#include <sys/ioctl.h>
#include "byte-order.h"
+#include "daemon.h"
+#include "dirs.h"
#include "dpif-linux.h"
#include "hash.h"
#include "hmap.h"
#include "list.h"
+#include "netdev-linux.h"
#include "netdev-provider.h"
#include "netlink.h"
#include "netlink-socket.h"
#include "packets.h"
#include "route-table.h"
#include "rtnetlink.h"
-#include "rtnetlink-link.h"
#include "shash.h"
#include "socket-util.h"
#include "vlog.h"
VLOG_DEFINE_THIS_MODULE(netdev_vport);
-static struct hmap name_map;
-static struct rtnetlink_notifier netdev_vport_link_notifier;
-
-struct name_node {
- struct hmap_node node; /* Node in name_map. */
- uint32_t ifi_index; /* Kernel interface index. */
-
- char ifname[IFNAMSIZ]; /* Interface name. */
-};
-
-struct netdev_vport_notifier {
- struct netdev_notifier notifier;
- struct list list_node;
- struct shash_node *shash_node;
-};
-
struct netdev_dev_vport {
struct netdev_dev netdev_dev;
struct ofpbuf *options;
+ int dp_ifindex; /* -1 if unknown. */
+ uint32_t port_no; /* UINT32_MAX if unknown. */
+ unsigned int change_seq;
};
struct netdev_vport {
struct shash *args);
};
-static struct shash netdev_vport_notifiers =
- SHASH_INITIALIZER(&netdev_vport_notifiers);
-
static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(5, 20);
static int netdev_vport_create(const struct netdev_class *, const char *,
- const struct shash *, struct netdev_dev **);
+ struct netdev_dev **);
static void netdev_vport_poll_notify(const struct netdev *);
static int tnl_port_config_from_nlattr(const struct nlattr *options,
size_t options_len,
struct nlattr *a[ODP_TUNNEL_ATTR_MAX + 1]);
-static void netdev_vport_tnl_iface_init(void);
-static void netdev_vport_link_change(const struct rtnetlink_link_change *,
- void *);
static const char *netdev_vport_get_tnl_iface(const struct netdev *netdev);
static bool
return "unknown";
}
-static int
-netdev_vport_init(void)
-{
- netdev_vport_tnl_iface_init();
- return 0;
-}
-
static int
netdev_vport_create(const struct netdev_class *netdev_class, const char *name,
- const struct shash *args,
struct netdev_dev **netdev_devp)
{
- const struct vport_class *vport_class = vport_class_cast(netdev_class);
- struct ofpbuf *options = NULL;
- struct shash fetched_args;
- int error;
-
- shash_init(&fetched_args);
-
- if (!shash_is_empty(args)) {
- /* Parse the provided configuration. */
- options = ofpbuf_new(64);
- error = vport_class->parse_config(name, netdev_class->type,
- args, options);
- } else {
- /* Fetch an existing configuration from the kernel.
