X-Git-Url: https://pintos-os.org/cgi-bin/gitweb.cgi?a=blobdiff_plain;f=lib%2Fnetdev-linux.c;h=27a123cd4ea9c87642d04f7d7dbb3110898e5205;hb=02e83e83b4c8d88f50317ba56fec53cd6d3fb909;hp=b5d30352379828d32a9f65de4009ac26a9bda503;hpb=1670c579a82921fedd8b2c20818919f6b5a9c330;p=openvswitch diff --git a/lib/netdev-linux.c b/lib/netdev-linux.c index b5d30352..27a123cd 100644 --- a/lib/netdev-linux.c +++ b/lib/netdev-linux.c @@ -24,11 +24,13 @@ #include #include #include +#include #include #include #include #include #include +#include #include #include #include @@ -37,9 +39,7 @@ #include #include #include -#include #include -#include #include #include #include @@ -58,12 +58,12 @@ #include "netdev-provider.h" #include "netdev-vport.h" #include "netlink.h" +#include "netlink-notifier.h" #include "netlink-socket.h" #include "ofpbuf.h" #include "openflow/openflow.h" #include "packets.h" #include "poll-loop.h" -#include "rtnetlink.h" #include "rtnetlink-link.h" #include "socket-util.h" #include "shash.h" @@ -73,7 +73,6 @@ VLOG_DEFINE_THIS_MODULE(netdev_linux); -COVERAGE_DEFINE(netdev_get_vlan_vid); COVERAGE_DEFINE(netdev_set_policing); COVERAGE_DEFINE(netdev_arp_lookup); COVERAGE_DEFINE(netdev_get_ifindex); @@ -90,13 +89,22 @@ COVERAGE_DEFINE(netdev_ethtool); #define ADVERTISED_Asym_Pause (1 << 14) #endif +/* These were introduced in Linux 2.6.24, so they might be missing if we + * have old headers. */ +#ifndef ETHTOOL_GFLAGS +#define ETHTOOL_GFLAGS 0x00000025 /* Get flags bitmap(ethtool_value) */ +#endif +#ifndef ETHTOOL_SFLAGS +#define ETHTOOL_SFLAGS 0x00000026 /* Set flags bitmap(ethtool_value) */ +#endif + /* This was introduced in Linux 2.6.25, so it might be missing if we have old * headers. */ #ifndef TC_RTAB_SIZE #define TC_RTAB_SIZE 1024 #endif -static struct rtnetlink_notifier netdev_linux_cache_notifier; +static struct nln_notifier *netdev_linux_cache_notifier = NULL; static int cache_notifier_refcount; enum { @@ -105,10 +113,8 @@ enum { VALID_IN4 = 1 << 2, VALID_IN6 = 1 << 3, VALID_MTU = 1 << 4, - VALID_CARRIER = 1 << 5, - VALID_IS_PSEUDO = 1 << 6, /* Represents is_internal and is_tap. */ - VALID_POLICING = 1 << 7, - VALID_HAVE_VPORT_STATS = 1 << 8 + VALID_POLICING = 1 << 5, + VALID_HAVE_VPORT_STATS = 1 << 6 }; struct tap_state { @@ -321,6 +327,9 @@ static unsigned int tc_buffer_per_jiffy(unsigned int rate); static struct tcmsg *tc_make_request(const struct netdev *, int type, unsigned int flags, struct ofpbuf *); static int tc_transact(struct ofpbuf *request, struct ofpbuf **replyp); +static int tc_add_del_ingress_qdisc(struct netdev *netdev, bool add); +static int tc_add_policer(struct netdev *netdev, int kbits_rate, + int kbits_burst); static int tc_parse_qdisc(const struct ofpbuf *, const char **kind, struct nlattr **options); @@ -346,6 +355,7 @@ struct netdev_dev_linux { struct shash_node *shash_node; 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. */ @@ -358,9 +368,8 @@ struct netdev_dev_linux { struct in_addr address, netmask; struct in6_addr in6; int mtu; - int carrier; - bool is_internal; /* Is this an openvswitch internal device? */ - bool is_tap; /* Is this a tuntap device? */ + bool carrier; + long long int carrier_resets; uint32_t kbits_rate; /* Policing data. */ uint32_t kbits_burst; bool have_vport_stats; @@ -382,15 +391,6 @@ static int af_inet_sock = -1; /* AF_INET, SOCK_DGRAM. */ /* A Netlink routing socket that is not subscribed to any multicast groups. */ static struct nl_sock *rtnl_sock; -struct netdev_linux_notifier { - struct netdev_notifier notifier; - struct list node; -}; - -static struct shash netdev_linux_notifiers = - SHASH_INITIALIZER(&netdev_linux_notifiers); -static struct rtnetlink_notifier netdev_linux_poll_notifier; - /* 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); @@ -415,8 +415,8 @@ static int set_etheraddr(const char *netdev_name, int hwaddr_family, 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_carrier_via_sysfs(const char *name, bool *carrier); static int af_packet_sock(void); -static void poll_notify(struct list *); static void netdev_linux_miimon_run(void); static void netdev_linux_miimon_wait(void); @@ -472,17 +472,27 @@ netdev_linux_init(void) static void netdev_linux_run(void) { - rtnetlink_link_notifier_run(); + rtnetlink_link_run(); netdev_linux_miimon_run(); } static void netdev_linux_wait(void) { - rtnetlink_link_notifier_wait(); + rtnetlink_link_wait(); netdev_linux_miimon_wait(); } +static void +netdev_dev_linux_changed(struct netdev_dev_linux *dev) +{ + dev->change_seq++; + if (!dev->change_seq) { + dev->change_seq++; + } + dev->cache_valid = 0; +} + static void netdev_linux_cache_cb(const struct rtnetlink_link_change *change, void *aux OVS_UNUSED) @@ -496,7 +506,13 @@ netdev_linux_cache_cb(const struct rtnetlink_link_change *change, if (is_netdev_linux_class(netdev_class)) { dev = netdev_dev_linux_cast(base_dev); - dev->cache_valid = 0; + + if (dev->carrier != change->running) { + dev->carrier = change->running; + dev->carrier_resets++; + } + + netdev_dev_linux_changed(dev); } } } else { @@ -506,38 +522,68 @@ netdev_linux_cache_cb(const struct rtnetlink_link_change *change, shash_init(&device_shash); netdev_dev_get_devices(&netdev_linux_class, &device_shash); SHASH_FOR_EACH (node, &device_shash) { + bool carrier; + dev = node->data; - dev->cache_valid = 0; + + get_carrier_via_sysfs(node->name, &carrier); + if (dev->carrier != carrier) { + dev->carrier = carrier; + dev->carrier_resets++; + } + + netdev_dev_linux_changed(dev); } shash_destroy(&device_shash); } } -/* Creates system and internal devices. */ static int -netdev_linux_create(const struct netdev_class *class, - const char *name, const struct shash *args, - struct netdev_dev **netdev_devp) +cache_notifier_ref(void) { - struct netdev_dev_linux *netdev_dev; - int error; + if (!cache_notifier_refcount) { + assert(!netdev_linux_cache_notifier); - if (!shash_is_empty(args)) { - VLOG_WARN("%s: arguments for %s devices should be empty", - name, class->type); - } + netdev_linux_cache_notifier = + rtnetlink_link_notifier_create(netdev_linux_cache_cb, NULL); - if (!cache_notifier_refcount) { - error = rtnetlink_link_notifier_register(&netdev_linux_cache_notifier, - netdev_linux_cache_cb, NULL); - if (error) { - return error; + if (!netdev_linux_cache_notifier) { + return EINVAL; } } cache_notifier_refcount++; + return 0; +} + +static void +cache_notifier_unref(void) +{ + assert(cache_notifier_refcount > 0); + if (!--cache_notifier_refcount) { + assert(netdev_linux_cache_notifier); + rtnetlink_link_notifier_destroy(netdev_linux_cache_notifier); + netdev_linux_cache_notifier = NULL; + } +} + +/* Creates system and internal devices. */ +static int +netdev_linux_create(const struct netdev_class *class, const char *name, + struct netdev_dev **netdev_devp) +{ + struct netdev_dev_linux *netdev_dev; + int error; + + error = cache_notifier_ref(); + if (error) { + return error; + } + netdev_dev = xzalloc(sizeof *netdev_dev); - netdev_dev_init(&netdev_dev->netdev_dev, name, args, class); + netdev_dev->change_seq = 1; + netdev_dev_init(&netdev_dev->netdev_dev, name, class); + get_carrier_via_sysfs(name, &netdev_dev->carrier); *netdev_devp = &netdev_dev->netdev_dev; return 0; @@ -551,8 +597,7 @@ netdev_linux_create(const struct netdev_class *class, * be unavailable to other reads for tap devices. */ static int netdev_linux_create_tap(const struct netdev_class *class OVS_UNUSED, - const char *name, const struct shash *args, - struct netdev_dev **netdev_devp) + const char *name, struct netdev_dev **netdev_devp) { struct netdev_dev_linux *netdev_dev; struct tap_state *state; @@ -560,19 +605,20 @@ netdev_linux_create_tap(const struct netdev_class *class OVS_UNUSED, struct ifreq ifr; int error; - if (!shash_is_empty(args)) { - VLOG_WARN("%s: arguments for TAP devices should be empty", name); - } - netdev_dev = xzalloc(sizeof *netdev_dev); state = &netdev_dev->state.tap; + error = cache_notifier_ref(); + if (error) { + goto error; + } + /* Open tap device. */ state->fd = open(tap_dev, O_RDWR); if (state->fd < 0) { error = errno; VLOG_WARN("opening \"%s\" failed: %s", tap_dev, strerror(error)); - goto error; + goto error_unref_notifier; } /* Create tap device. */ @@ -582,19 +628,21 @@ netdev_linux_create_tap(const struct netdev_class *class OVS_UNUSED, VLOG_WARN("%s: creating tap device failed: %s", name, strerror(errno)); error = errno; - goto error; + goto error_unref_notifier; } /* Make non-blocking. */ error = set_nonblocking(state->fd); if (error) { - goto error; + goto error_unref_notifier; } - netdev_dev_init(&netdev_dev->netdev_dev, name, args, &netdev_tap_class); + netdev_dev_init(&netdev_dev->netdev_dev, name, &netdev_tap_class); *netdev_devp = &netdev_dev->netdev_dev; return 0; +error_unref_notifier: + cache_notifier_unref(); error: free(netdev_dev); return error; @@ -621,24 +669,16 @@ netdev_linux_destroy(struct netdev_dev *netdev_dev_) netdev_dev->tc->ops->tc_destroy(netdev_dev->tc); } - if (class == &netdev_linux_class || class == &netdev_internal_class) { - cache_notifier_refcount--; - - if (!cache_notifier_refcount) { - rtnetlink_link_notifier_unregister(&netdev_linux_cache_notifier); - } - } else if (class == &netdev_tap_class) { + if (class == &netdev_tap_class) { destroy_tap(netdev_dev); - } else { - NOT_REACHED(); } - free(netdev_dev); + + cache_notifier_unref(); } static int -netdev_linux_open(struct netdev_dev *netdev_dev_, int ethertype, - struct netdev **netdevp) +netdev_linux_open(struct netdev_dev *netdev_dev_, struct netdev **netdevp) { struct netdev_dev_linux *netdev_dev = netdev_dev_linux_cast(netdev_dev_); struct