/*
- * Copyright (c) 2009, 2010, 2011, 2012, 2013 Nicira, Inc.
+ * Copyright (c) 2009, 2010, 2011, 2012, 2013, 2014 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 <linux/pkt_sched.h>
#include <linux/rtnetlink.h>
#include <linux/sockios.h>
-#include <linux/version.h>
#include <sys/types.h>
#include <sys/ioctl.h>
#include <sys/socket.h>
#include <unistd.h>
#include "coverage.h"
-#include "dpif-linux.h"
+#include "dpif-netlink.h"
+#include "dpif-netdev.h"
#include "dynamic-string.h"
#include "fatal-signal.h"
#include "hash.h"
#include "ofpbuf.h"
#include "openflow/openflow.h"
#include "ovs-atomic.h"
+#include "packet-dpif.h"
#include "packets.h"
#include "poll-loop.h"
#include "rtnetlink-link.h"
#define TC_RTAB_SIZE 1024
#endif
+/* Linux 2.6.21 introduced struct tpacket_auxdata.
+ * Linux 2.6.27 added the tp_vlan_tci member.
+ * Linux 3.0 defined TP_STATUS_VLAN_VALID.
+ * Linux 3.13 repurposed a padding member for tp_vlan_tpid and defined
+ * TP_STATUS_VLAN_TPID_VALID.
+ *
+ * With all this churn it's easiest to unconditionally define a replacement
+ * structure that has everything we want.
+ */
+#ifndef PACKET_AUXDATA
+#define PACKET_AUXDATA 8
+#endif
+#ifndef TP_STATUS_VLAN_VALID
+#define TP_STATUS_VLAN_VALID (1 << 4)
+#endif
+#ifndef TP_STATUS_VLAN_TPID_VALID
+#define TP_STATUS_VLAN_TPID_VALID (1 << 6)
+#endif
+#undef tpacket_auxdata
+#define tpacket_auxdata rpl_tpacket_auxdata
+struct tpacket_auxdata {
+ uint32_t tp_status;
+ uint32_t tp_len;
+ uint32_t tp_snaplen;
+ uint16_t tp_mac;
+ uint16_t tp_net;
+ uint16_t tp_vlan_tci;
+ uint16_t tp_vlan_tpid;
+};
+
+/* Linux 2.6.35 introduced IFLA_STATS64 and rtnl_link_stats64.
+ *
+ * Tests for rtnl_link_stats64 don't seem to consistently work, e.g. on
+ * 2.6.32-431.29.2.el6.x86_64 (see report at
+ * http://openvswitch.org/pipermail/dev/2014-October/047978.html). Maybe
+ * if_link.h is not self-contained on those kernels. It is easiest to
+ * unconditionally define a replacement. */
+#ifndef IFLA_STATS64
+#define IFLA_STATS64 23
+#endif
+#define rtnl_link_stats64 rpl_rtnl_link_stats64
+struct rtnl_link_stats64 {
+ uint64_t rx_packets;
+ uint64_t tx_packets;
+ uint64_t rx_bytes;
+ uint64_t tx_bytes;
+ uint64_t rx_errors;
+ uint64_t tx_errors;
+ uint64_t rx_dropped;
+ uint64_t tx_dropped;
+ uint64_t multicast;
+ uint64_t collisions;
+
+ uint64_t rx_length_errors;
+ uint64_t rx_over_errors;
+ uint64_t rx_crc_errors;
+ uint64_t rx_frame_errors;
+ uint64_t rx_fifo_errors;
+ uint64_t rx_missed_errors;
+
+ uint64_t tx_aborted_errors;
+ uint64_t tx_carrier_errors;
+ uint64_t tx_fifo_errors;
+ uint64_t tx_heartbeat_errors;
+ uint64_t tx_window_errors;
+
+ uint64_t rx_compressed;
+ uint64_t tx_compressed;
+};
+
enum {
VALID_IFINDEX = 1 << 0,
VALID_ETHERADDR = 1 << 1,
struct ovs_mutex mutex;
unsigned int cache_valid;
- unsigned int change_seq;
bool miimon; /* Link status of last poll. */
long long int miimon_interval; /* Miimon Poll rate. Disabled if <= 0. */
int tap_fd;
};
-struct netdev_rx_linux {
- struct netdev_rx up;
+struct netdev_rxq_linux {
+ struct netdev_rxq up;
bool is_tap;
int fd;
};
/* Polling miimon status for all ports causes performance degradation when
* handling a large number of ports. If there are no devices using miimon, then
- * we skip netdev_linux_miimon_run() and netdev_linux_miimon_wait(). */
-static atomic_int miimon_cnt = ATOMIC_VAR_INIT(0);
+ * we skip netdev_linux_miimon_run() and netdev_linux_miimon_wait().