- *
- * This case could be ambiguous with initializing a new vport with an
- * empty configuration, but none of the existing vport classes accept
- * an empty configuration. */
- struct dpif_linux_vport reply;
- struct ofpbuf *buf;
-
- error = dpif_linux_vport_get(name, &reply, &buf);
- if (!error) {
- /* XXX verify correct type */
- error = vport_class->unparse_config(name, netdev_class->type,
- reply.options,
- reply.options_len,
- &fetched_args);
- if (error) {
- VLOG_ERR_RL(&rl, "%s: failed to parse kernel config (%s)",
- name, strerror(error));
- } else {
- options = ofpbuf_clone_data(reply.options, reply.options_len);
- }
- ofpbuf_delete(buf);
- } else {
- VLOG_ERR_RL(&rl, "%s: vport query failed (%s)",
- name, strerror(error));
- }
- }
-
- if (!error) {
- struct netdev_dev_vport *dev;
+ struct netdev_dev_vport *dev;
- dev = xmalloc(sizeof *dev);
- netdev_dev_init(&dev->netdev_dev, name,
- shash_is_empty(&fetched_args) ? args : &fetched_args,
- netdev_class);
- dev->options = options;
+ dev = xmalloc(sizeof *dev);
+ netdev_dev_init(&dev->netdev_dev, name, netdev_class);
+ dev->options = NULL;
+ dev->dp_ifindex = -1;
+ dev->port_no = UINT32_MAX;
+ dev->change_seq = 1;
- *netdev_devp = &dev->netdev_dev;
- route_table_register();
- } else {
- ofpbuf_delete(options);
- }
-
- shash_destroy(&fetched_args);
+ *netdev_devp = &dev->netdev_dev;
+ route_table_register();
- return error;
+ return 0;
}
static void
}
static int
-netdev_vport_open(struct netdev_dev *netdev_dev_, int ethertype OVS_UNUSED,
- struct netdev **netdevp)
+netdev_vport_open(struct netdev_dev *netdev_dev_, struct netdev **netdevp)
{
struct netdev_vport *netdev;
free(netdev);
}
+static int
+netdev_vport_get_config(struct netdev_dev *dev_, struct shash *args)
+{
+ const struct netdev_class *netdev_class = netdev_dev_get_class(dev_);
+ const struct vport_class *vport_class = vport_class_cast(netdev_class);
+ struct netdev_dev_vport *dev = netdev_dev_vport_cast(dev_);
+ const char *name = netdev_dev_get_name(dev_);
+ int error;
+
+ if (!dev->options) {
+ struct dpif_linux_vport reply;
+ struct ofpbuf *buf;
+
+ error = dpif_linux_vport_get(name, &reply, &buf);
+ if (error) {
+ VLOG_ERR_RL(&rl, "%s: vport query failed (%s)",
+ name, strerror(error));
+ return error;
+ }
+
+ dev->options = ofpbuf_clone_data(reply.options, reply.options_len);
+ dev->dp_ifindex = reply.dp_ifindex;
+ dev->port_no = reply.port_no;
+ ofpbuf_delete(buf);
+ }
+
+ error = vport_class->unparse_config(name, netdev_class->type,
+ dev->options->data,
+ dev->options->size,
+ args);
+ if (error) {
+ VLOG_ERR_RL(&rl, "%s: failed to parse kernel config (%s)",
+ name, strerror(error));
+ }
+ return error;
+}
+
static int
netdev_vport_set_config(struct netdev_dev *dev_, const struct shash *args)
{
error = vport_class->parse_config(name, netdev_dev_get_type(dev_),
args, options);
if (!error
- && (options->size != dev->options->size
+ && (!dev->options
+ || options->size != dev->options->size
|| memcmp(options->data, dev->options->data, options->size))) {
struct dpif_linux_vport vport;
return error;
}
+static int
+netdev_vport_send(struct netdev *netdev, const void *data, size_t size)