netdev_linux *netdev; @@ -674,54 +714,6 @@ netdev_linux_open(struct netdev_dev *netdev_dev_, int ethertype, * directions appearing to be reversed. */ netdev->fd = netdev_dev->state.tap.fd; netdev_dev->state.tap.opened = true; - } else if (ethertype != NETDEV_ETH_TYPE_NONE) { - struct sockaddr_ll sll; - int protocol; - int ifindex; - - /* Create file descriptor. */ - protocol = (ethertype == NETDEV_ETH_TYPE_ANY ? ETH_P_ALL - : ethertype == NETDEV_ETH_TYPE_802_2 ? ETH_P_802_2 - : ethertype); - netdev->fd = socket(PF_PACKET, SOCK_RAW, - (OVS_FORCE int) htons(protocol)); - if (netdev->fd < 0) { - error = errno; - goto error; - } - - /* Set non-blocking mode. */ - error = set_nonblocking(netdev->fd); - if (error) { - goto error; - } - - /* Get ethernet device index. */ - error = get_ifindex(&netdev->netdev, &ifindex); - if (error) { - goto error; - } - - /* Bind to specific ethernet device. */ - memset(&sll, 0, sizeof sll); - sll.sll_family = AF_PACKET; - sll.sll_ifindex = ifindex; - if (bind(netdev->fd, - (struct sockaddr *) &sll, sizeof sll) < 0) { - error = errno; - VLOG_ERR("bind to %s failed: %s", netdev_dev_get_name(netdev_dev_), - strerror(error)); - goto error; - } - - /* Between the socket() and bind() calls above, the socket receives all - * packets of the requested type on all system interfaces. We do not - * want to receive that data, but there is no way to avoid it. So we - * must now drain out the receive queue. */ - error = drain_rcvbuf(netdev->fd); - if (error) { - goto error; - } } *netdevp = &netdev->netdev; @@ -744,26 +736,59 @@ netdev_linux_close(struct netdev *netdev_) free(netdev); } -/* Initializes 'sset' with a list of the names of all known network devices. */ static int -netdev_linux_enumerate(struct sset *sset) +netdev_linux_listen(struct netdev *netdev_) { - struct if_nameindex *names; - - names = if_nameindex(); - if (names) { - size_t i; + struct netdev_linux *netdev = netdev_linux_cast(netdev_); + struct sockaddr_ll sll; + int ifindex; + int error; + int fd; - for (i = 0; names[i].if_name != NULL; i++) { - sset_add(sset, names[i].if_name); - } - if_freenameindex(names); + if (netdev->fd >= 0) { return 0; - } else { - VLOG_WARN("could not obtain list of network device names: %s", - strerror(errno)); - return errno; } + + /* Create file descriptor. */ + fd = socket(PF_PACKET, SOCK_RAW, 0); + if (fd < 0) { + error = errno; + VLOG_ERR("failed to create raw socket (%s)", strerror(error)); + goto error; + } + + /* Set non-blocking mode. */ + error = set_nonblocking(fd); + if (error) { + goto error; + } + + /* Get ethernet device index. */ + error = get_ifindex(&netdev->netdev, &ifindex); + if (error) { + goto error; + } + + /* Bind to specific ethernet device. */ + 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); + if (bind(fd, (struct sockaddr *) &sll, sizeof sll) < 0) { + error = errno; + VLOG_ERR("%s: failed to bind raw socket (%s)", + netdev_get_name(netdev_), strerror(error)); + goto error; + } + + netdev->fd = fd; + return 0; + +error: + if (fd >= 0) { + close(fd); + } + return error; } static int @@ -772,13 +797,16 @@ netdev_linux_recv(struct netdev *netdev_, void *data, size_t size) struct netdev_linux *netdev = netdev_linux_cast(netdev_); if (netdev->fd < 0) { - /* Device was opened with NETDEV_ETH_TYPE_NONE. */ + /* Device is not listening. */ return -EAGAIN; } for (;;) { - ssize_t retval = read(netdev->fd, data, size); - if (retval >= 0) { + ssize_t retval = recv(netdev->fd, data, size, MSG_TRUNC); + if (retval > size) { + /* Received packet was longer than supplied buffer. */ + return -EMSGSIZE; + } else if (retval >= 0) { return retval; } else if (errno != EINTR) { if (errno != EAGAIN) { @@ -994,6 +1022,29 @@ netdev_linux_get_mtu(const struct netdev *netdev_, int *mtup) return 0; } +/* Sets the maximum size of transmitted (MTU) for given device using linux + * networking ioctl interface. + */ +static int +netdev_linux_set_mtu(const struct netdev *netdev_, int mtu) +{ + struct netdev_dev_linux *netdev_dev = + netdev_dev_linux_cast(netdev_get_dev(netdev_)); + struct ifreq ifr; + int error; + + ifr.ifr_mtu = mtu; + error = netdev_linux_do_ioctl(netdev_get_name(netdev_), &ifr, + SIOCSIFMTU, "SIOCSIFMTU"); + if (error) { + return error; + } + + netdev_dev->mtu = ifr.ifr_mtu; + netdev_dev->cache_valid |= VALID_MTU; + return 0; +} + /* Returns the ifindex of 'netdev', if successful, as a positive number. * On failure, returns a negative errno value. */ static int @@ -1010,62 +1061,20 @@ netdev_linux_get_carrier(const struct netdev *netdev_, bool *carrier) { struct netdev_dev_linux *netdev_dev = netdev_dev_linux_cast(netdev_get_dev(netdev_)); - int error = 0; - char *fn = NULL; - int fd = -1; if (netdev_dev->miimon_interval > 0) { *carrier = netdev_dev->miimon; - return 0; + } else { + *carrier = netdev_dev->carrier; } - if (!