ return EOPNOTSUPP;
+}
+
+int
+netdev_push_header(const struct netdev *netdev,
+ struct dp_packet_batch *batch,
+ const struct ovs_action_push_tnl *data)
+{
+ int i;
+
+ if (!netdev->netdev_class->push_header) {
+ return -EINVAL;
+ }
+
+ for (i = 0; i < batch->count; i++) {
+ netdev->netdev_class->push_header(batch->packets[i], data);
+ pkt_metadata_init(&batch->packets[i]->md, u32_to_odp(data->out_port));
+ }
+
+ return 0;
+}
+
/* Registers with the poll loop to wake up from the next call to poll_block()
* when the packet transmission queue has sufficient room to transmit a packet
* with netdev_send().
*
- * The kernel maintains a packet transmission queue, so the client is not
- * expected to do additional queuing of packets. Thus, this function is
- * unlikely to ever be used. It is included for completeness. */
+ * The network device is expected to maintain one or more packet
+ * transmission queues, so that the caller does not ordinarily have to
+ * do additional queuing of packets. 'qid' specifies the queue to use
+ * and can be ignored if the implementation does not support multiple
+ * queues. */
void
-netdev_send_wait(struct netdev *netdev)
+netdev_send_wait(struct netdev *netdev, int qid)
{
- void (*send_wait)(struct netdev *);
-
- send_wait = netdev_get_dev(netdev)->netdev_class->send_wait;
- if (send_wait) {
- send_wait(netdev);
+ if (netdev->netdev_class->send_wait) {
+ netdev->netdev_class->send_wait(netdev, qid);
}
}
/* Attempts to set 'netdev''s MAC address to 'mac'. Returns 0 if successful,
* otherwise a positive errno value. */
int
-netdev_set_etheraddr(struct netdev *netdev, const uint8_t mac[ETH_ADDR_LEN])
+netdev_set_etheraddr(struct netdev *netdev, const struct eth_addr mac)
{
- return netdev_get_dev(netdev)->netdev_class->set_etheraddr(netdev, mac);
+ return netdev->netdev_class->set_etheraddr(netdev, mac);
}
/* Retrieves 'netdev''s MAC address. If successful, returns 0 and copies the
* the MAC address into 'mac'. On failure, returns a positive errno value and
* clears 'mac' to all-zeros. */
int
-netdev_get_etheraddr(const struct netdev *netdev, uint8_t mac[ETH_ADDR_LEN])
+netdev_get_etheraddr(const struct netdev *netdev, struct eth_addr *mac)
{
- return netdev_get_dev(netdev)->netdev_class->get_etheraddr(netdev, mac);
+ return netdev->netdev_class->get_etheraddr(netdev, mac);
}
/* Returns the name of the network device that 'netdev' represents,
const char *
netdev_get_name(const struct netdev *netdev)
{
- return netdev_get_dev(netdev)->name;
+ return netdev->name;
}
/* Retrieves the MTU of 'netdev'. The MTU is the maximum size of transmitted
int
netdev_get_mtu(const struct netdev *netdev, int *mtup)
{
- const struct netdev_class *class = netdev_get_dev(netdev)->netdev_class;
+ const struct netdev_class *class = netdev->netdev_class;
int error;
error = class->get_mtu ? class->get_mtu(netdev, mtup) : EOPNOTSUPP;
*mtup = 0;
if (error != EOPNOTSUPP) {
VLOG_DBG_RL(&rl, "failed to retrieve MTU for network device %s: "
- "%s", netdev_get_name(netdev), strerror(error));
+ "%s", netdev_get_name(netdev), ovs_strerror(error));
}
}
return error;
int
netdev_set_mtu(const struct netdev *netdev, int mtu)
{
- const struct netdev_class *class = netdev_get_dev(netdev)->netdev_class;
+ const struct netdev_class *class = netdev->netdev_class;
int error;
error = class->set_mtu ? class->set_mtu(netdev, mtu) : EOPNOTSUPP;
if (error && error != EOPNOTSUPP) {
VLOG_DBG_RL(&rl, "failed to set MTU for network device %s: %s",
- netdev_get_name(netdev), strerror(error));
+ netdev_get_name(netdev), ovs_strerror(error));
}
return error;
{
int (*get_ifindex)(const struct netdev *);
- get_ifindex = netdev_get_dev(netdev)->netdev_class->get_ifindex;
+ get_ifindex = netdev->netdev_class->get_ifindex;
return get_ifindex ? get_ifindex(netdev) : -EOPNOTSUPP;
}
peer = &dummy[3];
}
- get_features = netdev_get_dev(netdev)->netdev_class->get_features;
+ get_features = netdev->netdev_class->get_features;
error = get_features
? get_features(netdev, current, advertised, supported,
peer)
netdev_set_advertisements(struct netdev *netdev,
enum netdev_features advertise)
{
- return (netdev_get_dev(netdev)->netdev_class->set_advertisements
- ? netdev_get_dev(netdev)->netdev_class->set_advertisements(
+ return (netdev->netdev_class->set_advertisements
+ ? netdev->netdev_class->set_advertisements(
netdev, advertise)
: EOPNOTSUPP);
}
-/* If 'netdev' has an assigned IPv4 address, sets '*address' to that address
- * and '*netmask' to its netmask and returns 0. Otherwise, returns a positive
- * errno value and sets '*address' to 0 (INADDR_ANY).