+{
+ struct netdev_dev *dev_ = netdev_get_dev(netdev);
+ struct netdev_dev_vport *dev = netdev_dev_vport_cast(dev_);
+
+ if (dev->dp_ifindex == -1) {
+ const char *name = netdev_get_name(netdev);
+ struct dpif_linux_vport reply;
+ struct ofpbuf *buf;
+ int error;
+
+ error = dpif_linux_vport_get(name, &reply, &buf);
+ if (error) {
+ VLOG_ERR_RL(&rl, "%s: failed to query vport for send (%s)",
+ name, strerror(error));
+ return error;
+ }
+ dev->dp_ifindex = reply.dp_ifindex;
+ dev->port_no = reply.port_no;
+ ofpbuf_delete(buf);
+ }
+
+ return dpif_linux_vport_send(dev->dp_ifindex, dev->port_no, data, size);
+}
+
static int
netdev_vport_set_etheraddr(struct netdev *netdev,
const uint8_t mac[ETH_ADDR_LEN])
return EOPNOTSUPP;
}
- stats->rx_packets = reply.stats->rx_packets;
- stats->tx_packets = reply.stats->tx_packets;
- stats->rx_bytes = reply.stats->rx_bytes;
- stats->tx_bytes = reply.stats->tx_bytes;
- stats->rx_errors = reply.stats->rx_errors;
- stats->tx_errors = reply.stats->tx_errors;
- stats->rx_dropped = reply.stats->rx_dropped;
- stats->tx_dropped = reply.stats->tx_dropped;
- stats->multicast = reply.stats->multicast;
- stats->collisions = reply.stats->collisions;
- stats->rx_length_errors = reply.stats->rx_length_errors;
- stats->rx_over_errors = reply.stats->rx_over_errors;
- stats->rx_crc_errors = reply.stats->rx_crc_errors;
- stats->rx_frame_errors = reply.stats->rx_frame_errors;
- stats->rx_fifo_errors = reply.stats->rx_fifo_errors;
- stats->rx_missed_errors = reply.stats->rx_missed_errors;
- stats->tx_aborted_errors = reply.stats->tx_aborted_errors;
- stats->tx_carrier_errors = reply.stats->tx_carrier_errors;
- stats->tx_fifo_errors = reply.stats->tx_fifo_errors;
- stats->tx_heartbeat_errors = reply.stats->tx_heartbeat_errors;
- stats->tx_window_errors = reply.stats->tx_window_errors;
+ netdev_stats_from_rtnl_link_stats64(stats, reply.stats);
ofpbuf_delete(buf);
struct dpif_linux_vport vport;
int err;
- rtnl_stats.rx_packets = stats->rx_packets;
- rtnl_stats.tx_packets = stats->tx_packets;
- rtnl_stats.rx_bytes = stats->rx_bytes;
- rtnl_stats.tx_bytes = stats->tx_bytes;
- rtnl_stats.rx_errors = stats->rx_errors;
- rtnl_stats.tx_errors = stats->tx_errors;
- rtnl_stats.rx_dropped = stats->rx_dropped;
- rtnl_stats.tx_dropped = stats->tx_dropped;
- rtnl_stats.multicast = stats->multicast;
- rtnl_stats.collisions = stats->collisions;
- rtnl_stats.rx_length_errors = stats->rx_length_errors;
- rtnl_stats.rx_over_errors = stats->rx_over_errors;
- rtnl_stats.rx_crc_errors = stats->rx_crc_errors;
- rtnl_stats.rx_frame_errors = stats->rx_frame_errors;
- rtnl_stats.rx_fifo_errors = stats->rx_fifo_errors;
- rtnl_stats.rx_missed_errors = stats->rx_missed_errors;
- rtnl_stats.tx_aborted_errors = stats->tx_aborted_errors;
- rtnl_stats.tx_carrier_errors = stats->tx_carrier_errors;
- rtnl_stats.tx_fifo_errors = stats->tx_fifo_errors;
- rtnl_stats.tx_heartbeat_errors = stats->tx_heartbeat_errors;
- rtnl_stats.tx_window_errors = stats->tx_window_errors;
+ netdev_stats_to_rtnl_link_stats64(&rtnl_stats, stats);
dpif_linux_vport_init(&vport);
vport.cmd = ODP_VPORT_CMD_SET;
shash_add(sh, "tunnel_egress_iface", xstrdup(iface));