(netdev_dev->cache_valid & VALID_CARRIER)) { - char line[8]; - int retval; - - fn = xasprintf("/sys/class/net/%s/carrier", - netdev_get_name(netdev_)); - fd = open(fn, O_RDONLY); - if (fd < 0) { - error = errno; - VLOG_WARN_RL(&rl, "%s: open failed: %s", fn, strerror(error)); - goto exit; - } - - retval = read(fd, line, sizeof line); - if (retval < 0) { - error = errno; - if (error == EINVAL) { - /* This is the normal return value when we try to check carrier - * if the network device is not up. */ - } else { - VLOG_WARN_RL(&rl, "%s: read failed: %s", fn, strerror(error)); - } - goto exit; - } else if (retval == 0) { - error = EPROTO; - VLOG_WARN_RL(&rl, "%s: unexpected end of file", fn); - goto exit; - } - - if (line[0] != '0' && line[0] != '1') { - error = EPROTO; - VLOG_WARN_RL(&rl, "%s: value is %c (expected 0 or 1)", - fn, line[0]); - goto exit; - } - netdev_dev->carrier = line[0] != '0'; - netdev_dev->cache_valid |= VALID_CARRIER; - } - *carrier = netdev_dev->carrier; - error = 0; + return 0; +} -exit: - if (fd >= 0) { - close(fd); - } - free(fn); - return error; +static long long int +netdev_linux_get_carrier_resets(const struct netdev *netdev) +{ + return netdev_dev_linux_cast(netdev_get_dev(netdev))->carrier_resets; } static int @@ -1161,14 +1170,8 @@ netdev_linux_miimon_run(void) netdev_linux_get_miimon(dev->netdev_dev.name, &miimon); if (miimon != dev->miimon) { - struct list *list; - dev->miimon = miimon; - list = shash_find_data(&netdev_linux_notifiers, - dev->netdev_dev.name); - if (list) { - poll_notify(list); - } + netdev_dev_linux_changed(dev); } timer_set_duration(&dev->miimon_timer, dev->miimon_interval); @@ -1223,21 +1226,6 @@ check_for_working_netlink_stats(void) } } -/* Brings the 'is_internal' and 'is_tap' members of 'netdev_dev' up-to-date. */ -static void -netdev_linux_update_is_pseudo(struct netdev_dev_linux *netdev_dev) -{ - if (!(netdev_dev->cache_valid & VALID_IS_PSEUDO)) { - const char *name = netdev_dev_get_name(&netdev_dev->netdev_dev); - const char *type = netdev_dev_get_type(&netdev_dev->netdev_dev); - - netdev_dev->is_tap = !strcmp(type, "tap"); - 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) { @@ -1246,37 +1234,123 @@ swap_uint64(uint64_t *a, uint64_t *b) *b = tmp; } -/* Retrieves current device stats for 'netdev'. */ -static int -netdev_linux_get_stats(const struct netdev *netdev_, - struct netdev_stats *stats) +static void +get_stats_via_vport(const struct netdev *netdev_, + struct netdev_stats *stats) { struct netdev_dev_linux *netdev_dev = netdev_dev_linux_cast(netdev_get_dev(netdev_)); - static int use_netlink_stats = -1; - int error; if (netdev_dev->have_vport_stats || !(netdev_dev->cache_valid & VALID_HAVE_VPORT_STATS)) { + int error; error = netdev_vport_get_stats(netdev_, stats); + if (error) { + VLOG_WARN_RL(&rl, "%s: obtaining netdev stats via vport failed " + "(%s)", netdev_get_name(netdev_), strerror(error)); + } netdev_dev->have_vport_stats = !error; netdev_dev->cache_valid |= VALID_HAVE_VPORT_STATS; } +} - if (!netdev_dev->have_vport_stats) { - if (use_netlink_stats < 0) { - use_netlink_stats = check_for_working_netlink_stats(); +static int +netdev_linux_sys_get_stats(const struct netdev *netdev_, + struct netdev_stats *stats) +{ + static int use_netlink_stats = -1; + int error; + + if (use_netlink_stats < 0) { + use_netlink_stats = check_for_working_netlink_stats(); + } + + if (use_netlink_stats) { + int ifindex; + + error = get_ifindex(netdev_, &ifindex); + if (!error) { + error = get_stats_via_netlink(ifindex, stats); } - if (use_netlink_stats) { - int ifindex; + } else { + error = get_stats_via_proc(netdev_get_name(netdev_), stats); + } - error = get_ifindex(netdev_, &ifindex); - if (!error) { - error = get_stats_via_netlink(ifindex, 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_, + struct netdev_stats *stats) +{ + struct netdev_dev_linux *netdev_dev = + netdev_dev_linux_cast(netdev_get_dev(netdev_)); + struct netdev_stats dev_stats; + int error; + + get_stats_via_vport(netdev_, stats); + + error = netdev_linux_sys_get_stats(netdev_, &dev_stats); + + if (error) { + if (!netdev_dev->have_vport_stats) { + return error; } else { - error = get_stats_via_proc(netdev_get_name(netdev_), stats); + return 0; + } + } + + if (!netdev_dev->have_vport_stats) { + /* stats not available from OVS then use ioctl stats. */ + *stats = dev_stats; + } else { + stats->rx_errors += dev_stats.rx_errors; + stats->tx_errors += dev_stats.tx_errors; + stats->rx_dropped += dev_stats.rx_dropped; + stats->tx_dropped += dev_stats.tx_dropped; + stats->multicast += dev_stats.multicast; + stats->collisions += dev_stats.collisions; + stats->rx_length_errors += dev_stats.rx_length_errors; + stats->rx_over_errors += dev_stats.rx_over_errors; + stats->rx_crc_errors += dev_stats.rx_crc_errors; + stats->rx_frame_errors += dev_stats.rx_frame_errors; + stats->rx_fifo_errors += dev_stats.rx_fifo_errors; + stats->rx_missed_errors += dev_stats.rx_missed_errors; + stats->tx_aborted_errors += dev_stats.tx_aborted_errors; + stats->tx_carrier_errors += dev_stats.tx_carrier_errors; + stats->tx_fifo_errors += dev_stats.tx_fifo_errors; + stats->tx_heartbeat_errors += dev_stats.tx_heartbeat_errors; + stats->tx_window_errors += dev_stats.tx_window_errors; + } + return 0; +} + +/* Retrieves current device stats for 'netdev-tap' netdev or + * netdev-internal. */ +static int +netdev_pseudo_get_stats(const struct netdev *netdev_, + struct netdev_stats *stats) +{ + struct netdev_dev_linux *netdev_dev = + netdev_dev_linux_cast(netdev_get_dev(netdev_)); + struct netdev_stats dev_stats; + int error; + + get_stats_via_vport(netdev_, stats); + + error = netdev_linux_sys_get_stats(netdev_, &dev_stats); + if (error) { + if (!netdev_dev->have_vport_stats) { + return error; + } else { + return 0; } } @@ -1286,9 +1360,8 @@ netdev_linux_get_stats(const struct netdev *netdev_, * them back here. This does not apply if we are getting stats from the * vport layer because it always tracks stats from the perspective of the * switch. */ - netdev_linux_update_is_pseudo(netdev_dev); - if (!error && !netdev_dev->have_vport_stats && - (netdev_dev->is_internal || netdev_dev->is_tap)) { + if (!netdev_dev->have_vport_stats) { + *stats = dev_stats; swap_uint64(&stats->rx_packets, &stats->tx_packets); swap_uint64(&stats->rx_bytes, &stats->tx_bytes); swap_uint64(&stats->rx_errors, &stats->tx_errors); @@ -1304,9 +1377,17 @@ netdev_linux_get_stats(const struct netdev *netdev_, stats->tx_fifo_errors = 0; stats->tx_heartbeat_errors = 0; stats->tx_window_errors = 0; - } + } else { + stats->rx_dropped += dev_stats.tx_dropped; + stats->tx_dropped += dev_stats.rx_dropped; - return error; + stats->rx_errors += dev_stats.tx_errors; + stats->tx_errors += dev_stats.rx_errors; + + stats->multicast += dev_stats.multicast; + stats->collisions += dev_stats.collisions; + } + return 0; } /* Stores the features supported by 'netdev' into each of '*current', @@ -1490,104 +1571,8 @@ netdev_linux_set_advertisements(struct netdev *netdev, uint32_t advertise) ETHTOOL_SSET, "ETHTOOL_SSET"); } -/* If 'netdev_name' is the name of a VLAN network device (e.g. one created with - * vconfig(8)), sets '*vlan_vid' to the VLAN VID associated with that device - * and returns 0. Otherwise returns a errno value (specifically ENOENT if - * 'netdev_name' is the name of a network device that is not a VLAN device) and - * sets '*vlan_vid' to -1. */ -static int -netdev_linux_get_vlan_vid(const struct netdev *netdev, int *vlan_vid) -{ - const char *netdev_name = netdev_get_name(netdev); - struct ds line = DS_EMPTY_INITIALIZER; - FILE *stream = NULL; - int error; - char *fn; - - COVERAGE_INC(netdev_get_vlan_vid); - fn = xasprintf("/proc/net/vlan/%s", netdev_name); - stream = fopen(fn, "r"); - if (!stream) { - error = errno; - goto done; - } - - if (ds_get_line(&line, stream)) { - if (ferror(stream)) { - error = errno; - VLOG_ERR_RL(&rl, "error reading \"%s\": %s", fn, strerror(errno)); - } else { - error = EPROTO; - VLOG_ERR_RL(&rl, "unexpected end of file reading \"%s\"", fn); - } - goto done; - } - - if (!sscanf(ds_cstr(&line), "%*s VID: %d", vlan_vid)) { - error = EPROTO; - VLOG_ERR_RL(&rl, "parse error reading \"%s\" line 1: \"%s\"", - fn, ds_cstr(&line)); - goto done; - } - - error = 0; - -done: - free(fn); - if (stream) { - fclose(stream); - } - ds_destroy(&line); - if (error) { - *vlan_vid = -1; - } - return error; -} - -#define POLICE_ADD_CMD "/sbin/tc qdisc add dev %s handle ffff: ingress" -#define POLICE_CONFIG_CMD "/sbin/tc filter add dev %s parent ffff: protocol ip prio 50 u32 match ip src 0.0.0.0/0 police rate %dkbit burst %dk mtu 65535 drop flowid :1" - -/* Remove ingress policing from 'netdev'. Returns 0 if successful, otherwise a - * positive errno value. - * - * This function is equivalent to running - * /sbin/tc qdisc del dev %s handle ffff: ingress - * but it is much, much faster. - */ -static int -netdev_linux_remove_policing(struct netdev *netdev) -{ - struct netdev_dev_linux *netdev_dev = - netdev_dev_linux_cast(netdev_get_dev(netdev)); - const char *netdev_name = netdev_get_name(netdev); - - struct ofpbuf request; - struct tcmsg *tcmsg; - int error; - - tcmsg = tc_make_request(netdev, RTM_DELQDISC, 0, &request); - if (!tcmsg) { - return ENODEV; - } - tcmsg->tcm_handle = tc_make_handle(0xffff, 0); - tcmsg->tcm_parent = TC_H_INGRESS; - nl_msg_put_string(&request, TCA_KIND, "ingress"); - nl_msg_put_unspec(&request, TCA_OPTIONS, NULL, 0); - - error = tc_transact(&request, NULL); - if (error && error != ENOENT && error != EINVAL) { - VLOG_WARN_RL(&rl, "%s: removing policing failed: %s", - netdev_name, strerror(error)); - return error; - } - - netdev_dev->kbits_rate = 0; - netdev_dev->kbits_burst = 0; - netdev_dev->cache_valid |= VALID_POLICING; - return 0; -} - -/* Attempts to set input rate limiting (policing) policy. */ +/* Attempts to set input rate limiting (policing) policy. Returns 0 if + * successful, otherwise a positive errno value. */ static int netdev_linux_set_policing(struct netdev *netdev, uint32_t kbits_rate, uint32_t kbits_burst) @@ -1595,7 +1580,7 @@ netdev_linux_set_policing(struct netdev *netdev, struct netdev_dev_linux *netdev_dev = netdev_dev_linux_cast(netdev_get_dev(netdev)); const char *netdev_name = netdev_get_name(netdev); - char command[1024]; + int error; COVERAGE_INC(netdev_set_policing); @@ -1610,27 +1595,34 @@ netdev_linux_set_policing(struct netdev *netdev, return 0; } - netdev_linux_remove_policing(netdev); + /* Remove any existing ingress qdisc. */ + error = tc_add_del_ingress_qdisc(netdev, false); + if (error) { + VLOG_WARN_RL(&rl, "%s: removing policing failed: %s", + netdev_name, strerror(error)); + return error; + } + if (kbits_rate) { - snprintf(command, sizeof(command), POLICE_ADD_CMD, netdev_name); - if (system(command) != 0) { - VLOG_WARN_RL(&rl, "%s: problem adding policing", netdev_name); - return -1; + error = tc_add_del_ingress_qdisc(netdev, true); + if (error) { + VLOG_WARN_RL(&rl, "%s: adding policing qdisc failed: %s", + netdev_name, strerror(error)); + return error; } - snprintf(command, sizeof(command), POLICE_CONFIG_CMD, netdev_name, - kbits_rate, kbits_burst); - if (system(command) != 0) { - VLOG_WARN_RL(&rl, "%s: problem configuring policing", - netdev_name); - return -1; + error = tc_add_policer(netdev, kbits_rate, kbits_burst); + if (error){ + VLOG_WARN_RL(&rl, "%s: adding policing action failed: %s", + netdev_name, strerror(error)); + return error; } - - netdev_dev->kbits_rate = kbits_rate; - netdev_dev->kbits_burst = kbits_burst; - netdev_dev->cache_valid |= VALID_POLICING; } + netdev_dev->kbits_rate = kbits_rate; + netdev_dev->kbits_burst = kbits_burst; + netdev_dev->cache_valid |= VALID_POLICING; + return 0; } @@ -2234,91 +2226,13 @@ netdev_linux_update_flags(struct netdev *netdev, enum netdev_flags off, return error; } -static void -poll_notify(struct list *list) -{ - struct netdev_linux_notifier *notifier; - LIST_FOR_EACH (notifier, node, list) { - struct netdev_notifier *n = ¬ifier->notifier; - n->cb(n); - } -} - -static void -netdev_linux_poll_cb(const struct rtnetlink_link_change *change, - void *aux OVS_UNUSED) -{ - if (change) { - struct list *list = shash_find_data(&netdev_linux_notifiers, - change->ifname); - if (list) { - poll_notify(list); - } - } else { - struct shash_node *node; - SHASH_FOR_EACH (node, &netdev_linux_notifiers) { - poll_notify(node->data); - } - } -} - -static int -netdev_linux_poll_add(struct netdev *netdev, - void (*cb)(struct netdev_notifier *), void *aux, - struct netdev_notifier **notifierp) -{ - const char *netdev_name = netdev_get_name(netdev); - struct netdev_linux_notifier *notifier; - struct list *list; - - if (shash_is_empty(&netdev_linux_notifiers)) { - int error; - error = rtnetlink_link_notifier_register(&netdev_linux_poll_notifier, - netdev_linux_poll_cb, NULL); - if (error) { - return error; - } - } - - list = shash_find_data(&netdev_linux_notifiers, netdev_name); - if (!list) { - list = xmalloc(sizeof *list); - list_init(list); - shash_add(&netdev_linux_notifiers, netdev_name, list); - } - - notifier = xmalloc(sizeof *notifier); - netdev_notifier_init(¬ifier->notifier, netdev, cb, aux); - list_push_back(list, ¬ifier->node); - *notifierp = ¬ifier->notifier; - return 0; -} - -static void -netdev_linux_poll_remove(struct netdev_notifier *notifier_) +static unsigned int +netdev_linux_change_seq(const struct netdev *netdev) { - struct netdev_linux_notifier *notifier = - CONTAINER_OF(notifier_, struct netdev_linux_notifier, notifier); - struct list *list; - - /* Remove 'notifier' from its list. */ - list = list_remove(¬ifier->node); - if (list_is_empty(list)) { - /* The list is now empty. Remove it from the hash and free it. */ - const char *netdev_name = netdev_get_name(notifier->notifier.netdev); - shash_delete(&netdev_linux_notifiers, - shash_find(&netdev_linux_notifiers, netdev_name)); - free(list); - } - free(notifier); - - /* If that was the last notifier, unregister. */ - if (shash_is_empty(&netdev_linux_notifiers)) { - rtnetlink_link_notifier_unregister(&netdev_linux_poll_notifier); - } + return netdev_dev_linux_cast(netdev_get_dev(netdev))->change_seq; } -#define NETDEV_LINUX_CLASS(NAME, CREATE, ENUMERATE, SET_STATS) \ +#define NETDEV_LINUX_CLASS(NAME, CREATE, GET_STATS, SET_STATS) \ { \ NAME, \ \ @@ -2328,13 +2242,13 @@ netdev_linux_poll_remove(struct