- *
- * The following error values have well-defined meanings:
- *
- * - EADDRNOTAVAIL: 'netdev' has no assigned IPv4 address.
- *
- * - 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. */
-int
-netdev_get_in4(const struct netdev *netdev,
- struct in_addr *address_, struct in_addr *netmask_)
-{
- struct in_addr address;
- struct in_addr netmask;
- int error;
-
- error = (netdev_get_dev(netdev)->netdev_class->get_in4
- ? netdev_get_dev(netdev)->netdev_class->get_in4(netdev,
- &address, &netmask)
- : EOPNOTSUPP);
- if (address_) {
- address_->s_addr = error ? 0 : address.s_addr;
- }
- if (netmask_) {
- netmask_->s_addr = error ? 0 : netmask.s_addr;
- }
- return error;
-}
-
/* Assigns 'addr' as 'netdev''s IPv4 address and 'mask' as its netmask. If
* 'addr' is INADDR_ANY, 'netdev''s IPv4 address is cleared. Returns a
* positive errno value. */
int
netdev_set_in4(struct netdev *netdev, struct in_addr addr, struct in_addr mask)
{
- return (netdev_get_dev(netdev)->netdev_class->set_in4
- ? netdev_get_dev(netdev)->netdev_class->set_in4(netdev, addr, mask)
+ return (netdev->netdev_class->set_in4
+ ? netdev->netdev_class->set_in4(netdev, addr, mask)
: EOPNOTSUPP);
}
int
netdev_get_in4_by_name(const char *device_name, struct in_addr *in4)
{
- struct netdev *netdev;
- int error;
+ struct in6_addr *mask, *addr6;
+ int err, n_in6, i;
+ struct netdev *dev;
- error = netdev_open(device_name, "system", &netdev);
- if (error) {
- in4->s_addr = htonl(0);
- return error;
+ err = netdev_open(device_name, NULL, &dev);
+ if (err) {
+ return err;
}
- error = netdev_get_in4(netdev, in4, NULL);
- netdev_close(netdev);
- return error;
+ err = netdev_get_addr_list(dev, &addr6, &mask, &n_in6);
+ if (err) {
+ goto out;
+ }
+
+ for (i = 0; i < n_in6; i++) {
+ if (IN6_IS_ADDR_V4MAPPED(&addr6[i])) {
+ in4->s_addr = in6_addr_get_mapped_ipv4(&addr6[i]);
+ goto out;
+ }
+ }
+ err = -ENOENT;
+out:
+ free(addr6);
+ free(mask);
+ netdev_close(dev);
+ return err;
+
}
/* Adds 'router' as a default IP gateway for the TCP/IP stack that corresponds
netdev_add_router(struct netdev *netdev, struct in_addr router)
{
COVERAGE_INC(netdev_add_router);
- return (netdev_get_dev(netdev)->netdev_class->add_router
- ? netdev_get_dev(netdev)->netdev_class->add_router(netdev, router)
+ return (netdev->netdev_class->add_router
+ ? netdev->netdev_class->add_router(netdev, router)
: EOPNOTSUPP);
}
const struct in_addr *host, struct in_addr *next_hop,
char **netdev_name)
{
- int error = (netdev_get_dev(netdev)->netdev_class->get_next_hop
- ? netdev_get_dev(netdev)->netdev_class->get_next_hop(
+ int error = (netdev->netdev_class->get_next_hop
+ ? netdev->netdev_class->get_next_hop(
host, next_hop, netdev_name)
: EOPNOTSUPP);
if (error) {
int
netdev_get_status(const struct netdev *netdev, struct smap *smap)
{
- struct netdev_dev *dev = netdev_get_dev(netdev);
-
- return (dev->netdev_class->get_status
- ? dev->netdev_class->get_status(netdev, smap)
+ return (netdev->netdev_class->get_status
+ ? netdev->netdev_class->get_status(netdev, smap)
: 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).
+/* Returns all assigned IP address to 'netdev' and returns 0.
+ * API allocates array of address and masks and set it to
+ * '*addr' and '*mask'.
+ * Otherwise, returns a positive errno value and sets '*addr', '*mask
+ * and '*n_addr' to NULL.
*
* The following error values have well-defined meanings:
*
*
* - EOPNOTSUPP: No IPv6 network stack attached to 'netdev'.
