/*
- * Copyright (c) 2009, 2010 Nicira Networks.
+ * Copyright (c) 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.
*/
#include <config.h>
+
+#include "netdev-linux.h"
+
#include <assert.h>
#include <errno.h>
#include <fcntl.h>
#include <unistd.h>
#include "coverage.h"
+#include "dpif-linux.h"
#include "dynamic-string.h"
#include "fatal-signal.h"
#include "hash.h"
#include "rtnetlink-link.h"
#include "socket-util.h"
#include "shash.h"
-#include "svec.h"
+#include "sset.h"
#include "vlog.h"
VLOG_DEFINE_THIS_MODULE(netdev_linux);
int fd;
};
-/* An AF_INET socket (used for ioctl operations). */
-static int af_inet_sock = -1;
+/* Sockets used for ioctl operations. */
+static int af_inet_sock = -1; /* AF_INET, SOCK_DGRAM. */
+static int af_packet_sock = -1; /* AF_PACKET, SOCK_RAW. */
/* A Netlink routing socket that is not subscribed to any multicast groups. */
static struct nl_sock *rtnl_sock;
status = af_inet_sock >= 0 ? 0 : errno;
if (status) {
VLOG_ERR("failed to create inet socket: %s", strerror(status));
+ } else {
+ /* Create AF_PACKET socket. */
+ af_packet_sock = socket(AF_PACKET, SOCK_RAW, 0);
+ status = af_packet_sock >= 0 ? 0 : errno;
+ if (status) {
+ VLOG_ERR("failed to create packet socket: %s",
+ strerror(status));
+ }
+ set_nonblocking(af_packet_sock);
}
/* Create rtnetlink socket. */
if (!status) {
- status = nl_sock_create(NETLINK_ROUTE, 0, 0, 0, &rtnl_sock);
+ status = nl_sock_create(NETLINK_ROUTE, &rtnl_sock);
if (status) {
VLOG_ERR_RL(&rl, "failed to create rtnetlink socket: %s",
strerror(status));
cache_notifier_refcount++;
netdev_dev = xzalloc(sizeof *netdev_dev);
- netdev_dev_init(&netdev_dev->netdev_dev, name, class);
+ netdev_dev_init(&netdev_dev->netdev_dev, name, args, class);
*netdev_devp = &netdev_dev->netdev_dev;
return 0;
/* Create tap device. */
ifr.ifr_flags = IFF_TAP | IFF_NO_PI;
- strncpy(ifr.ifr_name, name, sizeof ifr.ifr_name);
+ ovs_strzcpy(ifr.ifr_name, name, sizeof ifr.ifr_name);
if (ioctl(state->fd, TUNSETIFF, &ifr) == -1) {
VLOG_WARN("%s: creating tap device failed: %s", name,
strerror(errno));
goto error;
}
- netdev_dev_init(&netdev_dev->netdev_dev, name, &netdev_tap_class);
+ netdev_dev_init(&netdev_dev->netdev_dev, name, args, &netdev_tap_class);
*netdev_devp = &netdev_dev->netdev_dev;
return 0;
free(netdev);
}
-/* Initializes 'svec' with a list of the names of all known network devices. */
+/* Initializes 'sset' with a list of the names of all known network devices. */
static int
-netdev_linux_enumerate(struct svec *svec)
+netdev_linux_enumerate(struct sset *sset)
{
struct if_nameindex *names;
size_t i;
for (i = 0; names[i].if_name != NULL; i++) {
- svec_add(svec, names[i].if_name);
+ sset_add(sset, names[i].if_name);
}
if_freenameindex(names);
return 0;
netdev_linux_send(struct netdev *netdev_, const void *data, size_t size)
{
struct netdev_linux *netdev = netdev_linux_cast(netdev_);
+ for (;;) {
+ ssize_t retval;
- /* XXX should support sending even if 'ethertype' was NETDEV_ETH_TYPE_NONE.