+ *
+ * Readers do not depend on this variable synchronizing with the related
+ * changes in the device miimon status, so we can use atomic_count. */
+static atomic_count miimon_cnt = ATOMIC_COUNT_INIT(0);
static void netdev_linux_run(void);
struct in_addr addr);
static int get_etheraddr(const char *netdev_name, uint8_t ea[ETH_ADDR_LEN]);
static int set_etheraddr(const char *netdev_name, const uint8_t[ETH_ADDR_LEN]);
-static int get_stats_via_netlink(int ifindex, struct netdev_stats *stats);
-static int get_stats_via_proc(const char *netdev_name, struct netdev_stats *stats);
+static int get_stats_via_netlink(const struct netdev *, struct netdev_stats *);
static int af_packet_sock(void);
static bool netdev_linux_miimon_enabled(void);
static void netdev_linux_miimon_run(void);
static void netdev_linux_miimon_wait(void);
+static int netdev_linux_get_mtu__(struct netdev_linux *netdev, int *mtup);
static bool
is_netdev_linux_class(const struct netdev_class *netdev_class)
return CONTAINER_OF(netdev, struct netdev_linux, up);
}
-static struct netdev_rx_linux *
-netdev_rx_linux_cast(const struct netdev_rx *rx)
+static struct netdev_rxq_linux *
+netdev_rxq_linux_cast(const struct netdev_rxq *rx)
{
ovs_assert(is_netdev_linux_class(netdev_get_class(rx->netdev)));
- return CONTAINER_OF(rx, struct netdev_rx_linux, up);
+ return CONTAINER_OF(rx, struct netdev_rxq_linux, up);
}
\f
static void netdev_linux_update(struct netdev_linux *netdev,
static bool
netdev_linux_miimon_enabled(void)
{
- int miimon;
-
- atomic_read(&miimon_cnt, &miimon);
- return miimon > 0;
+ return atomic_count_get(&miimon_cnt) > 0;
}
static void
unsigned int ifi_flags, unsigned int mask)
OVS_REQUIRES(dev->mutex)
{
- dev->change_seq++;
- if (!dev->change_seq) {
- dev->change_seq++;
- }
+ netdev_change_seq_changed(&dev->up);
if ((dev->ifi_flags ^ ifi_flags) & IFF_RUNNING) {
dev->carrier_resets++;
netdev_linux_common_construct(struct netdev_linux *netdev)
{
ovs_mutex_init(&netdev->mutex);
- netdev->change_seq = 1;
}
/* Creates system and internal devices. */
}
if (netdev->miimon_interval > 0) {
- int junk;
- atomic_sub(&miimon_cnt, 1, &junk);
+ atomic_count_dec(&miimon_cnt);
}
ovs_mutex_destroy(&netdev->mutex);
free(netdev);
}
-static struct netdev_rx *
-netdev_linux_rx_alloc(void)
+static struct netdev_rxq *
+netdev_linux_rxq_alloc(void)
{
- struct netdev_rx_linux *rx = xzalloc(sizeof *rx);
+ struct netdev_rxq_linux *rx = xzalloc(sizeof *rx);
return &rx->up;
}
static int
-netdev_linux_rx_construct(struct netdev_rx *rx_)
+netdev_linux_rxq_construct(struct netdev_rxq *rxq_)
{
- struct netdev_rx_linux *rx = netdev_rx_linux_cast(rx_);
+ struct netdev_rxq_linux *rx = netdev_rxq_linux_cast(rxq_);
struct netdev *netdev_ = rx->up.netdev;
struct netdev_linux *netdev = netdev_linux_cast(netdev_);
int error;
rx->fd = netdev->tap_fd;
} else {
struct sockaddr_ll sll;
- int ifindex;
+ int ifindex, val;
/* Result of tcpdump -dd inbound */
static const struct sock_filter filt[] = {
{ 0x28, 0, 0, 0xfffff004 }, /* ldh [0] */
goto error;
}
+ val = 1;
+ if (setsockopt(rx->fd, SOL_PACKET, PACKET_AUXDATA, &val, sizeof val)) {
+ error = errno;
+ VLOG_ERR("%s: failed to mark socket for auxdata (%s)",
+ netdev_get_name(netdev_), ovs_strerror(error));
+ goto error;
+ }
+
/* Set non-blocking mode. */
error = set_nonblocking(rx->fd);
if (error) {
memset(&sll, 0, sizeof sll);
sll.sll_family = AF_PACKET;
sll.sll_ifindex = ifindex;
- sll.sll_protocol = (OVS_FORCE unsigned short int) htons(ETH_P_ALL);
+ sll.sll_protocol = htons(ETH_P_ALL);
if (bind(rx->fd, (struct sockaddr *) &sll, sizeof sll) < 0) {
error = errno;
VLOG_ERR("%s: failed to bind raw socket (%s)",
}
static void
-netdev_linux_rx_destruct(struct netdev_rx *rx_)