netdev_notifier *notifier_) \ CREATE, \ netdev_linux_destroy, \ + NULL, /* get_config */ \ NULL, /* set_config */ \ \ netdev_linux_open, \ netdev_linux_close, \ \ - ENUMERATE, \ - \ + netdev_linux_listen, \ netdev_linux_recv, \ netdev_linux_recv_wait, \ netdev_linux_drain, \ @@ -2345,15 +2259,16 @@ netdev_linux_poll_remove(struct netdev_notifier *notifier_) netdev_linux_set_etheraddr, \ netdev_linux_get_etheraddr, \ netdev_linux_get_mtu, \ + netdev_linux_set_mtu, \ netdev_linux_get_ifindex, \ netdev_linux_get_carrier, \ + netdev_linux_get_carrier_resets, \ netdev_linux_set_miimon_interval, \ - netdev_linux_get_stats, \ + GET_STATS, \ SET_STATS, \ \ netdev_linux_get_features, \ netdev_linux_set_advertisements, \ - netdev_linux_get_vlan_vid, \ \ netdev_linux_set_policing, \ netdev_linux_get_qos_types, \ @@ -2377,29 +2292,28 @@ netdev_linux_poll_remove(struct netdev_notifier *notifier_) \ netdev_linux_update_flags, \ \ - netdev_linux_poll_add, \ - netdev_linux_poll_remove \ + netdev_linux_change_seq \ } const struct netdev_class netdev_linux_class = NETDEV_LINUX_CLASS( "system", netdev_linux_create, - netdev_linux_enumerate, + netdev_linux_get_stats, NULL); /* set_stats */ const struct netdev_class netdev_tap_class = NETDEV_LINUX_CLASS( "tap", netdev_linux_create_tap, - NULL, /* enumerate */ + netdev_pseudo_get_stats, NULL); /* set_stats */ const struct netdev_class netdev_internal_class = NETDEV_LINUX_CLASS( "internal", netdev_linux_create, - NULL, /* enumerate */ + netdev_pseudo_get_stats, netdev_vport_set_stats); /* HTB traffic control class. */ @@ -2489,11 +2403,11 @@ htb_setup_class__(struct netdev *netdev, unsigned int handle, int error; int mtu; - netdev_get_mtu(netdev, &mtu); - if (mtu == INT_MAX) { + error = netdev_get_mtu(netdev, &mtu); + if (error) { VLOG_WARN_RL(&rl, "cannot set up HTB on device %s that lacks MTU", netdev_get_name(netdev)); - return EINVAL; + return error; } memset(&opt, 0, sizeof opt); @@ -2612,13 +2526,13 @@ htb_parse_class_details__(struct netdev *netdev, const char *max_rate_s = shash_find_data(details, "max-rate"); const char *burst_s = shash_find_data(details, "burst"); const char *priority_s = shash_find_data(details, "priority"); - int mtu; + int mtu, error; - netdev_get_mtu(netdev, &mtu); - if (mtu == INT_MAX) { + error = netdev_get_mtu(netdev, &mtu); + if (error) { VLOG_WARN_RL(&rl, "cannot parse HTB class on device %s that lacks MTU", netdev_get_name(netdev)); - return EINVAL; + return error; } /* HTB requires at least an mtu sized min-rate to send any traffic even @@ -3549,6 +3463,107 @@ tc_transact(struct ofpbuf *request, struct ofpbuf **replyp) return error; } +/* Adds or deletes a root ingress qdisc on 'netdev'. We use this for + * policing configuration. + * + * This function is equivalent to running the following when 'add' is true: + * /sbin/tc qdisc add dev handle ffff: ingress + * + * This function is equivalent to running the following when 'add' is false: + * /sbin/tc qdisc del dev handle ffff: ingress + * + * The configuration and stats may be seen with the following command: + * /sbin/tc -s qdisc show dev + * + * Returns 0 if successful, otherwise a positive errno value. + */ +static int +tc_add_del_ingress_qdisc(struct netdev *netdev, bool add) +{ + struct ofpbuf request; + struct tcmsg *tcmsg; + int error; + int type = add ? RTM_NEWQDISC : RTM_DELQDISC; + int flags = add ? NLM_F_EXCL | NLM_F_CREATE : 0; + + tcmsg = tc_make_request(netdev, type, flags, &request); + if (!tcmsg) { + return ENODEV; + } + tcmsg->tcm_handle = tc_make_handle(0xffff, 0); + tcmsg->tcm_parent = TC_H_INGRESS; + nl_msg_put_string(&request, TCA_KIND, "ingress"); + nl_msg_put_unspec(&request, TCA_OPTIONS, NULL, 0); + + error = tc_transact(&request, NULL); + if (error) { + /* If we're deleting the qdisc, don't worry about some of the + * error conditions. */ + if (!add && (error == ENOENT || error == EINVAL)) { + return 0; + } + return error; + } + + return 0; +} + +/* Adds a policer to 'netdev' with a rate of 'kbits_rate' and a burst size + * of 'kbits_burst'. + * + * This function is equivalent to running: + * /sbin/tc filter add dev parent ffff: protocol all prio 49 + * basic police rate kbit burst k + * mtu 65535 drop + * + * The configuration and stats may be seen with the following command: + * /sbin/tc -s filter show eth0 parent ffff: + * + * Returns 0 if successful, otherwise a positive errno value. + */ +static int +tc_add_policer(struct netdev *netdev, int kbits_rate, int kbits_burst) +{ + struct tc_police tc_police; + struct ofpbuf request; + struct tcmsg *tcmsg; + size_t basic_offset; + size_t police_offset; + int error; + int mtu = 65535; + + 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/8 * 1000, mtu); + tc_police.burst = tc_bytes_to_ticks(tc_police.rate.rate, + kbits_burst * 