- if (!netdev_open_default(iface, &egress_netdev)) {
+ if (!netdev_open(iface, "system", &egress_netdev)) {
shash_add(sh, "tunnel_egress_iface_carrier",
xstrdup(netdev_get_carrier(egress_netdev)
? "up" : "down"));
return 0;
}
-static char *
-make_poll_name(const struct netdev *netdev)
+static unsigned int
+netdev_vport_change_seq(const struct netdev *netdev)
{
- return xasprintf("%s:%s", netdev_get_type(netdev), netdev_get_name(netdev));
-}
-
-static int
-netdev_vport_poll_add(struct netdev *netdev,
- void (*cb)(struct netdev_notifier *), void *aux,
- struct netdev_notifier **notifierp)
-{
- char *poll_name = make_poll_name(netdev);
- struct netdev_vport_notifier *notifier;
- struct list *list;
- struct shash_node *shash_node;
-
- shash_node = shash_find_data(&netdev_vport_notifiers, poll_name);
- if (!shash_node) {
- list = xmalloc(sizeof *list);
- list_init(list);
- shash_node = shash_add(&netdev_vport_notifiers, poll_name, list);
- } else {
- list = shash_node->data;
- }
-
- notifier = xmalloc(sizeof *notifier);
- netdev_notifier_init(¬ifier->notifier, netdev, cb, aux);
- list_push_back(list, ¬ifier->list_node);
- notifier->shash_node = shash_node;
-
- *notifierp = ¬ifier->notifier;
- free(poll_name);
-
- return 0;
-}
-
-static void
-netdev_vport_poll_remove(struct netdev_notifier *notifier_)
-{
- struct netdev_vport_notifier *notifier =
- CONTAINER_OF(notifier_, struct netdev_vport_notifier, notifier);
-
- struct list *list;
-
- list = list_remove(¬ifier->list_node);
- if (list_is_empty(list)) {
- shash_delete(&netdev_vport_notifiers, notifier->shash_node);
- free(list);
- }
-
- free(notifier);
+ return netdev_dev_vport_cast(netdev_get_dev(netdev))->change_seq;
}
static void
netdev_vport_run(void)
{
- rtnetlink_link_notifier_run();
route_table_run();
}
static void
netdev_vport_wait(void)
{
- rtnetlink_link_notifier_wait();
route_table_wait();
}
\f
/* get_tnl_iface() implementation. */
-
-static struct name_node *
-name_node_lookup(int ifi_index)
-{
- struct name_node *nn;
-
- HMAP_FOR_EACH_WITH_HASH(nn, node, hash_int(ifi_index, 0), &name_map) {
- if (nn->ifi_index == ifi_index) {
- return nn;
- }
- }
-
- return NULL;
-}
-
-/* Queries the kernel for fresh data to populate the name map with. */
-static int
-netdev_vport_reset_names(void)
-{
- int error;
- struct nl_dump dump;
- struct rtgenmsg *rtmsg;
- struct ofpbuf request, reply;
- static struct nl_sock *rtnl_sock;
- struct name_node *nn, *nn_next;
-
- HMAP_FOR_EACH_SAFE(nn, nn_next, node, &name_map) {
- hmap_remove(&name_map, &nn->node);
- free(nn);
- }
-
- error = nl_sock_create(NETLINK_ROUTE, &rtnl_sock);
- if (error) {
- VLOG_WARN_RL(&rl, "Failed to create NETLINK_ROUTE socket");
- return error;
- }
-
- ofpbuf_init(&request, 0);
-
- nl_msg_put_nlmsghdr(&request, sizeof *rtmsg, RTM_GETLINK, NLM_F_REQUEST);
-
- rtmsg = ofpbuf_put_zeros(&request, sizeof *rtmsg);
- rtmsg->rtgen_family = AF_INET;
-
- nl_dump_start(&dump, rtnl_sock, &request);
-
- while (nl_dump_next(&dump, &reply)) {
- struct rtnetlink_link_change change;
-
- if (rtnetlink_link_parse(&reply, &change)) {
- netdev_vport_link_change(&change, NULL);
- }
- }
- nl_sock_destroy(rtnl_sock);