*
- * 'in6' may be null, in which case the address itself is not reported. */
+ * 'addr' may be null, in which case the address itself is not reported. */
int
-netdev_get_in6(const struct netdev *netdev, struct in6_addr *in6)
+netdev_get_addr_list(const struct netdev *netdev, struct in6_addr **addr,
+ struct in6_addr **mask, int *n_addr)
{
- struct in6_addr dummy;
int error;
- error = (netdev_get_dev(netdev)->netdev_class->get_in6
- ? netdev_get_dev(netdev)->netdev_class->get_in6(netdev,
- in6 ? in6 : &dummy)
- : EOPNOTSUPP);
- if (error && in6) {
- memset(in6, 0, sizeof *in6);
+ error = (netdev->netdev_class->get_addr_list
+ ? netdev->netdev_class->get_addr_list(netdev, addr, mask, n_addr): EOPNOTSUPP);
+ if (error && addr) {
+ *addr = NULL;
+ *mask = NULL;
+ *n_addr = 0;
}
+
return error;
}
/* On 'netdev', turns off the flags in 'off' and then turns on the flags in
- * 'on'. If 'permanent' is true, the changes will persist; otherwise, they
- * will be reverted when 'netdev' is closed or the program exits. Returns 0 if
- * successful, otherwise a positive errno value. */
+ * 'on'. Returns 0 if successful, otherwise a positive errno value. */
static int
do_update_flags(struct netdev *netdev, enum netdev_flags off,
enum netdev_flags on, enum netdev_flags *old_flagsp,
- bool permanent)
+ struct netdev_saved_flags **sfp)
+ OVS_EXCLUDED(netdev_mutex)
{
+ struct netdev_saved_flags *sf = NULL;
enum netdev_flags old_flags;
int error;
- error = netdev_get_dev(netdev)->netdev_class->update_flags(netdev,
- off & ~on, on, &old_flags);
+ error = 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",
off || on ? "set" : "get", netdev_get_name(netdev),
- strerror(error));
+ ovs_strerror(error));
old_flags = 0;
- } else if ((off || on) && !permanent) {
+ } else if ((off || on) && sfp) {
enum netdev_flags new_flags = (old_flags & ~off) | on;
enum netdev_flags changed_flags = old_flags ^ new_flags;
if (changed_flags) {
- if (!netdev->changed_flags) {
- netdev->save_flags = old_flags;
- }
- netdev->changed_flags |= changed_flags;
+ ovs_mutex_lock(&netdev_mutex);
+ *sfp = sf = xmalloc(sizeof *sf);
+ sf->netdev = netdev;
+ ovs_list_push_front(&netdev->saved_flags_list, &sf->node);
+ sf->saved_flags = changed_flags;
+ sf->saved_values = changed_flags & new_flags;
+
+ netdev->ref_cnt++;
+ ovs_mutex_unlock(&netdev_mutex);
}
}
+
if (old_flagsp) {
*old_flagsp = old_flags;
}
+ if (sfp) {
+ *sfp = sf;
+ }
+
return error;
}
netdev_get_flags(const struct netdev *netdev_, enum netdev_flags *flagsp)
{
struct netdev *netdev = CONST_CAST(struct netdev *, netdev_);
- return do_update_flags(netdev, 0, 0, flagsp, false);
+ return do_update_flags(netdev, 0, 0, flagsp, NULL);
}
/* Sets the flags for 'netdev' to 'flags'.
- * If 'permanent' is true, the changes will persist; otherwise, they
- * will be reverted when 'netdev' is closed or the program exits.
* Returns 0 if successful, otherwise a positive errno value. */
int
netdev_set_flags(struct netdev *netdev, enum netdev_flags flags,
- bool permanent)
+ struct netdev_saved_flags **sfp)
{
- return do_update_flags(netdev, -1, flags, NULL, permanent);
+ return do_update_flags(netdev, -1, flags, NULL, sfp);
}
-/* Turns on the specified 'flags' on 'netdev'.
- * If 'permanent' is true, the changes will persist; otherwise, they
- * will be reverted when 'netdev' is closed or the program exits.
- * Returns 0 if successful, otherwise a positive errno value. */
+/* Turns on the specified 'flags' on 'netdev':
+ *
+ * - On success, returns 0. If 'sfp' is nonnull, sets '*sfp' to a newly
+ * allocated 'struct netdev_saved_flags *' that may be passed to
+ * netdev_restore_flags() to restore the original values of 'flags' on
+ * 'netdev' (this will happen automatically at program termination if
+ * netdev_restore_flags() is never called) , or to NULL if no flags were
+ * actually changed.
+ *
+ * - On failure, returns a positive errno value. If 'sfp' is nonnull, sets
+ * '*sfp' to NULL. */
int
netdev_turn_flags_on(struct netdev *netdev, enum netdev_flags flags,
- bool permanent)
+ struct netdev_saved_flags **sfp)
{
- return do_update_flags(netdev, 0, flags, NULL, permanent);
+ return do_update_flags(netdev, 0, flags, NULL, sfp);
}
-/* Turns off the specified 'flags' on 'netdev'.
- * If 'permanent' is true, the changes will persist; otherwise, they
- * will be reverted when 'netdev' is closed or the program exits.
