X-Git-Url: http://git.onelab.eu/?a=blobdiff_plain;f=lib%2Fnetdev-linux.c;h=5bbaf632f26d38f4172660a2c57fd7ad50ae215a;hb=2f980d741737fd3296d8063d800753a2d77bd730;hp=27a123cd4ea9c87642d04f7d7dbb3110898e5205;hpb=02e83e83b4c8d88f50317ba56fec53cd6d3fb909;p=sliver-openvswitch.git diff --git a/lib/netdev-linux.c b/lib/netdev-linux.c index 27a123cd4..5bbaf632f 100644 --- a/lib/netdev-linux.c +++ b/lib/netdev-linux.c @@ -1,5 +1,5 @@ /* - * Copyright (c) 2009, 2010, 2011 Nicira Networks. + * Copyright (c) 2009, 2010, 2011, 2012, 2013 Nicira, Inc. * * Licensed under the Apache License, Version 2.0 (the "License"); * you may not use this file except in compliance with the License. @@ -18,15 +18,14 @@ #include "netdev-linux.h" -#include #include #include #include #include +#include #include #include #include -#include #include #include #include @@ -57,18 +56,19 @@ #include "hmap.h" #include "netdev-provider.h" #include "netdev-vport.h" -#include "netlink.h" #include "netlink-notifier.h" #include "netlink-socket.h" +#include "netlink.h" #include "ofpbuf.h" #include "openflow/openflow.h" #include "packets.h" #include "poll-loop.h" #include "rtnetlink-link.h" -#include "socket-util.h" #include "shash.h" +#include "socket-util.h" #include "sset.h" #include "timer.h" +#include "unaligned.h" #include "vlog.h" VLOG_DEFINE_THIS_MODULE(netdev_linux); @@ -78,7 +78,9 @@ COVERAGE_DEFINE(netdev_arp_lookup); COVERAGE_DEFINE(netdev_get_ifindex); COVERAGE_DEFINE(netdev_get_hwaddr); COVERAGE_DEFINE(netdev_set_hwaddr); -COVERAGE_DEFINE(netdev_ethtool); +COVERAGE_DEFINE(netdev_get_ethtool); +COVERAGE_DEFINE(netdev_set_ethtool); + /* These were introduced in Linux 2.6.14, so they might be missing if we have * old headers. */ @@ -114,12 +116,9 @@ enum { VALID_IN6 = 1 << 3, VALID_MTU = 1 << 4, VALID_POLICING = 1 << 5, - VALID_HAVE_VPORT_STATS = 1 << 6 -}; - -struct tap_state { - int fd; - bool opened; + VALID_VPORT_STAT_ERROR = 1 << 6, + VALID_DRVINFO = 1 << 7, + VALID_FEATURES = 1 << 8, }; /* Traffic control. */ @@ -136,6 +135,8 @@ struct tc { * Written only by TC implementation. */ }; +#define TC_INITIALIZER(TC, OPS) { OPS, HMAP_INITIALIZER(&(TC)->queues) } + /* One traffic control queue. * * Each TC implementation subclasses this with whatever additional data it @@ -143,6 +144,7 @@ struct tc { struct tc_queue { struct hmap_node hmap_node; /* In struct tc's "queues" hmap. */ unsigned int queue_id; /* OpenFlow queue ID. */ + long long int created; /* Time queue was created, in msecs. */ }; /* A particular kind of traffic control. Each implementation generally maps to @@ -179,7 +181,7 @@ struct tc_ops { * * (This function is null for tc_ops_other, which cannot be installed. For * other TC classes it should always be nonnull.) */ - int (*tc_install)(struct netdev *netdev, const struct shash *details); + int (*tc_install)(struct netdev *netdev, const struct smap *details); /* Called when the netdev code determines (through a Netlink query) that * this TC class's qdisc is installed on 'netdev', but we didn't install @@ -219,7 +221,7 @@ struct tc_ops { * * This function may be null if 'tc' is not configurable. */ - int (*qdisc_get)(const struct netdev *netdev, struct shash *details); + int (*qdisc_get)(const struct netdev *netdev, struct smap *details); /* Reconfigures 'netdev->tc' according to 'details', performing any * required Netlink calls to complete the reconfiguration. @@ -230,7 +232,7 @@ struct tc_ops { * * This function may be null if 'tc' is not configurable. */ - int (*qdisc_set)(struct netdev *, const struct shash *details); + int (*qdisc_set)(struct netdev *, const struct smap *details); /* Retrieves details of 'queue' on 'netdev->tc' into 'details'. 