-
- return nl_dump_done(&dump);
-}
-
-static void
-netdev_vport_link_change(const struct rtnetlink_link_change *change,
- void *aux OVS_UNUSED)
-{
-
- if (!change) {
- netdev_vport_reset_names();
- } else if (change->nlmsg_type == RTM_NEWLINK) {
- struct name_node *nn;
-
- if (name_node_lookup(change->ifi_index)) {
- return;
- }
-
- nn = xzalloc(sizeof *nn);
- nn->ifi_index = change->ifi_index;
-
- strncpy(nn->ifname, change->ifname, IFNAMSIZ);
- nn->ifname[IFNAMSIZ - 1] = '\0';
-
- hmap_insert(&name_map, &nn->node, hash_int(nn->ifi_index, 0));
- } else if (change->nlmsg_type == RTM_DELLINK) {
- struct name_node *nn;
-
- nn = name_node_lookup(change->ifi_index);
-
- if (nn) {
- hmap_remove(&name_map, &nn->node);
- free(nn);
- }
-
- } else {
- VLOG_WARN_RL(&rl, "Received unexpected rtnetlink message type %d",
- change->nlmsg_type);
- }
-}
-
-static void
-netdev_vport_tnl_iface_init(void)
-{
- static bool tnl_iface_is_init = false;
-
- if (!tnl_iface_is_init) {
- hmap_init(&name_map);
-
- rtnetlink_link_notifier_register(&netdev_vport_link_notifier,
- netdev_vport_link_change, NULL);
-
- netdev_vport_reset_names();
- tnl_iface_is_init = true;
- }
-}
-
static const char *
netdev_vport_get_tnl_iface(const struct netdev *netdev)
{
struct nlattr *a[ODP_TUNNEL_ATTR_MAX + 1];
- int ifindex;
- uint32_t route;
+ ovs_be32 route;
struct netdev_dev_vport *ndv;
+ static char name[IFNAMSIZ];
ndv = netdev_dev_vport_cast(netdev_get_dev(netdev));
if (tnl_port_config_from_nlattr(ndv->options->data, ndv->options->size,
}
route = nl_attr_get_be32(a[ODP_TUNNEL_ATTR_DST_IPV4]);
- if (route_table_get_ifindex(route, &ifindex)) {
- struct name_node *nn;
- HMAP_FOR_EACH_WITH_HASH(nn, node, hash_int(ifindex, 0), &name_map) {
- if (nn->ifi_index == ifindex) {
- return nn->ifname;
- }
- }
+ if (route_table_get_name(route, name)) {
+ return name;
}
return NULL;
static void
netdev_vport_poll_notify(const struct netdev *netdev)
{
- char *poll_name = make_poll_name(netdev);
- struct list *list = shash_find_data(&netdev_vport_notifiers,
- poll_name);
+ struct netdev_dev_vport *ndv;
- if (list) {
- struct netdev_vport_notifier *notifier;
+ ndv = netdev_dev_vport_cast(netdev_get_dev(netdev));
- LIST_FOR_EACH (notifier, list_node, list) {
- struct netdev_notifier *n = ¬ifier->notifier;
- n->cb(n);
- }
+ ndv->change_seq++;
+ if (!ndv->change_seq) {
+ ndv->change_seq++;
}
-
- free(poll_name);
}
\f
/* Code specific to individual vport types. */
ovs_be32 daddr = htonl(0);
uint32_t flags;
- flags = TNL_F_PMTUD | TNL_F_HDR_CACHE;
+ flags = TNL_F_DF_DEFAULT | TNL_F_PMTUD | TNL_F_HDR_CACHE;
if (!strcmp(type, "gre")) {
is_gre = true;
} else if (!strcmp(type, "ipsec_gre")) {
if (!strcmp(node->data, "true")) {
flags |= TNL_F_CSUM;
}
+ } else if (!strcmp(node->name, "df_inherit")) {
+ if (!strcmp(node->data, "true")) {
+ flags |= TNL_F_DF_INHERIT;
+ }
+ } else if (!strcmp(node->name, "df_default")) {
+ if (!strcmp(node->data, "false")) {
+ flags &= ~TNL_F_DF_DEFAULT;
+ }
} else if (!strcmp(node->name, "pmtud")) {
if (!strcmp(node->data, "false")) {
flags &= ~TNL_F_PMTUD;
*/
use_ssl_cert = shash_find_data(args, "use_ssl_cert");