- * Returns 0 if successful, otherwise a positive errno value. */
+/* Turns off the specified 'flags' on 'netdev'. See netdev_turn_flags_on() for
+ * details of the interface. */
int
netdev_turn_flags_off(struct netdev *netdev, enum netdev_flags flags,
- bool permanent)
+ struct netdev_saved_flags **sfp)
{
- return do_update_flags(netdev, flags, 0, NULL, permanent);
+ return do_update_flags(netdev, flags, 0, NULL, sfp);
+}
+
+/* Restores the flags that were saved in 'sf', and destroys 'sf'.
+ * Does nothing if 'sf' is NULL. */
+void
+netdev_restore_flags(struct netdev_saved_flags *sf)
+ OVS_EXCLUDED(netdev_mutex)
+{
+ if (sf) {
+ struct netdev *netdev = sf->netdev;
+ enum netdev_flags old_flags;
+
+ netdev->netdev_class->update_flags(netdev,
+ sf->saved_flags & sf->saved_values,
+ sf->saved_flags & ~sf->saved_values,
+ &old_flags);
+
+ ovs_mutex_lock(&netdev_mutex);
+ ovs_list_remove(&sf->node);
+ free(sf);
+ netdev_unref(netdev);
+ }
}
/* Looks up the ARP table entry for 'ip' on 'netdev'. If one exists and can be
* ENXIO indicates that there is no ARP table entry for 'ip' on 'netdev'. */
int
netdev_arp_lookup(const struct netdev *netdev,
- ovs_be32 ip, uint8_t mac[ETH_ADDR_LEN])
+ ovs_be32 ip, struct eth_addr *mac)
{
- int error = (netdev_get_dev(netdev)->netdev_class->arp_lookup
- ? netdev_get_dev(netdev)->netdev_class->arp_lookup(netdev,
- ip, mac)
+ int error = (netdev->netdev_class->arp_lookup
+ ? netdev->netdev_class->arp_lookup(netdev, ip, mac)
: EOPNOTSUPP);
if (error) {
- memset(mac, 0, ETH_ADDR_LEN);
+ *mac = eth_addr_zero;
}
return error;
}
return false;
}
- if (!netdev_get_dev(netdev)->netdev_class->get_carrier) {
+ if (!netdev->netdev_class->get_carrier) {
return true;
}
- error = netdev_get_dev(netdev)->netdev_class->get_carrier(netdev,
- &carrier);
+ error = netdev->netdev_class->get_carrier(netdev, &carrier);
if (error) {
VLOG_DBG("%s: failed to get network device carrier status, assuming "
- "down: %s", netdev_get_name(netdev), strerror(error));
+ "down: %s", netdev_get_name(netdev), ovs_strerror(error));
carrier = false;
}
long long int
netdev_get_carrier_resets(const struct netdev *netdev)
{
- return (netdev_get_dev(netdev)->netdev_class->get_carrier_resets
- ? netdev_get_dev(netdev)->netdev_class->get_carrier_resets(netdev)
+ return (netdev->netdev_class->get_carrier_resets
+ ? netdev->netdev_class->get_carrier_resets(netdev)
: 0);
}
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)
+ return (netdev->netdev_class->set_miimon_interval
+ ? netdev->netdev_class->set_miimon_interval(netdev, interval)
: EOPNOTSUPP);
}
{
int error;
+ /* Statistics are initialized before passing it to particular device
+ * implementation so all values are filtered out by default. */
+ memset(stats, 0xFF, sizeof *stats);
+
COVERAGE_INC(netdev_get_stats);
- error = (netdev_get_dev(netdev)->netdev_class->get_stats
- ? netdev_get_dev(netdev)->netdev_class->get_stats(netdev, stats)
+ error = (netdev->netdev_class->get_stats
+ ? netdev->netdev_class->get_stats(netdev, stats)
: EOPNOTSUPP);
if (error) {
+ /* In case of error all statistics are filtered out */
memset(stats, 0xff, sizeof *stats);
}
return error;
}
-/* Attempts to change the stats for 'netdev' to those provided in 'stats'.
- * Returns 0 if successful, otherwise a positive errno value.