'queue' is * one of the 'struct tc_queue's within 'netdev->tc->queues'. @@ -246,7 +248,7 @@ struct tc_ops { * This function may be null if 'tc' does not have queues ('n_queues' is * 0). */ int (*class_get)(const struct netdev *netdev, const struct tc_queue *queue, - struct shash *details); + struct smap *details); /* Configures or reconfigures 'queue_id' on 'netdev->tc' according to * 'details', perfoming any required Netlink calls to complete the @@ -260,7 +262,7 @@ struct tc_ops { * This function may be null if 'tc' does not have queues or its queues are * not configurable. */ int (*class_set)(struct netdev *, unsigned int queue_id, - const struct shash *details); + const struct smap *details); /* Deletes 'queue' from 'netdev->tc'. 'queue' is one of the 'struct * tc_queue's within 'netdev->tc->queues'. @@ -308,7 +310,7 @@ static const struct tc_ops tc_ops_hfsc; static const struct tc_ops tc_ops_default; static const struct tc_ops tc_ops_other; -static const struct tc_ops *tcs[] = { +static const struct tc_ops *const tcs[] = { &tc_ops_htb, /* Hierarchy token bucket (see tc-htb(8)). */ &tc_ops_hfsc, /* Hierarchical fair service curve. */ &tc_ops_default, /* Default qdisc (see tc-pfifo_fast(8)). */ @@ -350,10 +352,9 @@ static void tc_put_rtab(struct ofpbuf *, uint16_t type, const struct tc_ratespec *rate); static int tc_calc_buffer(unsigned int Bps, int mtu, uint64_t burst_bytes); -struct netdev_dev_linux { - struct netdev_dev netdev_dev; +struct netdev_linux { + struct netdev up; - struct shash_node *shash_node; unsigned int cache_valid; unsigned int change_seq; @@ -368,29 +369,40 @@ struct netdev_dev_linux { struct in_addr address, netmask; struct in6_addr in6; int mtu; - bool carrier; + unsigned int ifi_flags; long long int carrier_resets; uint32_t kbits_rate; /* Policing data. */ uint32_t kbits_burst; - bool have_vport_stats; + int vport_stats_error; /* Cached error code from vport_get_stats(). + 0 or an errno value. */ + int netdev_mtu_error; /* Cached error code from SIOCGIFMTU or SIOCSIFMTU. */ + int ether_addr_error; /* Cached error code from set/get etheraddr. */ + int netdev_policing_error; /* Cached error code from set policing. */ + int get_features_error; /* Cached error code from ETHTOOL_GSET. */ + int get_ifindex_error; /* Cached error code from SIOCGIFINDEX. */ + + enum netdev_features current; /* Cached from ETHTOOL_GSET. */ + enum netdev_features advertised; /* Cached from ETHTOOL_GSET. */ + enum netdev_features supported; /* Cached from ETHTOOL_GSET. */ + + struct ethtool_drvinfo drvinfo; /* Cached from ETHTOOL_GDRVINFO. */ struct tc *tc; - union { - struct tap_state tap; - } state; + /* For devices of class netdev_tap_class only. */ + int tap_fd; }; -struct netdev_linux { - struct netdev netdev; +struct netdev_rx_linux { + struct netdev_rx up; + bool