if (!use_ssl_cert || strcmp(use_ssl_cert, "true")) {
- VLOG_WARN("%s: 'peer_cert' requires 'certificate' argument",
- name);
+ VLOG_ERR("%s: 'peer_cert' requires 'certificate' argument",
+ name);
return EINVAL;
}
ipsec_mech_set = true;
|| !strcmp(node->name, "private_key")
|| !strcmp(node->name, "use_ssl_cert"))) {
/* Ignore options not used by the netdev. */
- } else if (is_gre && (!strcmp(node->name, "key") &&
- !strcmp(node->name, "in_key") &&
+ } else if (is_gre && (!strcmp(node->name, "key") ||
+ !strcmp(node->name, "in_key") ||
!strcmp(node->name, "out_key"))) {
/* Handled separately below. */
} else {
}
if (is_ipsec) {
+ char *file_name = xasprintf("%s/%s", ovs_rundir(),
+ "ovs-monitor-ipsec.pid");
+ pid_t pid = read_pidfile(file_name);
+ free(file_name);
+ if (pid < 0) {
+ VLOG_ERR("%s: IPsec requires the ovs-monitor-ipsec daemon",
+ name);
+ return EINVAL;
+ }
+
if (shash_find(args, "peer_cert") && shash_find(args, "psk")) {
- VLOG_WARN("%s: cannot define both 'peer_cert' and 'psk'", name);
+ VLOG_ERR("%s: cannot define both 'peer_cert' and 'psk'", name);
return EINVAL;
}
if (!ipsec_mech_set) {
- VLOG_WARN("%s: IPsec requires an 'peer_cert' or psk' argument",
- name);
+ VLOG_ERR("%s: IPsec requires an 'peer_cert' or psk' argument",
+ name);
return EINVAL;
}
}
}
if (!daddr) {
- VLOG_WARN("%s: %s type requires valid 'remote_ip' argument",
- name, type);
+ VLOG_ERR("%s: %s type requires valid 'remote_ip' argument",
+ name, type);
return EINVAL;
}
nl_msg_put_be32(options, ODP_TUNNEL_ATTR_DST_IPV4, daddr);
if (flags & TNL_F_CSUM) {
smap_add(args, "csum", "true");
}
+ if (flags & TNL_F_DF_INHERIT) {
+ smap_add(args, "df_inherit", "true");
+ }
+ if (!(flags & TNL_F_DF_DEFAULT)) {
+ smap_add(args, "df_default", "false");
+ }
if (!(flags & TNL_F_PMTUD)) {
smap_add(args, "pmtud", "false");
}
peer = shash_find_data(args, "peer");
if (!peer) {
- VLOG_WARN("%s: patch type requires valid 'peer' argument", name);
+ VLOG_ERR("%s: patch type requires valid 'peer' argument", name);
return EINVAL;
}
if (shash_count(args) > 1) {
- VLOG_WARN("%s: patch type takes only a 'peer' argument", name);
+ VLOG_ERR("%s: patch type takes only a 'peer' argument", name);
return EINVAL;
}
if (strlen(peer) >= IFNAMSIZ) {
- VLOG_WARN("%s: patch 'peer' arg too long", name);
+ VLOG_ERR("%s: patch 'peer' arg too long", name);
return EINVAL;
}
if (!strcmp(name, peer)) {
- VLOG_WARN("%s: patch peer must not be self", name);
+ VLOG_ERR("%s: patch peer must not be self", name);
return EINVAL;
}
}
\f
#define VPORT_FUNCTIONS(GET_STATUS) \
- netdev_vport_init, \
+ NULL, \
netdev_vport_run, \
netdev_vport_wait, \
\
netdev_vport_create, \
netdev_vport_destroy, \
+ netdev_vport_get_config, \
netdev_vport_set_config, \
\
netdev_vport_open, \
\
NULL, /* enumerate */ \
\
+ NULL, /* listen */ \
NULL, /* recv */ \
NULL, /* recv_wait */ \
NULL, /* drain */ \
\
- NULL, /* send */ \
+ netdev_vport_send, /* send */ \
NULL, /* send_wait */ \
\
netdev_vport_set_etheraddr, \
\
netdev_vport_update_flags, \
\
- netdev_vport_poll_add, \
- netdev_vport_poll_remove,
+ netdev_vport_change_seq
void
netdev_vport_register(void)
git://git.onelab.eu / sliver-openvswitch.git / blobdiff