- */
- if (netdev->fd < 0) {
- return EPIPE;
- }
+ if (netdev->fd < 0) {
+ /* Use our AF_PACKET socket to send to this device. */
+ struct sockaddr_ll sll;
+ struct msghdr msg;
+ struct iovec iov;
+ int ifindex;
+ int error;
+
+ error = get_ifindex(netdev_, &ifindex);
+ if (error) {
+ return error;
+ }
+
+ /* We don't bother setting most fields in sockaddr_ll because the
+ * kernel ignores them for SOCK_RAW. */
+ memset(&sll, 0, sizeof sll);
+ sll.sll_family = AF_PACKET;
+ sll.sll_ifindex = ifindex;
+
+ iov.iov_base = (void *) data;
+ iov.iov_len = size;
+
+ msg.msg_name = &sll;
+ msg.msg_namelen = sizeof sll;
+ msg.msg_iov = &iov;
+ msg.msg_iovlen = 1;
+ msg.msg_control = NULL;
+ msg.msg_controllen = 0;
+ msg.msg_flags = 0;
+
+ retval = sendmsg(af_packet_sock, &msg, 0);
+ } else {
+ /* Use the netdev's own fd to send to this device. This is
+ * essential for tap devices, because packets sent to a tap device
+ * with an AF_PACKET socket will loop back to be *received* again
+ * on the tap device. */
+ retval = write(netdev->fd, data, size);
+ }
- for (;;) {
- ssize_t retval = write(netdev->fd, data, size);
if (retval < 0) {
/* The Linux AF_PACKET implementation never blocks waiting for room
* for packets, instead returning ENOBUFS. Translate this into
}
static int
-netdev_linux_get_miimon(const struct netdev *netdev_, bool *miimon)
+netdev_linux_do_miimon(const struct netdev *netdev, int cmd,
+ const char *cmd_name, struct mii_ioctl_data *data)
{
- int error;
struct ifreq ifr;
- const char *name = netdev_get_name(netdev_);
+ int error;
- *miimon = false;
memset(&ifr, 0, sizeof ifr);
+ memcpy(&ifr.ifr_data, data, sizeof *data);
+ error = netdev_linux_do_ioctl(netdev_get_name(netdev),
+ &ifr, cmd, cmd_name);
+ memcpy(data, &ifr.ifr_data, sizeof *data);
- error = netdev_linux_do_ioctl(name, &ifr, SIOCGMIIPHY, "SIOCGMIIPHY");
- if (!error) {
- struct mii_ioctl_data *data = (struct mii_ioctl_data *)&ifr.ifr_data;
+ return error;
+}
- /* data->phy_id is filled out by previous SIOCGMIIPHY ioctl call. */
- data->reg_num = MII_BMSR;
- error = netdev_linux_do_ioctl(name, &ifr, SIOCGMIIREG, "SIOCGMIIREG");
+static int
+netdev_linux_get_miimon(const struct netdev *netdev, bool *miimon)
+{
+ const char *name = netdev_get_name(netdev);
+ struct mii_ioctl_data data;
+ int error;
+
+ *miimon = false;
+
+ memset(&data, 0, sizeof data);
+ error = netdev_linux_do_miimon(netdev, SIOCGMIIPHY, "SIOCGMIIPHY", &data);
+ if (!error) {
+ /* data.phy_id is filled out by previous SIOCGMIIPHY miimon call. */
+ data.reg_num = MII_BMSR;
+ error = netdev_linux_do_miimon(netdev, SIOCGMIIREG, "SIOCGMIIREG",
+ &data);
if (!error) {
- *miimon = !!(data->val_out & BMSR_LSTATUS);
+ *miimon = !!(data.val_out & BMSR_LSTATUS);
} else {
VLOG_WARN_RL(&rl, "%s: failed to query MII", name);
}
} else {
struct ethtool_cmd ecmd;
- struct ethtool_value *eval = (struct ethtool_value *) &ecmd;
VLOG_DBG_RL(&rl, "%s: failed to query MII, falling back to ethtool",