+netdev_linux_rxq_destruct(struct netdev_rxq *rxq_)
{
- struct netdev_rx_linux *rx = netdev_rx_linux_cast(rx_);
+ struct netdev_rxq_linux *rx = netdev_rxq_linux_cast(rxq_);
if (!rx->is_tap) {
close(rx->fd);
}
static void
-netdev_linux_rx_dealloc(struct netdev_rx *rx_)
+netdev_linux_rxq_dealloc(struct netdev_rxq *rxq_)
{
- struct netdev_rx_linux *rx = netdev_rx_linux_cast(rx_);
+ struct netdev_rxq_linux *rx = netdev_rxq_linux_cast(rxq_);
free(rx);
}
+static ovs_be16
+auxdata_to_vlan_tpid(const struct tpacket_auxdata *aux)
+{
+ if (aux->tp_status & TP_STATUS_VLAN_TPID_VALID) {
+ return htons(aux->tp_vlan_tpid);
+ } else {
+ return htons(ETH_TYPE_VLAN);
+ }
+}
+
+static bool
+auxdata_has_vlan_tci(const struct tpacket_auxdata *aux)
+{
+ return aux->tp_vlan_tci || aux->tp_status & TP_STATUS_VLAN_VALID;
+}
+
static int
-netdev_linux_rx_recv(struct netdev_rx *rx_, void *data, size_t size)
+netdev_linux_rxq_recv_sock(int fd, struct ofpbuf *buffer)
{
- struct netdev_rx_linux *rx = netdev_rx_linux_cast(rx_);
+ size_t size;
ssize_t retval;
+ struct iovec iov;
+ struct cmsghdr *cmsg;
+ union {
+ struct cmsghdr cmsg;
+ char buffer[CMSG_SPACE(sizeof(struct tpacket_auxdata))];
+ } cmsg_buffer;
+ struct msghdr msgh;
+
+ /* Reserve headroom for a single VLAN tag */
+ ofpbuf_reserve(buffer, VLAN_HEADER_LEN);
+ size = ofpbuf_tailroom(buffer);
+
+ iov.iov_base = ofpbuf_data(buffer);
+ iov.iov_len = size;
+ msgh.msg_name = NULL;
+ msgh.msg_namelen = 0;
+ msgh.msg_iov = &iov;
+ msgh.msg_iovlen = 1;
+ msgh.msg_control = &cmsg_buffer;
+ msgh.msg_controllen = sizeof cmsg_buffer;
+ msgh.msg_flags = 0;
do {
- retval = (rx->is_tap
- ? read(rx->fd, data, size)
- : recv(rx->fd, data, size, MSG_TRUNC));
+ retval = recvmsg(fd, &msgh, MSG_TRUNC);
} while (retval < 0 && errno == EINTR);
- if (retval >= 0) {
- return retval > size ? -EMSGSIZE : retval;
- } else {
- if (errno != EAGAIN) {
+ if (retval < 0) {
+ return errno;
+ } else if (retval > size) {
+ return EMSGSIZE;
+ }
+
+ ofpbuf_set_size(buffer, ofpbuf_size(buffer) + retval);
+
+ for (cmsg = CMSG_FIRSTHDR(&msgh); cmsg; cmsg = CMSG_NXTHDR(&msgh, cmsg)) {
+ const struct tpacket_auxdata *aux;
+
+ if (cmsg->cmsg_level != SOL_PACKET
+ || cmsg->cmsg_type != PACKET_AUXDATA
+ || cmsg->cmsg_len < CMSG_LEN(sizeof(struct tpacket_auxdata))) {
+ continue;
+ }
+
+ aux = ALIGNED_CAST(struct tpacket_auxdata *, CMSG_DATA(cmsg));
+ if (auxdata_has_vlan_tci(aux)) {
+ if (retval < ETH_HEADER_LEN) {
+ return EINVAL;
+ }
+
+ eth_push_vlan(buffer, auxdata_to_vlan_tpid(aux),
+ htons(aux->tp_vlan_tci));
+ break;
+ }
+ }
+
+ return 0;
+}
+
+static int
+netdev_linux_rxq_recv_tap(int fd, struct ofpbuf *buffer)
+{
+ ssize_t retval;
+ size_t size = ofpbuf_tailroom(buffer);
+
+ do {
+ retval = read(fd, ofpbuf_data(buffer), size);
+ } while (retval < 0 && errno == EINTR);
+
+ if (retval < 0) {
+ return errno;
+ } else if (retval > size) {
+ return EMSGSIZE;
+ }
+
+ ofpbuf_set_size(buffer, ofpbuf_size(buffer) + retval);
+ return 0;
+}
+
+static int
+netdev_linux_rxq_recv(struct netdev_rxq *rxq_, struct dpif_packet **packets,
+ int *c)
+{
+ struct netdev_rxq_linux *rx = netdev_rxq_linux_cast(rxq_);
+ struct netdev *netdev = rx->up.netdev;
+ struct dpif_packet *packet;
+ struct ofpbuf *buffer;
+ ssize_t retval;
+ int mtu;
+
+ if (netdev_linux_get_mtu__(netdev_linux_cast(netdev), &mtu)) {
+ mtu = ETH_PAYLOAD_MAX;
+ }
+
+ packet = dpif_packet_new_with_headroom(VLAN_ETH_HEADER_LEN + mtu,
+ DP_NETDEV_HEADROOM);
+ buffer = &packet->ofpbuf;
+
+ retval = (rx->is_tap
+ ? netdev_linux_rxq_recv_tap(rx->fd, buffer)
+ : netdev_linux_rxq_recv_sock(rx->fd, buffer));
+
+ if (retval) {
+ if (retval != EAGAIN && retval != EMSGSIZE) {
VLOG_WARN_RL(&rl, "error receiving Ethernet packet on %s: %s",
- ovs_strerror(errno), netdev_rx_get_name(rx_));