- *
- * This will probably fail for most network devices. Some devices might only
- * allow setting their stats to 0. */
-int
-netdev_set_stats(struct netdev *netdev, const struct netdev_stats *stats)
-{
- return (netdev_get_dev(netdev)->netdev_class->set_stats
- ? netdev_get_dev(netdev)->netdev_class->set_stats(netdev, stats)
- : EOPNOTSUPP);
-}
-
/* Attempts to set input rate limiting (policing) policy, such that up to
* 'kbits_rate' kbps of traffic is accepted, with a maximum accumulative burst
* size of 'kbits' kb. */
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,
+ return (netdev->netdev_class->set_policing
+ ? netdev->netdev_class->set_policing(netdev,
kbits_rate, kbits_burst)
: EOPNOTSUPP);
}
int
netdev_get_qos_types(const struct netdev *netdev, struct sset *types)
{
- const struct netdev_class *class = netdev_get_dev(netdev)->netdev_class;
+ const struct netdev_class *class = netdev->netdev_class;
return (class->get_qos_types
? class->get_qos_types(netdev, types)
: 0);
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;
+ const struct netdev_class *class = netdev->netdev_class;
if (*type) {
int retval = (class->get_qos_capabilities
netdev_get_qos(const struct netdev *netdev,
const char **typep, struct smap *details)
{
- const struct netdev_class *class = netdev_get_dev(netdev)->netdev_class;
+ const struct netdev_class *class = netdev->netdev_class;
int retval;
if (class->get_qos) {
netdev_set_qos(struct netdev *netdev,
const char *type, const struct smap *details)
{
- const struct netdev_class *class = netdev_get_dev(netdev)->netdev_class;
+ const struct netdev_class *class = netdev->netdev_class;
if (!type) {
type = "";
if (class->set_qos) {
if (!details) {
- static struct smap empty = SMAP_INITIALIZER(&empty);
+ static const struct smap empty = SMAP_INITIALIZER(&empty);
details = ∅
}
return class->set_qos(netdev, type, details);
netdev_get_queue(const struct netdev *netdev,
unsigned int queue_id, struct smap *details)
{
- const struct netdev_class *class = netdev_get_dev(netdev)->netdev_class;
+ const struct netdev_class *class = netdev->netdev_class;
int retval;
retval = (class->get_queue
netdev_set_queue(struct netdev *netdev,
unsigned int queue_id, const struct smap *details)
{
- const struct netdev_class *class = netdev_get_dev(netdev)->netdev_class;
+ const struct netdev_class *class = netdev->netdev_class;
return (class->set_queue
? class->set_queue(netdev, queue_id, details)
: EOPNOTSUPP);
int
netdev_delete_queue(struct netdev *netdev, unsigned int queue_id)
{
- const struct netdev_class *class = netdev_get_dev(netdev)->netdev_class;
+ const struct netdev_class *class = 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. */
+ * positive errno value and fills 'stats' with values indicating unsupported
+ * statistics. */
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;
+ const struct netdev_class *class = 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);
+ stats->tx_bytes = UINT64_MAX;
+ stats->tx_packets = UINT64_MAX;
+ stats->tx_errors = UINT64_MAX;
+ stats->created = LLONG_MIN;
}
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.
+/* Initializes 'dump' to begin dumping the queues in a netdev.
+ *
+ * This function provides no status indication. An error status for the entire
+ * dump operation is provided when it is completed by calling
+ * netdev_queue_dump_done().
+ */
+void
+netdev_queue_dump_start(struct netdev_queue_dump *dump,
+ const struct netdev *netdev)
+{
+ dump->netdev = netdev_ref(netdev);
+ if (netdev->netdev_class->queue_dump_start) {
+ dump->error = netdev->netdev_class->queue_dump_start(netdev,
+ &dump->state);
+ } else {
+ dump->error = EOPNOTSUPP;
+ }
+}
+
+/* Attempts to retrieve another queue from 'dump', which must have been
+ * initialized with netdev_queue_dump_start(). On success, stores a new queue
+ * ID into '*queue_id', fills 'details' with configuration details for the
+ * queue, and returns true. On failure, returns false.
*
- * Calling this function may be more efficient than calling netdev_get_queue()
- * for every queue.
+ * Queues are not necessarily dumped in increasing order of queue ID (or any
+ * other predictable order).
*
- * 'cb' must not modify or free the 'details' argument passed in. It may
- * delete or modify the queue passed in as its 'queue_id' argument. It may
- * modify but must not delete any other queue within 'netdev'. 'cb' should not
- * add new queues because this may cause some queues to be visited twice or not
- * at all.
+ * Failure might indicate an actual error or merely that the last queue has
+ * been dumped. An error status for the entire dump operation is provided when
+ * it is completed by calling netdev_queue_dump_done().
*
- * Returns 0 if successful, otherwise a positive errno value. On error, some
- * configured queues may not have been included in the iteration. */
+ * 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 smap_init()) before calling
+ * this function. This function will clear and replace its contents. The
+ * caller must free 'details' when it is no longer needed (e.g. with
+ * smap_destroy()). */
+bool
+netdev_queue_dump_next(struct netdev_queue_dump *dump,
+ unsigned int *queue_id, struct smap *details)
+{
+ const struct netdev *netdev = dump->netdev;
+
+ if (dump->error) {
+ return false;
+ }
+
+ dump->error = netdev->netdev_class->queue_dump_next(netdev, dump->state,
+ queue_id, details);
+
+ if (dump->error) {
+ netdev->netdev_class->queue_dump_done(netdev, dump->state);
+ return false;
+ }
+ return true;
+}
+
+/* Completes queue table dump operation 'dump', which must have been
+ * initialized with netdev_queue_dump_start(). Returns 0 if the dump operation
+ * was error-free, otherwise a positive errno value describing the problem. */
int
-netdev_dump_queues(const struct netdev *netdev,
- netdev_dump_queues_cb *cb, void *aux)
+netdev_queue_dump_done(struct netdev_queue_dump *dump)
{
- const struct netdev_class *class = netdev_get_dev(netdev)->netdev_class;
- return (class->dump_queues
- ? class->dump_queues(netdev, cb, aux)
- : EOPNOTSUPP);
+ const struct netdev *netdev = dump->netdev;
+ if (!dump->error && netdev->netdev_class->queue_dump_done) {
+ dump->error = netdev->netdev_class->queue_dump_done(netdev,
+ dump->state);
+ }
+ netdev_close(dump->netdev);
+ return dump->error == EOF ? 0 : dump->error;
}
/* Iterates over all of 'netdev''s queues, calling 'cb' with the queue's ID,
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;
+ const struct netdev_class *class = 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'.