name);
error = netdev_linux_do_ethtool(name, &ecmd, ETHTOOL_GLINK,
"ETHTOOL_GLINK");
if (!error) {
- *miimon = !!eval->data;
+ struct ethtool_value eval;
+
+ memcpy(&eval, &ecmd, sizeof eval);
+ *miimon = !!eval.data;
} else {
VLOG_WARN_RL(&rl, "%s: ethtool link status failed", name);
}
const char *type = netdev_dev_get_type(&netdev_dev->netdev_dev);
netdev_dev->is_tap = !strcmp(type, "tap");
- netdev_dev->is_internal = false;
- if (!netdev_dev->is_tap) {
- struct ethtool_drvinfo drvinfo;
- int error;
-
- memset(&drvinfo, 0, sizeof drvinfo);
- error = netdev_linux_do_ethtool(name,
- (struct ethtool_cmd *)&drvinfo,
- ETHTOOL_GDRVINFO,
- "ETHTOOL_GDRVINFO");
-
- if (!error && !strcmp(drvinfo.driver, "openvswitch")) {
- netdev_dev->is_internal = true;
- }
- }
-
+ netdev_dev->is_internal = (!netdev_dev->is_tap
+ && dpif_linux_is_internal_device(name));
netdev_dev->cache_valid |= VALID_IS_PSEUDO;
}
}
static void
swap_uint64(uint64_t *a, uint64_t *b)
{
- *a ^= *b;
- *b ^= *a;
- *a ^= *b;
+ uint64_t tmp = *a;
+ *a = *b;
+ *b = tmp;
}
/* Retrieves current device stats for 'netdev'. */
* bitmap of "enum ofp_port_features" bits, in host byte order. Returns 0 if
* successful, otherwise a positive errno value. */
static int
-netdev_linux_get_features(struct netdev *netdev,
+netdev_linux_get_features(const struct netdev *netdev,
uint32_t *current, uint32_t *advertised,
uint32_t *supported, uint32_t *peer)
{
static int
netdev_linux_get_qos_types(const struct netdev *netdev OVS_UNUSED,
- struct svec *types)
+ struct sset *types)
{
const struct tc_ops **opsp;
for (opsp = tcs; *opsp != NULL; opsp++) {
const struct tc_ops *ops = *opsp;
if (ops->tc_install && ops->ovs_name[0] != '\0') {
- svec_add(types, ops->ovs_name);
+ sset_add(types, ops->ovs_name);
}
}
return 0;
int ioctl_nr, const char *ioctl_name, struct in_addr addr)
{
struct ifreq ifr;
- strncpy(ifr.ifr_name, netdev_get_name(netdev), sizeof ifr.ifr_name);
+ ovs_strzcpy(ifr.ifr_name, netdev_get_name(netdev), sizeof ifr.ifr_name);
make_in4_sockaddr(&ifr.ifr_addr, addr);
return netdev_linux_do_ioctl(netdev_get_name(netdev), &ifr, ioctl_nr,
while (fgets(line, sizeof line, stream)) {
if (++ln >= 2) {
char iface[17];
- uint32_t dest, gateway, mask;
+ ovs_be32 dest, gateway, mask;
int refcnt, metric, mtu;
unsigned int flags, use, window, irtt;
return ENXIO;
}
+static int
+netdev_linux_get_status(const struct netdev *netdev, struct shash *sh)
+{
+ struct ethtool_drvinfo drvinfo;
+ int error;
+
+ memset(&drvinfo, 0, sizeof drvinfo);
+ error = netdev_linux_do_ethtool(netdev_get_name(netdev),
+ (struct ethtool_cmd *)&drvinfo,
+ ETHTOOL_GDRVINFO,
+ "ETHTOOL_GDRVINFO");
+ if (!error) {
+ shash_add(sh, "driver_name", xstrdup(drvinfo.driver));
+ shash_add(sh, "driver_version", xstrdup(drvinfo.version));
+ shash_add(sh, "firmware_version", xstrdup(drvinfo.fw_version));
+ }
+
+ return error;
+}
+
/* Looks up the ARP table entry for 'ip' on 'netdev'. If one exists and can be