+ ovs_strerror(errno), netdev_rxq_get_name(rxq_));
}
- return -errno;
+ dpif_packet_delete(packet);
+ } else {
+ dp_packet_pad(buffer);
+ dpif_packet_set_dp_hash(packet, 0);
+ packets[0] = packet;
+ *c = 1;
}
+
+ return retval;
}
static void
-netdev_linux_rx_wait(struct netdev_rx *rx_)
+netdev_linux_rxq_wait(struct netdev_rxq *rxq_)
{
- struct netdev_rx_linux *rx = netdev_rx_linux_cast(rx_);
+ struct netdev_rxq_linux *rx = netdev_rxq_linux_cast(rxq_);
poll_fd_wait(rx->fd, POLLIN);
}
static int
-netdev_linux_rx_drain(struct netdev_rx *rx_)
+netdev_linux_rxq_drain(struct netdev_rxq *rxq_)
{
- struct netdev_rx_linux *rx = netdev_rx_linux_cast(rx_);
+ struct netdev_rxq_linux *rx = netdev_rxq_linux_cast(rxq_);
if (rx->is_tap) {
struct ifreq ifr;
- int error = af_inet_ifreq_ioctl(netdev_rx_get_name(rx_), &ifr,
+ int error = af_inet_ifreq_ioctl(netdev_rxq_get_name(rxq_), &ifr,
SIOCGIFTXQLEN, "SIOCGIFTXQLEN");
if (error) {
return error;
* The kernel maintains a packet transmission queue, so the caller is not
* expected to do additional queuing of packets. */
static int
-netdev_linux_send(struct netdev *netdev_, const void *data, size_t size)
+netdev_linux_send(struct netdev *netdev_, int qid OVS_UNUSED,
+ struct dpif_packet **pkts, int cnt, bool may_steal)
{
- for (;;) {
+ int i;
+ int error = 0;
+
+ /* 'i' is incremented only if there's no error */
+ for (i = 0; i < cnt;) {
+ const void *data = ofpbuf_data(&pkts[i]->ofpbuf);
+ size_t size = ofpbuf_size(&pkts[i]->ofpbuf);
ssize_t retval;
if (!is_tap_netdev(netdev_)) {
/* The Linux AF_PACKET implementation never blocks waiting for room
* for packets, instead returning ENOBUFS. Translate this into
* EAGAIN for the caller. */
- if (errno == ENOBUFS) {
- return EAGAIN;
- } else if (errno == EINTR) {
+ error = errno == ENOBUFS ? EAGAIN : errno;
+ if (error == EINTR) {
+ /* continue without incrementing 'i', i.e. retry this packet */
continue;
- } else if (errno != EAGAIN) {
- VLOG_WARN_RL(&rl, "error sending Ethernet packet on %s: %s",
- netdev_get_name(netdev_), ovs_strerror(errno));
}
- return errno;
+ break;
} else if (retval != size) {
- VLOG_WARN_RL(&rl, "sent partial Ethernet packet (%"PRIuSIZE"d bytes of "
- "%"PRIuSIZE") on %s", retval, size, netdev_get_name(netdev_));
- return EMSGSIZE;
- } else {
- return 0;
+ VLOG_WARN_RL(&rl, "sent partial Ethernet packet (%"PRIuSIZE" bytes"
+ " of %"PRIuSIZE") on %s", retval, size,
+ netdev_get_name(netdev_));
+ error = EMSGSIZE;
+ break;
+ }
+
+ /* Process the next packet in the batch */
+ i++;
+ }
+
+ if (may_steal) {
+ for (i = 0; i < cnt; i++) {
+ dpif_packet_delete(pkts[i]);
}
}
+
+ if (error && error != EAGAIN) {
+ VLOG_WARN_RL(&rl, "error sending Ethernet packet on %s: %s",
+ netdev_get_name(netdev_), ovs_strerror(error));
+ }
+
+ return error;
+
}
/* Registers with the poll loop to wake up from the next call to poll_block()
* expected to do additional queuing of packets. Thus, this function is
* unlikely to ever be used. It is included for completeness. */
static void
-netdev_linux_send_wait(struct netdev *netdev)
+netdev_linux_send_wait(struct netdev *netdev, int qid OVS_UNUSED)
{
if (is_tap_netdev(netdev)) {
/* TAP device always accepts packets.*/
ovs_mutex_lock(&netdev->mutex);
interval = interval > 0 ? MAX(interval, 100) : 0;
if (netdev->miimon_interval != interval) {
- int junk;
-
if (interval && !netdev->miimon_interval) {
- atomic_add(&miimon_cnt, 1, &junk);
+ atomic_count_inc(&miimon_cnt);
} else if (!interval && netdev->miimon_interval) {
- atomic_sub(&miimon_cnt, 1, &junk);
+ atomic_count_dec(&miimon_cnt);
}
netdev->miimon_interval = interval;
shash_destroy(&device_shash);
}
-/* Check whether we can we use RTM_GETLINK to get network device statistics.