+\f
+/* Returns the class type of '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)
+ * The caller must not free the returned value. */
+const char *
+netdev_get_type(const struct netdev *netdev)
{
- return netdev_get_dev(netdev)->netdev_class->change_seq(netdev);
+ return netdev->netdev_class->type;
}
-\f
-/* 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.
+
+/* Returns the class associated with 'netdev'. */
+const struct netdev_class *
+netdev_get_class(const struct netdev *netdev)
+{
+ return netdev->netdev_class;
+}
+
+/* Returns the netdev with 'name' or NULL if there is none.
*
- * 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 netdev_class *netdev_class)
+ * The caller must free the returned netdev with netdev_close(). */
+struct netdev *
+netdev_from_name(const char *name)
+ OVS_EXCLUDED(netdev_mutex)
{
- ovs_assert(!shash_find(&netdev_dev_shash, name));
+ struct netdev *netdev;
+
+ ovs_mutex_lock(&netdev_mutex);
+ netdev = shash_find_data(&netdev_shash, name);
+ if (netdev) {
+ netdev->ref_cnt++;
+ }
+ ovs_mutex_unlock(&netdev_mutex);
- memset(netdev_dev, 0, sizeof *netdev_dev);
- netdev_dev->netdev_class = netdev_class;
- netdev_dev->name = xstrdup(name);
- netdev_dev->node = shash_add(&netdev_dev_shash, name, netdev_dev);
+ return netdev;
}
-/* Undoes the results of initialization.
+/* Fills 'device_list' with devices that match 'netdev_class'.
*
- * Normally this function does not need to be called as netdev_close has
- * the same effect when the refcount drops to zero.
- * However, it may be called by providers due to an error on creation
- * that occurs after initialization. It this case netdev_close() would
- * never be called. */
+ * The caller is responsible for initializing and destroying 'device_list' and
+ * must close each device on the list. */
void
-netdev_dev_uninit(struct netdev_dev *netdev_dev, bool destroy)
+netdev_get_devices(const struct netdev_class *netdev_class,
+ struct shash *device_list)
+ OVS_EXCLUDED(netdev_mutex)
{
- char *name = netdev_dev->name;
-
- ovs_assert(!netdev_dev->ref_cnt);
+ struct shash_node *node;
- shash_delete(&netdev_dev_shash, netdev_dev->node);
+ ovs_mutex_lock(&netdev_mutex);
+ SHASH_FOR_EACH (node, &netdev_shash) {
+ struct netdev *dev = node->data;
- if (destroy) {
- netdev_dev->netdev_class->destroy(netdev_dev);
+ if (dev->netdev_class == netdev_class) {
+ dev->ref_cnt++;
+ shash_add(device_list, node->name, node->data);
+ }
}
- free(name);
+ ovs_mutex_unlock(&netdev_mutex);
}
-/* Returns the class type of 'netdev_dev'.
+/* Extracts pointers to all 'netdev-vports' into an array 'vports'
+ * and returns it. Stores the size of the array into '*size'.
*
- * The caller must not free the returned value. */
-const char *
-netdev_dev_get_type(const struct netdev_dev *netdev_dev)
+ * The caller is responsible for freeing 'vports' and must close
+ * each 'netdev-vport' in the list. */
+struct netdev **
+netdev_get_vports(size_t *size)
+ OVS_EXCLUDED(netdev_mutex)
{
- return netdev_dev->netdev_class->type;
+ struct netdev **vports;
+ struct shash_node *node;
+ size_t n = 0;
+
+ if (!size) {
+ return NULL;
+ }
+
+ /* Explicitly allocates big enough chunk of memory. */
+ vports = xmalloc(shash_count(&netdev_shash) * sizeof *vports);
+ ovs_mutex_lock(&netdev_mutex);
+ SHASH_FOR_EACH (node, &netdev_shash) {
+ struct netdev *dev = node->data;
+
+ if (netdev_vport_is_vport_class(dev->netdev_class)) {
+ dev->ref_cnt++;
+ vports[n] = dev;
+ n++;
+ }
+ }
+ ovs_mutex_unlock(&netdev_mutex);
+ *size = n;
+
+ return vports;
}
-/* Returns the class associated with 'netdev_dev'. */
-const struct netdev_class *
-netdev_dev_get_class(const struct netdev_dev *netdev_dev)
+const char *
+netdev_get_type_from_name(const char *name)
{
- return netdev_dev->netdev_class;
+ struct netdev *dev = netdev_from_name(name);
+ const char *type = dev ? netdev_get_type(dev) : NULL;
+ netdev_close(dev);
+ return type;
+}
+\f
+struct netdev *
+netdev_rxq_get_netdev(const struct netdev_rxq *rx)
+{
+ ovs_assert(rx->netdev->ref_cnt > 0);
+ return rx->netdev;
}
-/* Returns the name of 'netdev_dev'.