* successfully retrieved, it stores the corresponding MAC address in 'mac' and
* returns 0. Otherwise, it returns a positive errno value; in particular,
* ENXIO indicates that there is not ARP table entry for 'ip' on 'netdev'. */
static int
netdev_linux_arp_lookup(const struct netdev *netdev,
- uint32_t ip, uint8_t mac[ETH_ADDR_LEN])
+ ovs_be32 ip, uint8_t mac[ETH_ADDR_LEN])
{
struct arpreq r;
struct sockaddr_in sin;
int retval;
memset(&r, 0, sizeof r);
+ memset(&sin, 0, sizeof sin);
sin.sin_family = AF_INET;
sin.sin_addr.s_addr = ip;
sin.sin_port = 0;
memcpy(&r.arp_pa, &sin, sizeof sin);
r.arp_ha.sa_family = ARPHRD_ETHER;
r.arp_flags = 0;
- strncpy(r.arp_dev, netdev_get_name(netdev), sizeof r.arp_dev);
+ ovs_strzcpy(r.arp_dev, netdev_get_name(netdev), sizeof r.arp_dev);
COVERAGE_INC(netdev_arp_lookup);
retval = ioctl(af_inet_sock, SIOCGARP, &r) < 0 ? errno : 0;
if (!retval) {
\
CREATE, \
netdev_linux_destroy, \
- NULL, /* reconfigure */ \
+ NULL, /* set_config */ \
\
netdev_linux_open, \
netdev_linux_close, \
netdev_linux_get_in6, \
netdev_linux_add_router, \
netdev_linux_get_next_hop, \
- NULL, /* get_status */ \
+ netdev_linux_get_status, \
netdev_linux_arp_lookup, \
\
netdev_linux_update_flags, \
return CONTAINER_OF(netdev_dev->tc, struct htb, tc);
}
-static struct htb *
+static void
htb_install__(struct netdev *netdev, uint64_t max_rate)
{
struct netdev_dev_linux *netdev_dev =
htb->max_rate = max_rate;
netdev_dev->tc = &htb->tc;
-
- return htb;
}
/* Create an HTB qdisc.
int mtu;
netdev_get_mtu(netdev, &mtu);
+ if (mtu == INT_MAX) {
+ VLOG_WARN_RL(&rl, "cannot set up HTB on device %s that lacks MTU",
+ netdev_get_name(netdev));
+ return EINVAL;
+ }
memset(&opt, 0, sizeof opt);
tc_fill_rate(&opt.rate, class->min_rate, mtu);
const char *priority_s = shash_find_data(details, "priority");
int mtu;
- /* min-rate. Don't allow a min-rate below 1500 bytes/s. */
- if (!min_rate_s) {
- /* min-rate is required. */
+ netdev_get_mtu(netdev, &mtu);
+ if (mtu == INT_MAX) {
+ VLOG_WARN_RL(&rl, "cannot parse HTB class on device %s that lacks MTU",
+ netdev_get_name(netdev));
return EINVAL;
}
- hc->min_rate = strtoull(min_rate_s, NULL, 10) / 8;
- hc->min_rate = MAX(hc->min_rate, 1500);
+
+ /* HTB requires at least an mtu sized min-rate to send any traffic even
+ * on uncongested links. */
+ hc->min_rate = min_rate_s ? strtoull(min_rate_s, NULL, 10) / 8 : 0;
+ hc->min_rate = MAX(hc->min_rate, mtu);
hc->min_rate = MIN(hc->min_rate, htb->max_rate);
/* max-rate */
* doesn't include the Ethernet header, we need to add at least 14 (18?) to
* the MTU. We actually add 64, instead of 14, as a guard against
* additional headers get tacked on somewhere that we're not aware of. */
- netdev_get_mtu(netdev, &mtu);
hc->burst = burst_s ? strtoull(burst_s, NULL, 10) / 8 : 0;
hc->burst = MAX(hc->burst, mtu + 64);
struct ofpbuf msg;
struct nl_dump dump;
struct htb_class hc;
- struct htb *htb;
/* Get qdisc options. */
hc.max_rate = 0;