- * In pre-2.6.19 kernels, this was only available if wireless extensions were
- * enabled. */
-static bool
-check_for_working_netlink_stats(void)
-{
- /* Decide on the netdev_get_stats() implementation to use. Netlink is
- * preferable, so if that works, we'll use it. */
- int ifindex = do_get_ifindex("lo");
- if (ifindex < 0) {
- VLOG_WARN("failed to get ifindex for lo, "
- "obtaining netdev stats from proc");
- return false;
- } else {
- struct netdev_stats stats;
- int error = get_stats_via_netlink(ifindex, &stats);
- if (!error) {
- VLOG_DBG("obtaining netdev stats via rtnetlink");
- return true;
- } else {
- VLOG_INFO("RTM_GETLINK failed (%s), obtaining netdev stats "
- "via proc (you are probably running a pre-2.6.19 "
- "kernel)", ovs_strerror(error));
- return false;
- }
- }
-}
-
static void
swap_uint64(uint64_t *a, uint64_t *b)
{
netdev_stats_from_ovs_vport_stats(struct netdev_stats *dst,
const struct ovs_vport_stats *src)
{
- dst->rx_packets = get_unaligned_u64(&src->rx_packets);
- dst->tx_packets = get_unaligned_u64(&src->tx_packets);
- dst->rx_bytes = get_unaligned_u64(&src->rx_bytes);
- dst->tx_bytes = get_unaligned_u64(&src->tx_bytes);
- dst->rx_errors = get_unaligned_u64(&src->rx_errors);
- dst->tx_errors = get_unaligned_u64(&src->tx_errors);
- dst->rx_dropped = get_unaligned_u64(&src->rx_dropped);
- dst->tx_dropped = get_unaligned_u64(&src->tx_dropped);
+ dst->rx_packets = get_32aligned_u64(&src->rx_packets);
+ dst->tx_packets = get_32aligned_u64(&src->tx_packets);
+ dst->rx_bytes = get_32aligned_u64(&src->rx_bytes);
+ dst->tx_bytes = get_32aligned_u64(&src->tx_bytes);
+ dst->rx_errors = get_32aligned_u64(&src->rx_errors);
+ dst->tx_errors = get_32aligned_u64(&src->tx_errors);
+ dst->rx_dropped = get_32aligned_u64(&src->rx_dropped);
+ dst->tx_dropped = get_32aligned_u64(&src->tx_dropped);
dst->multicast = 0;
dst->collisions = 0;
dst->rx_length_errors = 0;
static int
get_stats_via_vport__(const struct netdev *netdev, struct netdev_stats *stats)
{
- struct dpif_linux_vport reply;
+ struct dpif_netlink_vport reply;
struct ofpbuf *buf;
int error;
- error = dpif_linux_vport_get(netdev_get_name(netdev), &reply, &buf);
+ error = dpif_netlink_vport_get(netdev_get_name(netdev), &reply, &buf);
if (error) {
return error;
} else if (!reply.stats) {
}
}
-static int
-netdev_linux_sys_get_stats(const struct netdev *netdev_,
- struct netdev_stats *stats)
-{
- static struct ovsthread_once once = OVSTHREAD_ONCE_INITIALIZER;
- static int use_netlink_stats;
- int error;
-
- if (ovsthread_once_start(&once)) {
- use_netlink_stats = check_for_working_netlink_stats();
- ovsthread_once_done(&once);
- }
-
- if (use_netlink_stats) {
- int ifindex;
-
- error = get_ifindex(netdev_, &ifindex);
- if (!error) {
- error = get_stats_via_netlink(ifindex, stats);
- }
- } else {
- error = get_stats_via_proc(netdev_get_name(netdev_), stats);
- }
-
- if (error) {
- VLOG_WARN_RL(&rl, "%s: linux-sys get stats failed %d",
- netdev_get_name(netdev_), error);
- }
- return error;
-
-}
-
/* Retrieves current device stats for 'netdev-linux'. */
static int
netdev_linux_get_stats(const struct netdev *netdev_,
ovs_mutex_lock(&netdev->mutex);
get_stats_via_vport(netdev_, stats);
- error = netdev_linux_sys_get_stats(netdev_, &dev_stats);
+ error = get_stats_via_netlink(netdev_, &dev_stats);
if (error) {
if (!netdev->vport_stats_error) {
error = 0;
}
} else if (netdev->vport_stats_error) {
- /* stats not available from OVS then use ioctl stats. */
+ /* stats not available from OVS then use netdev stats. */
*stats = dev_stats;
} else {
+ /* Use kernel netdev's packet and byte counts since vport's counters