- *
- * The caller must not free the returned value. */
const char *
-netdev_dev_get_name(const struct netdev_dev *netdev_dev)
+netdev_rxq_get_name(const struct netdev_rxq *rx)
{
- return netdev_dev->name;
+ return netdev_get_name(netdev_rxq_get_netdev(rx));
}
-/* Returns the netdev_dev with 'name' or NULL if there is none.
- *
- * The caller must not free the returned value. */
-struct netdev_dev *
-netdev_dev_from_name(const char *name)
+int
+netdev_rxq_get_queue_id(const struct netdev_rxq *rx)
{
- return shash_find_data(&netdev_dev_shash, name);
+ return rx->queue_id;
}
-/* Fills 'device_list' with devices that match 'netdev_class'.
- *
- * The caller is responsible for initializing and destroying 'device_list'
- * but the contained netdev_devs must not be freed. */
-void
-netdev_dev_get_devices(const struct netdev_class *netdev_class,
- struct shash *device_list)
+static void
+restore_all_flags(void *aux OVS_UNUSED)
{
struct shash_node *node;
- SHASH_FOR_EACH (node, &netdev_dev_shash) {
- struct netdev_dev *dev = node->data;
- if (dev->netdev_class == netdev_class) {
- shash_add(device_list, node->name, node->data);
+ SHASH_FOR_EACH (node, &netdev_shash) {
+ struct netdev *netdev = node->data;
+ const struct netdev_saved_flags *sf;
+ enum netdev_flags saved_values;
+ enum netdev_flags saved_flags;
+
+ saved_values = saved_flags = 0;
+ LIST_FOR_EACH (sf, node, &netdev->saved_flags_list) {
+ saved_flags |= sf->saved_flags;
+ saved_values &= ~sf->saved_flags;
+ saved_values |= sf->saved_flags & sf->saved_values;
+ }
+ if (saved_flags) {
+ enum netdev_flags old_flags;
+
+ netdev->netdev_class->update_flags(netdev,
+ saved_flags & saved_values,
+ saved_flags & ~saved_values,
+ &old_flags);
}
}
}
-/* Initializes 'netdev' as a instance of the netdev_dev.
- *
- * This function adds 'netdev' to a netdev-owned linked list, so it is very
- * important that 'netdev' only be freed after calling netdev_close(). */
-void
-netdev_init(struct netdev *netdev, struct netdev_dev *netdev_dev)
+uint64_t
+netdev_get_change_seq(const struct netdev *netdev)
{
- memset(netdev, 0, sizeof *netdev);
- netdev->netdev_dev = netdev_dev;
- list_push_back(&netdev_list, &netdev->node);
+ return netdev->change_seq;
}
-/* Undoes the results of initialization.
- *
- * Normally this function only needs to be called from netdev_close().
- * However, it may be called by providers due to an error on opening
- * that occurs after initialization. It this case netdev_close() would
- * never be called. */
+#ifndef _WIN32
+/* This implementation is shared by Linux and BSD. */
+
+static struct ifaddrs *if_addr_list;
+static struct ovs_mutex if_addr_list_lock = OVS_MUTEX_INITIALIZER;
+
void
-netdev_uninit(struct netdev *netdev, bool close)
+netdev_get_addrs_list_flush(void)
{
- /* Restore flags that we changed, if any. */
- int error = restore_flags(netdev);
- list_remove(&netdev->node);
- if (error) {
- VLOG_WARN("failed to restore network device flags on %s: %s",
- netdev_get_name(netdev), strerror(error));
+ ovs_mutex_lock(&if_addr_list_lock);
+ if (if_addr_list) {
+ freeifaddrs(if_addr_list);
+ if_addr_list = NULL;
}
+ ovs_mutex_unlock(&if_addr_list_lock);
+}
- if (close) {
- netdev_get_dev(netdev)->netdev_class->close(netdev);
+int
+netdev_get_addrs(const char dev[], struct in6_addr **paddr,
+ struct in6_addr **pmask, int *n_in)
+{
+ struct in6_addr *addr_array, *mask_array;
+ const struct ifaddrs *ifa;
+ int cnt = 0, i = 0;
+
+ ovs_mutex_lock(&if_addr_list_lock);
+ if (!if_addr_list) {
+ int err;
+
+ err = getifaddrs(&if_addr_list);
+ if (err) {
+ ovs_mutex_unlock(&if_addr_list_lock);
+ return -err;
+ }
}
-}
+ for (ifa = if_addr_list; ifa; ifa = ifa->ifa_next) {
+ if (ifa->ifa_addr != NULL) {
+ int family;
-/* Returns the class type of 'netdev'.