is_tap; int fd; }; +static const struct netdev_rx_class netdev_rx_linux_class; + /* Sockets used for ioctl operations. */ 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; - /* 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); @@ -403,19 +415,17 @@ static int netdev_linux_do_ioctl(const char *name, struct ifreq *, int cmd, const char *cmd_name); static int netdev_linux_get_ipv4(const struct netdev *, struct in_addr *, int cmd, const char *cmd_name); -static int get_flags(const struct netdev *, int *flagsp); -static int set_flags(struct netdev *, int flags); +static int get_flags(const struct netdev *, unsigned int *flags); +static int set_flags(const char *, unsigned int flags); static int do_get_ifindex(const char *netdev_name); static int get_ifindex(const struct netdev *, int *ifindexp); static int do_set_addr(struct netdev *netdev, int ioctl_nr, const char *ioctl_name, 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, int hwaddr_family, - const uint8_t[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_carrier_via_sysfs(const char *name, bool *carrier); static int af_packet_sock(void); static void netdev_linux_miimon_run(void); static void netdev_linux_miimon_wait(void); @@ -426,23 +436,25 @@ is_netdev_linux_class(const struct netdev_class *netdev_class) return netdev_class->init == netdev_linux_init; } -static struct netdev_dev_linux * -netdev_dev_linux_cast(const struct netdev_dev *netdev_dev) +static bool +is_tap_netdev(const struct netdev *netdev) { - const struct netdev_class *netdev_class = netdev_dev_get_class(netdev_dev); - assert(is_netdev_linux_class(netdev_class)); - - return CONTAINER_OF(netdev_dev, struct netdev_dev_linux, netdev_dev); + return netdev_get_class(netdev) == &netdev_tap_class; } static struct netdev_linux * netdev_linux_cast(const struct netdev *netdev) { - struct netdev_dev *netdev_dev = netdev_get_dev(netdev); - const struct netdev_class *netdev_class = netdev_dev_get_class(netdev_dev); - assert(is_netdev_linux_class(netdev_class)); + ovs_assert(is_netdev_linux_class(netdev_get_class(netdev))); - return CONTAINER_OF(netdev, struct netdev_linux, netdev); + return CONTAINER_OF(netdev, struct netdev_linux, up); +} + +static struct netdev_rx_linux * +netdev_rx_linux_cast(const struct netdev_rx *rx) +{ + netdev_rx_assert_class(rx, &netdev_rx_linux_class); + return CONTAINER_OF(rx, struct netdev_rx_linux, up); } static int @@ -454,16 +466,7 @@ netdev_linux_init(void) af_inet_sock = socket(AF_INET, SOCK_DGRAM, 0); status = af_inet_sock >= 0 ? 0 : errno; if (status) { - VLOG_ERR("failed to create inet socket: %s", strerror(status)); - } - - /* Create rtnetlink socket. */ - if (!status) { - status = nl_sock_create(NETLINK_ROUTE, &rtnl_sock); - if (status) { - VLOG_ERR_RL(&rl, "failed to create rtnetlink socket: %s", - strerror(status)); - } + VLOG_ERR("failed to create inet socket: %s", ovs_strerror(status)); } } return status; @@ -484,55 +487,75 @@ netdev_linux_wait(void) } static void -netdev_dev_linux_changed(struct netdev_dev_linux *dev) +netdev_linux_changed(struct netdev_linux *dev, + unsigned int ifi_flags, unsigned int mask) { dev->change_seq++; if (!dev->change_seq) { dev->change_seq++; } - dev->cache_valid = 0; + + if ((dev->ifi_flags ^ ifi_flags) & IFF_RUNNING) { + dev->carrier_resets++; + } + dev->ifi_flags = ifi_flags; + + dev->cache_valid &= mask; +} + +static void +netdev_linux_update(struct netdev_linux *dev, + const struct rtnetlink_link_change *change) +{ + if (change->nlmsg_type == RTM_NEWLINK) { + /* Keep drv-info */ + netdev_linux_changed(dev, change->ifi_flags, VALID_DRVINFO); + + /* Update netdev from rtnl-change msg. */ + if (change->mtu) { + dev->mtu = change->mtu; + dev->cache_valid |= VALID_MTU; + dev->netdev_mtu_error = 0; + } + + if (!eth_addr_is_zero(change->addr)) { + memcpy(dev->etheraddr, change->addr, ETH_ADDR_LEN); + dev->cache_valid |= VALID_ETHERADDR; + dev->ether_addr_error = 0; + } + + dev->ifindex = change->ifi_index; + dev->cache_valid |= VALID_IFINDEX; + dev->get_ifindex_error = 0; + + } else { + netdev_linux_changed(dev, change->ifi_flags, 0); + } } static void netdev_linux_cache_cb(const struct rtnetlink_link_change *change, void *aux OVS_UNUSED) { - struct netdev_dev_linux *dev; if (change) { - struct netdev_dev *base_dev = netdev_dev_from_name(change->ifname); - if (base_dev) { - const struct netdev_class *netdev_class = - netdev_dev_get_class(base_dev); - - if (is_netdev_linux_class(netdev_class)) { - dev = netdev_dev_linux_cast(base_dev); - - if (dev->carrier != change->running) { - dev->carrier = change->running; - dev->carrier_resets++; - } - - netdev_dev_linux_changed(dev); - } + struct netdev *base_dev = netdev_from_name(change->ifname); + if (base_dev && is_netdev_linux_class(netdev_get_class(base_dev))) { + netdev_linux_update(netdev_linux_cast(base_dev), change); } } else { struct shash device_shash; struct shash_node *node; shash_init(&device_shash); - netdev_dev_get_devices(&netdev_linux_class, &device_shash); + netdev_get_devices(&netdev_linux_class, &device_shash); SHASH_FOR_EACH (node, &device_shash) { - bool carrier; + struct netdev *netdev = node->data; + struct netdev_linux *dev = netdev_linux_cast(netdev); + unsigned int flags; - dev = node->data; - - get_carrier_via_sysfs(node->name, &carrier); - if (dev->carrier != carrier) { - dev->carrier = carrier; - dev->carrier_resets++; - } - - netdev_dev_linux_changed(dev); + get_flags(&dev->up, &flags); + netdev_linux_changed(dev, flags, 0); + netdev_close(netdev); } shash_destroy(&device_shash); } @@ -542,7 +565,7 @@ static int cache_notifier_ref(void) { if (!cache_notifier_refcount) { - assert(!netdev_linux_cache_notifier); + ovs_assert(!netdev_linux_cache_notifier); netdev_linux_cache_notifier = rtnetlink_link_notifier_create(netdev_linux_cache_cb, NULL); @@ -559,9 +582,9 @@ cache_notifier_ref(void) static void cache_notifier_unref(void) { - assert(cache_notifier_refcount > 0); + ovs_assert(cache_notifier_refcount > 0); if (!--cache_notifier_refcount) { - assert(netdev_linux_cache_notifier); + ovs_assert(netdev_linux_cache_notifier); rtnetlink_link_notifier_destroy(netdev_linux_cache_notifier); netdev_linux_cache_notifier = NULL; } @@ -570,9 +593,9 @@ cache_notifier_unref(void) /* 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 **netdevp) { - struct netdev_dev_linux *netdev_dev; + struct netdev_linux *netdev; int error; error = cache_notifier_ref(); @@ -580,12 +603,26 @@ netdev_linux_create(const struct netdev_class *class, const