+ * do not reflect packet counts on the wire when GSO, TSO or GRO are
+ * enabled. */
+ stats->rx_packets = dev_stats.rx_packets;
+ stats->rx_bytes = dev_stats.rx_bytes;
+ stats->tx_packets = dev_stats.tx_packets;
+ stats->tx_bytes = dev_stats.tx_bytes;
+
stats->rx_errors += dev_stats.rx_errors;
stats->tx_errors += dev_stats.tx_errors;
stats->rx_dropped += dev_stats.rx_dropped;
ovs_mutex_lock(&netdev->mutex);
get_stats_via_vport(netdev_, stats);
- error = netdev_linux_sys_get_stats(netdev_, &dev_stats);
+ error = get_stats_via_netlink(netdev_, &dev_stats);
if (error) {
if (!netdev->vport_stats_error) {
error = 0;
stats->tx_heartbeat_errors = 0;
stats->tx_window_errors = 0;
} else {
+ /* Use kernel netdev's packet and byte counts since vport counters
+ * do not reflect packet counts on the wire when GSO, TSO or GRO
+ * are enabled. */
+ stats->rx_packets = dev_stats.tx_packets;
+ stats->rx_bytes = dev_stats.tx_bytes;
+ stats->tx_packets = dev_stats.rx_packets;
+ stats->tx_bytes = dev_stats.rx_bytes;
+
stats->rx_dropped += dev_stats.tx_dropped;
stats->tx_dropped += dev_stats.rx_dropped;
return error;
}
-static int
-netdev_internal_set_stats(struct netdev *netdev,
- const struct netdev_stats *stats)
-{
- struct ovs_vport_stats vport_stats;
- struct dpif_linux_vport vport;
- int err;
-
- vport_stats.rx_packets = stats->rx_packets;
- vport_stats.tx_packets = stats->tx_packets;
- vport_stats.rx_bytes = stats->rx_bytes;
- vport_stats.tx_bytes = stats->tx_bytes;
- vport_stats.rx_errors = stats->rx_errors;
- vport_stats.tx_errors = stats->tx_errors;
- vport_stats.rx_dropped = stats->rx_dropped;
- vport_stats.tx_dropped = stats->tx_dropped;
-
- dpif_linux_vport_init(&vport);
- vport.cmd = OVS_VPORT_CMD_SET;
- vport.name = netdev_get_name(netdev);
- vport.stats = &vport_stats;
-
- err = dpif_linux_vport_transact(&vport, NULL, NULL);
-
- /* If the vport layer doesn't know about the device, that doesn't mean it
- * doesn't exist (after all were able to open it when netdev_open() was
- * called), it just means that it isn't attached and we'll be getting
- * stats a different way. */
- if (err == ENODEV) {
- err = EOPNOTSUPP;
- }
-
- return err;
-}
-
static void
netdev_linux_read_features(struct netdev_linux *netdev)
{
return error;
}
+struct queue_dump_state {
+ struct nl_dump dump;
+ struct ofpbuf buf;
+};
+
static bool
-start_queue_dump(const struct netdev *netdev, struct nl_dump *dump)
+start_queue_dump(const struct netdev *netdev, struct queue_dump_state *state)
{
struct ofpbuf request;
struct tcmsg *tcmsg;
return false;
}
tcmsg->tcm_parent = 0;
- nl_dump_start(dump, NETLINK_ROUTE, &request);
+ nl_dump_start(&state->dump, NETLINK_ROUTE, &request);
ofpbuf_uninit(&request);
+
+ ofpbuf_init(&state->buf, NL_DUMP_BUFSIZE);
return true;
}
+static int
+finish_queue_dump(struct queue_dump_state *state)
+{
+ ofpbuf_uninit(&state->buf);
+ return nl_dump_done(&state->dump);
+}
+
struct netdev_linux_queue_state {
unsigned int *queues;
size_t cur_queue;
ovs_mutex_lock(&netdev->mutex);
error = tc_query_qdisc(netdev_);
if (!error) {
- struct nl_dump dump;
+ struct queue_dump_state state;
if (!netdev->tc->ops->class_dump_stats) {
error = EOPNOTSUPP;
- } else if (!start_queue_dump(netdev_, &dump)) {
+ } else if (!start_queue_dump(netdev_, &state)) {
error = ENODEV;
} else {
struct ofpbuf msg;
int retval;
- while (nl_dump_next(&dump, &msg)) {
+ while (nl_dump_next(&state.dump, &msg, &state.buf)) {
retval = netdev->tc->ops->class_dump_stats(netdev_, &msg,
cb, aux);
if (retval) {
}
}
- retval = nl_dump_done(&dump);
+ retval = finish_queue_dump(&state);
if (retval) {
error = retval;
}
return error;
}
-static unsigned int