- *
- * The caller must not free the returned value. */
-const char *
-netdev_get_type(const struct netdev *netdev)
-{
- return netdev_get_dev(netdev)->netdev_class->type;
-}
+ family = ifa->ifa_addr->sa_family;
+ if (family == AF_INET || family == AF_INET6) {
+ if (!strncmp(ifa->ifa_name, dev, IFNAMSIZ)) {
+ cnt++;
+ }
+ }
+ }
+ }
+ if (!cnt) {
+ ovs_mutex_unlock(&if_addr_list_lock);
+ return EADDRNOTAVAIL;
+ }
+ addr_array = xzalloc(sizeof *addr_array * cnt);
+ mask_array = xzalloc(sizeof *mask_array * cnt);
+ for (ifa = if_addr_list; ifa; ifa = ifa->ifa_next) {
+ int family;
-const char *
-netdev_get_type_from_name(const char *name)
-{
- const struct netdev_dev *dev = netdev_dev_from_name(name);
- return dev ? netdev_dev_get_type(dev) : NULL;
+ if (strncmp(ifa->ifa_name, dev, IFNAMSIZ) || ifa->ifa_addr == NULL) {
+ continue;
+ }
+
+ family = ifa->ifa_addr->sa_family;
+ if (family == AF_INET) {
+ const struct sockaddr_in *sin;
+
+ sin = ALIGNED_CAST(const struct sockaddr_in *, ifa->ifa_addr);
+ in6_addr_set_mapped_ipv4(&addr_array[i], sin->sin_addr.s_addr);
+ sin = (struct sockaddr_in *) &ifa->ifa_netmask;
+ in6_addr_set_mapped_ipv4(&mask_array[i], sin->sin_addr.s_addr);
+ i++;
+ } else if (family == AF_INET6) {
+ const struct sockaddr_in6 *sin6;
+
+ sin6 = ALIGNED_CAST(const struct sockaddr_in6 *, ifa->ifa_addr);
+ memcpy(&addr_array[i], &sin6->sin6_addr, sizeof *addr_array);
+ sin6 = (struct sockaddr_in6 *) &ifa->ifa_netmask;
+ memcpy(&mask_array[i], &sin6->sin6_addr, sizeof *mask_array);
+ i++;
+ }
+ }
+ ovs_mutex_unlock(&if_addr_list_lock);
+ if (paddr) {
+ *n_in = cnt;
+ *paddr = addr_array;
+ *pmask = mask_array;
+ } else {
+ free(addr_array);
+ free(mask_array);
+ }
+ return 0;
}
+#endif
-struct netdev_dev *
-netdev_get_dev(const struct netdev *netdev)
+void
+netdev_wait_reconf_required(struct netdev *netdev)
{
- return netdev->netdev_dev;
+ seq_wait(netdev->reconfigure_seq, netdev->last_reconfigure_seq);
}
-\f
-/* Restore the network device flags on 'netdev' to those that were active
- * before we changed them. Returns 0 if successful, otherwise a positive
- * errno value.
- *
- * To avoid reentry, the caller must ensure that fatal signals are blocked. */
-static int
-restore_flags(struct netdev *netdev)
+
+bool
+netdev_is_reconf_required(struct netdev *netdev)
{
- if (netdev->changed_flags) {
- enum netdev_flags restore = netdev->save_flags & netdev->changed_flags;
- enum netdev_flags old_flags;
- return netdev_get_dev(netdev)->netdev_class->update_flags(netdev,
- netdev->changed_flags & ~restore,
- restore, &old_flags);
- }
- return 0;
+ return seq_read(netdev->reconfigure_seq) != netdev->last_reconfigure_seq;
}
-/* Close all netdevs on shutdown so they can do any needed cleanup such as
- * destroying devices, restoring flags, etc. */
-static void
-close_all_netdevs(void *aux OVS_UNUSED)
+/* Give a chance to 'netdev' to reconfigure some of its parameters.
+ *
+ * If a module uses netdev_send() and netdev_rxq_recv(), it must call this
+ * function when netdev_is_reconf_required() returns true.
+ *
+ * Return 0 if successful, otherwise a positive errno value. If the
+ * reconfiguration fails the netdev will not be able to send or receive
+ * packets.
+ *
+ * When this function is called, no call to netdev_rxq_recv() or netdev_send()
+ * must be issued. */
+int
+netdev_reconfigure(struct netdev *netdev)
{
- struct netdev *netdev, *next;
- LIST_FOR_EACH_SAFE(netdev, next, node, &netdev_list) {
- netdev_close(netdev);
- }
+ const struct netdev_class *class = netdev->netdev_class;
+
+ netdev->last_reconfigure_seq = seq_read(netdev->reconfigure_seq);
+
+ return (class->reconfigure
+ ? class->reconfigure(netdev)
+ : EOPNOTSUPP);
}