htb_query_class__(netdev, tc_make_handle(1, 0xfffe), 0, &hc, NULL);
- htb = htb_install__(netdev, hc.max_rate);
+ htb_install__(netdev, hc.max_rate);
/* Get queues. */
if (!start_queue_dump(netdev, &dump)) {
return CONTAINER_OF(queue, struct hfsc_class, tc_queue);
}
-static struct hfsc *
+static void
hfsc_install__(struct netdev *netdev, uint32_t max_rate)
{
struct netdev_dev_linux * netdev_dev;
tc_init(&hfsc->tc, &tc_ops_hfsc);
hfsc->max_rate = max_rate;
netdev_dev->tc = &hfsc->tc;
-
- return hfsc;
}
static void
min_rate_s = shash_find_data(details, "min-rate");
max_rate_s = shash_find_data(details, "max-rate");
- if (!min_rate_s) {
- return EINVAL;
- }
-
- min_rate = strtoull(min_rate_s, NULL, 10) / 8;
- min_rate = MAX(min_rate, 1500);
+ min_rate = min_rate_s ? strtoull(min_rate_s, NULL, 10) / 8 : 0;
+ min_rate = MAX(min_rate, 1);
min_rate = MIN(min_rate, hfsc->max_rate);
max_rate = (max_rate_s
hfsc_tc_load(struct netdev *netdev, struct ofpbuf *nlmsg OVS_UNUSED)
{
struct ofpbuf msg;
- struct hfsc *hfsc;
struct nl_dump dump;
struct hfsc_class hc;
hc.max_rate = 0;
hfsc_query_class__(netdev, tc_make_handle(1, 0xfffe), 0, &hc, NULL);
- hfsc = hfsc_install__(netdev, hc.max_rate);
+ hfsc_install__(netdev, hc.max_rate);
if (!start_queue_dump(netdev, &dump)) {
return ENODEV;
unsigned int min_burst = tc_buffer_per_jiffy(Bps) + mtu;
return tc_bytes_to_ticks(Bps, MAX(burst_bytes, min_burst));
}
-
+\f
+/* Public utility functions. */
+
+#define COPY_NETDEV_STATS \
+ dst->rx_packets = src->rx_packets; \
+ dst->tx_packets = src->tx_packets; \
+ dst->rx_bytes = src->rx_bytes; \
+ dst->tx_bytes = src->tx_bytes; \
+ dst->rx_errors = src->rx_errors; \
+ dst->tx_errors = src->tx_errors; \
+ dst->rx_dropped = src->rx_dropped; \
+ dst->tx_dropped = src->tx_dropped; \
+ dst->multicast = src->multicast; \
+ dst->collisions = src->collisions; \
+ dst->rx_length_errors = src->rx_length_errors; \
+ dst->rx_over_errors = src->rx_over_errors; \
+ dst->rx_crc_errors = src->rx_crc_errors; \
+ dst->rx_frame_errors = src->rx_frame_errors; \
+ dst->rx_fifo_errors = src->rx_fifo_errors; \
+ dst->rx_missed_errors = src->rx_missed_errors; \
+ dst->tx_aborted_errors = src->tx_aborted_errors; \
+ dst->tx_carrier_errors = src->tx_carrier_errors; \
+ dst->tx_fifo_errors = src->tx_fifo_errors; \
+ dst->tx_heartbeat_errors = src->tx_heartbeat_errors; \
+ dst->tx_window_errors = src->tx_window_errors
+
+/* Copies 'src' into 'dst', performing format conversion in the process. */
+void
+netdev_stats_from_rtnl_link_stats(struct netdev_stats *dst,
+ const struct rtnl_link_stats *src)
+{
+ COPY_NETDEV_STATS;
+}
+
+/* Copies 'src' into 'dst', performing format conversion in the process. */
+void
+netdev_stats_from_rtnl_link_stats64(struct netdev_stats *dst,
+ const struct rtnl_link_stats64 *src)
+{
+ COPY_NETDEV_STATS;
+}
+
+/* Copies 'src' into 'dst', performing format conversion in the process. */
+void
+netdev_stats_to_rtnl_link_stats64(struct rtnl_link_stats64 *dst,
+ const struct netdev_stats *src)