1024); + + tcmsg = tc_make_request(netdev, RTM_NEWTFILTER, + NLM_F_EXCL | NLM_F_CREATE, &request); + if (!tcmsg) { + return ENODEV; + } + tcmsg->tcm_parent = tc_make_handle(0xffff, 0); + tcmsg->tcm_info = tc_make_handle(49, + (OVS_FORCE uint16_t) htons(ETH_P_ALL)); + + nl_msg_put_string(&request, TCA_KIND, "basic"); + basic_offset = nl_msg_start_nested(&request, TCA_OPTIONS); + police_offset = nl_msg_start_nested(&request, TCA_BASIC_POLICE); + nl_msg_put_unspec(&request, TCA_POLICE_TBF, &tc_police, sizeof tc_police); + tc_put_rtab(&request, TCA_POLICE_RATE, &tc_police.rate); + nl_msg_end_nested(&request, police_offset); + nl_msg_end_nested(&request, basic_offset); + + error = tc_transact(&request, NULL); + if (error) { + return error; + } + + return 0; +} + static void read_psched(void) { @@ -4009,58 +4024,90 @@ tc_calc_buffer(unsigned int Bps, int mtu, uint64_t burst_bytes) return tc_bytes_to_ticks(Bps, MAX(burst_bytes, min_burst)); } -/* 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 +/* Linux-only functions declared in netdev-linux.h */ -/* 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) +/* Returns a fd for an AF_INET socket or a negative errno value. */ +int +netdev_linux_get_af_inet_sock(void) { - COPY_NETDEV_STATS; + int error = netdev_linux_init(); + return error ? -error : af_inet_sock; } -/* 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) +/* Modifies the 'flag' bit in ethtool's flags field for 'netdev'. If + * 'enable' is true, the bit is set. Otherwise, it is cleared. */ +int +netdev_linux_ethtool_set_flag(struct netdev *netdev, uint32_t flag, + const char *flag_name, bool enable) { - COPY_NETDEV_STATS; + const char *netdev_name = netdev_get_name(netdev); + struct ethtool_value evalue; + uint32_t new_flags; + int error; + + memset(&evalue, 0, sizeof evalue); + error = netdev_linux_do_ethtool(netdev_name, + (struct ethtool_cmd *)&evalue, + ETHTOOL_GFLAGS, "ETHTOOL_GFLAGS"); + if (error) { + return error; + } + + evalue.data = new_flags = (evalue.data & ~flag) | (enable ? flag : 0); + error = netdev_linux_do_ethtool(netdev_name, + (struct ethtool_cmd *)&evalue, + ETHTOOL_SFLAGS, "ETHTOOL_SFLAGS"); + if (error) { + return error; + } + + memset(&evalue, 0, sizeof evalue); + error = netdev_linux_do_ethtool(netdev_name, + (struct ethtool_cmd *)&evalue, + ETHTOOL_GFLAGS, "ETHTOOL_GFLAGS"); + if (error) { + return error; + } + + if (new_flags != evalue.data) { + VLOG_WARN_RL(&rl, "attempt to %s ethtool %s flag on network " + "device %s failed", enable ? "enable" : "disable", + flag_name, netdev_name); + return EOPNOTSUPP; + } + + return 0; } + +/* Utility functions. */ /* 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) +static void +netdev_stats_from_rtnl_link_stats(struct netdev_stats *dst, + const struct rtnl_link_stats *src) { - COPY_NETDEV_STATS; - dst->rx_compressed = 0; - dst->tx_compressed = 0; + 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; } - -/* Utility functions. */ static int get_stats_via_netlink(int ifindex, struct netdev_stats *stats) @@ -4169,6 +4216,58 @@ get_stats_via_proc(const char *netdev_name, struct netdev_stats *stats) return ENODEV; } +static int +get_carrier_via_sysfs(const char *name, bool *carrier) +{ + char line[8]; + int retval; + + int error = 0; + char *fn = NULL; + int fd = -1; + + *carrier = false; + + fn = xasprintf("/sys/class/net/%s/carrier", name); + fd = open(fn, O_RDONLY); + if (fd < 0) { + error = errno; + VLOG_WARN_RL(&rl, "%s: open failed: %s", fn, strerror(error)); + goto exit; + } + + retval = read(fd, line, sizeof line); + if (retval < 0) { + error = errno; + if (error == EINVAL) { + /* This is the normal return value when we try to check carrier if + * the network device is not up. */ + } else { + VLOG_WARN_RL(&rl, "%s: read failed: %s", fn, strerror(error)); + } + goto exit; + } else if (retval == 0) { + error = EPROTO; + VLOG_WARN_RL(&rl, "%s: unexpected end of file", fn); + goto exit; + } + + if (line[0] != '0' && line[0] != '1') { + error = EPROTO; + VLOG_WARN_RL(&rl, "%s: value is %c (expected 0 or 1)", fn, line[0]); + goto exit; + } + *carrier = line[0] != '0'; + error = 0; + +exit: + if (fd >= 0) { + close(fd); + } + free(fn); + return error; +} + static int get_flags(const struct netdev *netdev, int *flags) { @@ -4234,8 +4333,12 @@ get_etheraddr(const char *netdev_name, uint8_t ea[ETH_ADDR_LEN]) 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", - netdev_name, strerror(errno)); + /* ENODEV probably means that a vif disappeared asynchronously and + * hasn't been removed from the database yet, so reduce the log level + * to INFO for that case. */ + VLOG(errno == ENODEV ? VLL_INFO : VLL_ERR, + "ioctl(SIOCGIFHWADDR) on %s device failed: %s", + netdev_name, strerror(errno)); return errno; } hwaddr_family = ifr.ifr_hwaddr.sa_family;