-netdev_linux_change_seq(const struct netdev *netdev_)
-{
- struct netdev_linux *netdev = netdev_linux_cast(netdev_);
- unsigned int change_seq;
-
- ovs_mutex_lock(&netdev->mutex);
- change_seq = netdev->change_seq;
- ovs_mutex_unlock(&netdev->mutex);
-
- return change_seq;
-}
-
-#define NETDEV_LINUX_CLASS(NAME, CONSTRUCT, GET_STATS, SET_STATS, \
+#define NETDEV_LINUX_CLASS(NAME, CONSTRUCT, GET_STATS, \
GET_FEATURES, GET_STATUS) \
{ \
NAME, \
NULL, /* get_config */ \
NULL, /* set_config */ \
NULL, /* get_tunnel_config */ \
+ NULL, /* get_numa_id */ \
+ NULL, /* set_multiq */ \
\
netdev_linux_send, \
netdev_linux_send_wait, \
netdev_linux_get_carrier_resets, \
netdev_linux_set_miimon_interval, \
GET_STATS, \
- SET_STATS, \
\
GET_FEATURES, \
netdev_linux_set_advertisements, \
\
netdev_linux_update_flags, \
\
- netdev_linux_change_seq, \
- \
- netdev_linux_rx_alloc, \
- netdev_linux_rx_construct, \
- netdev_linux_rx_destruct, \
- netdev_linux_rx_dealloc, \
- netdev_linux_rx_recv, \
- netdev_linux_rx_wait, \
- netdev_linux_rx_drain, \
+ netdev_linux_rxq_alloc, \
+ netdev_linux_rxq_construct, \
+ netdev_linux_rxq_destruct, \
+ netdev_linux_rxq_dealloc, \
+ netdev_linux_rxq_recv, \
+ netdev_linux_rxq_wait, \
+ netdev_linux_rxq_drain, \
}
const struct netdev_class netdev_linux_class =
"system",
netdev_linux_construct,
netdev_linux_get_stats,
- NULL, /* set_stats */
netdev_linux_get_features,
netdev_linux_get_status);
"tap",
netdev_linux_construct_tap,
netdev_tap_get_stats,
- NULL, /* set_stats */
netdev_linux_get_features,
netdev_linux_get_status);
"internal",
netdev_linux_construct,
netdev_internal_get_stats,
- netdev_internal_set_stats,
NULL, /* get_features */
netdev_internal_get_status);
\f
htb_tc_load(struct netdev *netdev, struct ofpbuf *nlmsg OVS_UNUSED)
{
struct ofpbuf msg;
- struct nl_dump dump;
+ struct queue_dump_state state;
struct htb_class hc;
/* Get qdisc options. */
htb_install__(netdev, hc.max_rate);
/* Get queues. */
- if (!start_queue_dump(netdev, &dump)) {
+ if (!start_queue_dump(netdev, &state)) {
return ENODEV;
}
- while (nl_dump_next(&dump, &msg)) {
+ while (nl_dump_next(&state.dump, &msg, &state.buf)) {
unsigned int queue_id;
if (!htb_parse_tcmsg__(&msg, &queue_id, &hc, NULL)) {
htb_update_queue__(netdev, queue_id, &hc);
}
}
- nl_dump_done(&dump);
+ finish_queue_dump(&state);
return 0;
}
hfsc_tc_load(struct netdev *netdev, struct ofpbuf *nlmsg OVS_UNUSED)
{
struct ofpbuf msg;
- struct nl_dump dump;
+ struct queue_dump_state state;
struct hfsc_class hc;
hc.max_rate = 0;
hfsc_query_class__(netdev, tc_make_handle(1, 0xfffe), 0, &hc, NULL);
hfsc_install__(netdev, hc.max_rate);
- if (!start_queue_dump(netdev, &dump)) {
+ if (!start_queue_dump(netdev, &state)) {
return ENODEV;
}
- while (nl_dump_next(&dump, &msg)) {
+ while (nl_dump_next(&state.dump, &msg, &state.buf)) {
unsigned int queue_id;
if (!hfsc_parse_tcmsg__(&msg, &queue_id, &hc, NULL)) {
}
}
- nl_dump_done(&dump);
+ finish_queue_dump(&state);
return 0;
}
memset(&tc_police, 0, sizeof tc_police);
tc_police.action = TC_POLICE_SHOT;
tc_police.mtu = mtu;
- tc_fill_rate(&tc_police.rate, (kbits_rate * 1000)/8, mtu);
+ tc_fill_rate(&tc_police.rate, ((uint64_t) kbits_rate * 1000)/8, mtu);
tc_police.burst = tc_bytes_to_ticks(tc_police.rate.rate,
kbits_burst * 1024);
dst->tx_window_errors = src->tx_window_errors;
}
-static int
-get_stats_via_netlink(int ifindex, struct netdev_stats *stats)
+/* Copies 'src' into 'dst', performing format conversion in the process. */
+static void
+netdev_stats_from_rtnl_link_stats64(struct netdev_stats *dst,
+ const struct rtnl_link_stats64 *src)
{
- /* Policy for RTNLGRP_LINK messages.