char *name, return error; } - netdev_dev = xzalloc(sizeof *netdev_dev); - netdev_dev->change_seq = 1; - netdev_dev_init(&netdev_dev->netdev_dev, name, class); - get_carrier_via_sysfs(name, &netdev_dev->carrier); + netdev = xzalloc(sizeof *netdev); + netdev->change_seq = 1; + netdev_init(&netdev->up, name, class); + error = get_flags(&netdev->up, &netdev->ifi_flags); + if (error == ENODEV) { + if (class != &netdev_internal_class) { + /* The device does not exist, so don't allow it to be opened. */ + netdev_uninit(&netdev->up, false); + cache_notifier_unref(); + free(netdev); + return ENODEV; + } else { + /* "Internal" netdevs have to be created as netdev objects before + * they exist in the kernel, because creating them in the kernel + * happens by passing a netdev object to dpif_port_add(). + * Therefore, ignore the error. */ + } + } - *netdev_devp = &netdev_dev->netdev_dev; + *netdevp = &netdev->up; return 0; } @@ -597,16 +634,15 @@ netdev_linux_create(const struct netdev_class *class, const char *name, * be unavailable to other reads for tap devices. */ static int netdev_linux_create_tap(const struct netdev_class *class OVS_UNUSED, - const char *name, struct netdev_dev **netdev_devp) + const char *name, struct netdev **netdevp) { - struct netdev_dev_linux *netdev_dev; - struct tap_state *state; + struct netdev_linux *netdev; static const char tap_dev[] = "/dev/net/tun"; struct ifreq ifr; int error; - netdev_dev = xzalloc(sizeof *netdev_dev); - state = &netdev_dev->state.tap; + netdev = xzalloc(sizeof *netdev); + netdev->change_seq = 1; error = cache_notifier_ref(); if (error) { @@ -614,174 +650,133 @@ netdev_linux_create_tap(const struct netdev_class *class OVS_UNUSED, } /* Open tap device. */ - state->fd = open(tap_dev, O_RDWR); - if (state->fd < 0) { + netdev->tap_fd = open(tap_dev, O_RDWR); + if (netdev->tap_fd < 0) { error = errno; - VLOG_WARN("opening \"%s\" failed: %s", tap_dev, strerror(error)); + VLOG_WARN("opening \"%s\" failed: %s", tap_dev, ovs_strerror(error)); goto error_unref_notifier; } /* Create tap device. */ ifr.ifr_flags = IFF_TAP | IFF_NO_PI; ovs_strzcpy(ifr.ifr_name, name, sizeof ifr.ifr_name); - if (ioctl(state->fd, TUNSETIFF, &ifr) == -1) { + if (ioctl(netdev->tap_fd, TUNSETIFF, &ifr) == -1) { VLOG_WARN("%s: creating tap device failed: %s", name, - strerror(errno)); + ovs_strerror(errno)); error = errno; - goto error_unref_notifier; + goto error_close; } /* Make non-blocking. */ - error = set_nonblocking(state->fd); + error = set_nonblocking(netdev->tap_fd); if (error) { - goto error_unref_notifier; + goto error_close; } - netdev_dev_init(&netdev_dev->netdev_dev, name, &netdev_tap_class); - *netdev_devp = &netdev_dev->netdev_dev; + netdev_init(&netdev->up, name, &netdev_tap_class); + *netdevp = &netdev->up; return 0; +error_close: + close(netdev->tap_fd); error_unref_notifier: cache_notifier_unref(); error: - free(netdev_dev); + free(netdev); return error; } static void -destroy_tap(struct netdev_dev_linux *netdev_dev) -{ - struct tap_state *state = &netdev_dev->state.tap; - - if (state->fd >= 0) { - close(state->fd); - } -} - -/* Destroys the netdev device 'netdev_dev_'. */ -static void -netdev_linux_destroy(struct netdev_dev *netdev_dev_) +netdev_linux_destroy(struct