+{
+ COPY_NETDEV_STATS;
+}
\f
/* Utility functions. */
struct ofpbuf request;
struct ofpbuf *reply;
struct ifinfomsg *ifi;
- const struct rtnl_link_stats *rtnl_stats;
struct nlattr *attrs[ARRAY_SIZE(rtnlgrp_link_policy)];
int error;
return EPROTO;
}
- rtnl_stats = nl_attr_get(attrs[IFLA_STATS]);
- stats->rx_packets = rtnl_stats->rx_packets;
- stats->tx_packets = rtnl_stats->tx_packets;
- stats->rx_bytes = rtnl_stats->rx_bytes;
- stats->tx_bytes = rtnl_stats->tx_bytes;
- stats->rx_errors = rtnl_stats->rx_errors;
- stats->tx_errors = rtnl_stats->tx_errors;
- stats->rx_dropped = rtnl_stats->rx_dropped;
- stats->tx_dropped = rtnl_stats->tx_dropped;
- stats->multicast = rtnl_stats->multicast;
- stats->collisions = rtnl_stats->collisions;
- stats->rx_length_errors = rtnl_stats->rx_length_errors;
- stats->rx_over_errors = rtnl_stats->rx_over_errors;
- stats->rx_crc_errors = rtnl_stats->rx_crc_errors;
- stats->rx_frame_errors = rtnl_stats->rx_frame_errors;
- stats->rx_fifo_errors = rtnl_stats->rx_fifo_errors;
- stats->rx_missed_errors = rtnl_stats->rx_missed_errors;
- stats->tx_aborted_errors = rtnl_stats->tx_aborted_errors;
- stats->tx_carrier_errors = rtnl_stats->tx_carrier_errors;
- stats->tx_fifo_errors = rtnl_stats->tx_fifo_errors;
- stats->tx_heartbeat_errors = rtnl_stats->tx_heartbeat_errors;
- stats->tx_window_errors = rtnl_stats->tx_window_errors;
+ netdev_stats_from_rtnl_link_stats(stats, nl_attr_get(attrs[IFLA_STATS]));
ofpbuf_delete(reply);
{
struct ifreq ifr;
- strncpy(ifr.ifr_name, netdev_name, sizeof ifr.ifr_name);
+ ovs_strzcpy(ifr.ifr_name, netdev_name, sizeof ifr.ifr_name);
COVERAGE_INC(netdev_get_ifindex);
if (ioctl(af_inet_sock, SIOCGIFINDEX, &ifr) < 0) {
VLOG_WARN_RL(&rl, "ioctl(SIOCGIFINDEX) on %s device failed: %s",
int hwaddr_family;
memset(&ifr, 0, sizeof ifr);
- strncpy(ifr.ifr_name, netdev_name, sizeof ifr.ifr_name);
+ ovs_strzcpy(ifr.ifr_name, netdev_name, sizeof ifr.ifr_name);
COVERAGE_INC(netdev_get_hwaddr);
if (ioctl(af_inet_sock, SIOCGIFHWADDR, &ifr) < 0) {
VLOG_ERR("ioctl(SIOCGIFHWADDR) on %s device failed: %s",
struct ifreq ifr;
memset(&ifr, 0, sizeof ifr);
- strncpy(ifr.ifr_name, netdev_name, sizeof ifr.ifr_name);
+ ovs_strzcpy(ifr.ifr_name, netdev_name, sizeof ifr.ifr_name);
ifr.ifr_hwaddr.sa_family = hwaddr_family;
memcpy(ifr.ifr_hwaddr.sa_data, mac, ETH_ADDR_LEN);
COVERAGE_INC(netdev_set_hwaddr);
struct ifreq ifr;
memset(&ifr, 0, sizeof ifr);
- strncpy(ifr.ifr_name, name, sizeof ifr.ifr_name);
+ ovs_strzcpy(ifr.ifr_name, name, sizeof ifr.ifr_name);
ifr.ifr_data = (caddr_t) ecmd;
ecmd->cmd = cmd;
netdev_linux_do_ioctl(const char *name, struct ifreq *ifr, int cmd,
const char *cmd_name)
{
- strncpy(ifr->ifr_name, name, sizeof ifr->ifr_name);
+ ovs_strzcpy(ifr->ifr_name, name, sizeof ifr->ifr_name);
if (ioctl(af_inet_sock, cmd, ifr) == -1) {
VLOG_DBG_RL(&rl, "%s: ioctl(%s) failed: %s", name, cmd_name,
strerror(errno));
git://git.onelab.eu / sliver-openvswitch.git / blobdiff