- *
- * There are *many* more fields in these messages, but currently we only
- * care about these fields. */
- static const struct nl_policy rtnlgrp_link_policy[] = {
- [IFLA_IFNAME] = { .type = NL_A_STRING, .optional = false },
- [IFLA_STATS] = { .type = NL_A_UNSPEC, .optional = true,
- .min_len = sizeof(struct rtnl_link_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;
+}
+static int
+get_stats_via_netlink(const struct netdev *netdev_, struct netdev_stats *stats)
+{
struct ofpbuf request;
struct ofpbuf *reply;
- struct ifinfomsg *ifi;
- struct nlattr *attrs[ARRAY_SIZE(rtnlgrp_link_policy)];
int error;
ofpbuf_init(&request, 0);
- nl_msg_put_nlmsghdr(&request, sizeof *ifi, RTM_GETLINK, NLM_F_REQUEST);
- ifi = ofpbuf_put_zeros(&request, sizeof *ifi);
- ifi->ifi_family = PF_UNSPEC;
- ifi->ifi_index = ifindex;
+ nl_msg_put_nlmsghdr(&request,
+ sizeof(struct ifinfomsg) + NL_ATTR_SIZE(IFNAMSIZ),
+ RTM_GETLINK, NLM_F_REQUEST);
+ ofpbuf_put_zeros(&request, sizeof(struct ifinfomsg));
+ nl_msg_put_string(&request, IFLA_IFNAME, netdev_get_name(netdev_));
error = nl_transact(NETLINK_ROUTE, &request, &reply);
ofpbuf_uninit(&request);
if (error) {
return error;
}
- if (!nl_policy_parse(reply, NLMSG_HDRLEN + sizeof(struct ifinfomsg),
- rtnlgrp_link_policy,
- attrs, ARRAY_SIZE(rtnlgrp_link_policy))) {
- ofpbuf_delete(reply);
- return EPROTO;
- }
-
- if (!attrs[IFLA_STATS]) {
- VLOG_WARN_RL(&rl, "RTM_GETLINK reply lacks stats");
- ofpbuf_delete(reply);
- return EPROTO;
+ if (ofpbuf_try_pull(reply, NLMSG_HDRLEN + sizeof(struct ifinfomsg))) {
+ const struct nlattr *a = nl_attr_find(reply, 0, IFLA_STATS64);
+ if (a && nl_attr_get_size(a) >= sizeof(struct rtnl_link_stats64)) {
+ netdev_stats_from_rtnl_link_stats64(stats, nl_attr_get(a));
+ error = 0;
+ } else {
+ const struct nlattr *a = nl_attr_find(reply, 0, IFLA_STATS);
+ if (a && nl_attr_get_size(a) >= sizeof(struct rtnl_link_stats)) {
+ netdev_stats_from_rtnl_link_stats(stats, nl_attr_get(a));
+ error = 0;
+ } else {
+ VLOG_WARN_RL(&rl, "RTM_GETLINK reply lacks stats");
+ error = EPROTO;
+ }
+ }
+ } else {
+ VLOG_WARN_RL(&rl, "short RTM_GETLINK reply");
+ error = EPROTO;
}
- netdev_stats_from_rtnl_link_stats(stats, nl_attr_get(attrs[IFLA_STATS]));
ofpbuf_delete(reply);
-
- return 0;
-}
-
-static int
-get_stats_via_proc(const char *netdev_name, struct netdev_stats *stats)
-{
- static const char fn[] = "/proc/net/dev";
- char line[1024];
- FILE *stream;
- int ln;
-
- stream = fopen(fn, "r");
- if (!stream) {
- VLOG_WARN_RL(&rl, "%s: open failed: %s", fn, ovs_strerror(errno));
- return errno;
- }
-
- ln = 0;
- while (fgets(line, sizeof line, stream)) {
- if (++ln >= 3) {
- char devname[16];
-#define X64 "%"SCNu64
- if (!ovs_scan(line,
- " %15[^:]:"
- X64 X64 X64 X64 X64 X64 X64 "%*u"
- X64 X64 X64 X64 X64 X64 X64 "%*u",
- devname,
- &stats->rx_bytes,
- &stats->rx_packets,
- &stats->rx_errors,
- &stats->rx_dropped,
- &stats->rx_fifo_errors,
- &stats->rx_frame_errors,
- &stats->multicast,
- &stats->tx_bytes,
- &stats->tx_packets,
- &stats->tx_errors,
- &stats->tx_dropped,
- &stats->tx_fifo_errors,
- &stats->collisions,
- &stats->tx_carrier_errors)) {
- VLOG_WARN_RL(&rl, "%s:%d: parse error", fn, ln);
- } else if (!strcmp(devname, netdev_name)) {
- stats->rx_length_errors = UINT64_MAX;
- stats->rx_over_errors = UINT64_MAX;
- stats->rx_crc_errors = UINT64_MAX;
- stats->rx_missed_errors = UINT64_MAX;
- stats->tx_aborted_errors = UINT64_MAX;
- stats->tx_heartbeat_errors = UINT64_MAX;
- stats->tx_window_errors = UINT64_MAX;
- fclose(stream);
- return 0;
- }
- }
- }
- VLOG_WARN_RL(&rl, "%s: no stats for %s", fn, netdev_name);
- fclose(stream);
- return ENODEV;
+ return error;
}
static int
projects / cascardo / ovs.git / blobdiff