netdev *netdev_) { - struct netdev_dev_linux *netdev_dev = netdev_dev_linux_cast(netdev_dev_); - const struct netdev_class *class = netdev_dev_get_class(netdev_dev_); + struct netdev_linux *netdev = netdev_linux_cast(netdev_); - if (netdev_dev->tc && netdev_dev->tc->ops->tc_destroy) { - netdev_dev->tc->ops->tc_destroy(netdev_dev->tc); + if (netdev->tc && netdev->tc->ops->tc_destroy) { + netdev->tc->ops->tc_destroy(netdev->tc); } - if (class == &netdev_tap_class) { - destroy_tap(netdev_dev); + if (netdev_get_class(netdev_) == &netdev_tap_class + && netdev->tap_fd >= 0) + { + close(netdev->tap_fd); } - free(netdev_dev); + free(netdev); cache_notifier_unref(); } static int -netdev_linux_open(struct netdev_dev *netdev_dev_, struct netdev **netdevp) +netdev_linux_rx_open(struct netdev *netdev_, struct netdev_rx **rxp) { - struct netdev_dev_linux *netdev_dev = netdev_dev_linux_cast(netdev_dev_); - struct netdev_linux *netdev; - enum netdev_flags flags; + struct netdev_linux *netdev = netdev_linux_cast(netdev_); + bool is_tap = is_tap_netdev(netdev_); + struct netdev_rx_linux *rx; int error; + int fd; - /* Allocate network device. */ - netdev = xzalloc(sizeof *netdev); - netdev->fd = -1; - netdev_init(&netdev->netdev, netdev_dev_); + if (is_tap) { + fd = netdev->tap_fd; + } else { + struct sockaddr_ll sll; + int ifindex; + /* Result of tcpdump -dd inbound */ + static struct sock_filter filt[] = { + { 0x28, 0, 0, 0xfffff004 }, /* ldh [0] */ + { 0x15, 0, 1, 0x00000004 }, /* jeq #4 jt 2 jf 3 */ + { 0x6, 0, 0, 0x00000000 }, /* ret #0 */ + { 0x6, 0, 0, 0x0000ffff } /* ret #65535 */ + }; + static struct sock_fprog fprog = { ARRAY_SIZE(filt), filt }; - /* Verify that the device really exists, by attempting to read its flags. - * (The flags might be cached, in which case this won't actually do an - * ioctl.) - * - * Don't do this for "internal" netdevs, though, because those have to be - * created as netdev objects before they exist in the kernel, because - * creating them in the kernel happens by passing a netdev object to - * dpif_port_add(). */ - if (netdev_dev_get_class(netdev_dev_) != &netdev_internal_class) { - error = netdev_get_flags(&netdev->netdev, &flags); - if (error == ENODEV) { + /* Create file descriptor. */ + fd = socket(PF_PACKET, SOCK_RAW, 0); + if (fd < 0) { + error = errno; + VLOG_ERR("failed to create raw socket (%s)", ovs_strerror(error)); goto error; } - } - - if (!strcmp(netdev_dev_get_type(netdev_dev_), "tap") && - !netdev_dev->state.tap.opened) { - - /* We assume that the first user of the tap device is the primary user - * and give them the tap FD. Subsequent users probably just expect - * this to be a system device so open it normally to avoid send/receive - * directions appearing to be reversed. */ - netdev->fd = netdev_dev->state.tap.fd; - netdev_dev->state.tap.opened = true; - } - - *netdevp = &netdev->netdev; - return 0; - -error: - netdev_uninit(&netdev->netdev, true); - return error; -} - -/* Closes and destroys 'netdev'. */ -static void -netdev_linux_close(struct netdev *netdev_) -{ - struct netdev_linux *netdev = netdev_linux_cast(netdev_); - - if (netdev->fd > 0 && strcmp(netdev_get_type(netdev_), "tap")) { - close(netdev->fd); - } - free(netdev); -} - -static int -netdev_linux_listen(struct netdev *netdev_) -{ - struct netdev_linux *netdev = netdev_linux_cast(netdev_); - struct sockaddr_ll sll; - int ifindex; - int error; - int fd; - if (netdev->fd >= 0) { - return 0; - } + /* Set non-blocking mode. */ + error = set_nonblocking(fd); + if (error) { + goto error; + } - /* 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; - } + /* Get ethernet device index. */ + error = get_ifindex(&netdev->up, &ifindex); + if (error) { + goto error; + } - /* Set non-blocking mode. */ - error = set_nonblocking(fd); - 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_), ovs_strerror(error)); + goto error; + } - /* Get ethernet device index. */ - error = get_ifindex(&netdev->netdev, &ifindex); - if (error) { - goto error; + /* Filter for only inbound packets. */ + error = setsockopt(fd, SOL_SOCKET, SO_ATTACH_FILTER, &fprog, + sizeof fprog); + if (error) { + error = errno; + VLOG_ERR("%s: failed attach filter (%s)", + netdev_get_name(netdev_), ovs_strerror(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; - } + rx = xmalloc(sizeof *rx); + netdev_rx_init(&rx->up, netdev_, &netdev_rx_linux_class); + rx->is_tap = is_tap; + rx->fd = fd; - netdev->fd = fd; + *rxp = &rx->up; return 0; error: @@ -791,62 +786,62 @@ error: return error; } -static int -netdev_linux_recv(struct netdev *netdev_, void *data, size_t size) +static void +netdev_rx_linux_destroy(struct netdev_rx *rx_) { - struct netdev_linux *netdev = netdev_linux_cast(netdev_); + struct netdev_rx_linux *rx = netdev_rx_linux_cast(rx_); - if (netdev->fd < 0) { - /* Device is not listening. */ - return -EAGAIN; + if (!rx->is_tap) { + close(rx->fd); } + free(rx); +} - for (;;) { - 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) { - VLOG_WARN_RL(&rl, "error receiving Ethernet packet on %s: %s", - strerror(errno), netdev_get_name(netdev_)); - } - return -errno; +static int +netdev_rx_linux_recv(struct netdev_rx *rx_, void *data, size_t size) +{ + struct netdev_rx_linux *rx = netdev_rx_linux_cast(rx_); + ssize_t retval; + + do { + retval = (rx->is_tap + ? read(rx->fd, data, size) + : recv(rx->fd, data, size, MSG_TRUNC)); + } while (retval < 0 && errno == EINTR); + + if (retval >= 0) { + return retval > size ? -EMSGSIZE : retval; + } else { + if (errno != EAGAIN) { + VLOG_WARN_RL(&rl, "error receiving Ethernet packet on %s: %s", + ovs_strerror(errno), netdev_rx_get_name(rx_)); } + return -errno; } } -/* Registers with the poll loop to wake up from the next call to poll_block() - * when a packet is ready to be received with netdev_recv() on 'netdev'. */ static void -netdev_linux_recv_wait(struct netdev *netdev_) +netdev_rx_linux_wait(struct netdev_rx *rx_) { - struct netdev_linux *netdev = netdev_linux_cast(netdev_); - if (netdev->fd >= 0) { - poll_fd_wait(netdev->fd, POLLIN); - } + struct netdev_rx_linux *rx = netdev_rx_linux_cast(rx_); + poll_fd_wait(rx->fd, POLLIN); } -/* Discards all packets waiting to be received from 'netdev'. */ static int -netdev_linux_drain(struct netdev *netdev_) +netdev_rx_linux_drain(struct netdev_rx *rx_) { - struct netdev_linux *netdev = netdev_linux_cast(netdev_); - if (netdev->fd < 0) { - return 0; - } else if (!strcmp(netdev_get_type(netdev_), "tap")) { + struct netdev_rx_linux *rx = netdev_rx_linux_cast(rx_); + if (rx->is_tap) { struct ifreq ifr; - int error = netdev_linux_do_ioctl(netdev_get_name(netdev_), &ifr, + int error = netdev_linux_do_ioctl(netdev_rx_get_name(rx_), &ifr, SIOCGIFTXQLEN, "SIOCGIFTXQLEN"); if (error) { return error; } - drain_fd(netdev->fd, ifr.ifr_qlen); + drain_fd(rx->fd, ifr.ifr_qlen); return 0; } else { - return drain_rcvbuf(netdev->fd); + return drain_rcvbuf(rx->fd); } } @@ -862,11 +857,10 @@ netdev_linux_drain(struct netdev *netdev_) static int netdev_linux_send(struct netdev *netdev_, const void *data, size_t size) { - struct netdev_linux *netdev = netdev_linux_cast(netdev_); for (;;) { ssize_t retval; - if (netdev->fd < 0) { + if (!is_tap_netdev(netdev_)) { /* Use our AF_PACKET socket to send to this device. */ struct sockaddr_ll sll; struct msghdr msg; @@ -877,7 +871,7 @@ netdev_linux_send(struct netdev *netdev_, const void *data, size_t size) sock = af_packet_sock(); if (sock < 0) { - return sock; + return -sock; } error = get_ifindex(netdev_, &ifindex); @@ -891,7 +885,7 @@ netdev_linux_send(struct netdev *netdev_, const void *data, size_t size) sll.sll_family = AF_PACKET; sll.sll_ifindex = ifindex; - iov.iov_base = (void *) data; + iov.iov_base = CONST_CAST(void *, data); iov.iov_len = size; msg.msg_name = &sll; @@ -904,11 +898,14 @@ netdev_linux_send(struct netdev *netdev_, const void *data, size_t size) retval = sendmsg(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); + /* Use the tap 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. This doesn't occur on other interface types + * because we attach a socket filter to the rx socket. */ + struct netdev_linux *netdev = netdev_linux_cast(netdev_); + + retval = write(netdev->tap_fd, data, size); } if (retval < 0) { @@ -921,7 +918,7 @@ netdev_linux_send(struct netdev *netdev_, const void *data, size_t size) continue; } else if (errno != EAGAIN) { VLOG_WARN_RL(&rl, "error sending Ethernet packet on %s: %s", - netdev_get_name(netdev_), strerror(errno)); + netdev_get_name(netdev_), ovs_strerror(errno)); } return errno; } else if (retval != size) { @@ -942,14 +939,9 @@ netdev_linux_send(struct netdev *netdev_, const void *data, size_t size) * 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) { - struct netdev_linux *netdev = netdev_linux_cast(netdev_); - if (netdev->fd < 0) { - /* Nothing to do. */ - } else if (strcmp(netdev_get_type(netdev_), "tap")) { - poll_fd_wait(netdev->fd, POLLOUT); - } else { + if (is_tap_netdev(netdev)) { /* TAP device always accepts packets.*/ poll_immediate_wake(); } @@ -961,41 +953,58 @@ static int netdev_linux_set_etheraddr(struct netdev *netdev_, const uint8_t mac[ETH_ADDR_LEN]) { - struct netdev_dev_linux *netdev_dev = - netdev_dev_linux_cast(netdev_get_dev(netdev_)); + struct netdev_linux *netdev = netdev_linux_cast(netdev_); + struct netdev_saved_flags *sf = NULL; int error; - if (!(netdev_dev->cache_valid & VALID_ETHERADDR) - || !eth_addr_equals(netdev_dev->etheraddr, mac)) { - error = set_etheraddr(netdev_get_name(netdev_), ARPHRD_ETHER, mac); + if (netdev->cache_valid & VALID_ETHERADDR) { + if (netdev->ether_addr_error) { + return netdev->ether_addr_error; + } + if (eth_addr_equals(netdev->etheraddr, mac)) { + return 0; + } + netdev->cache_valid &= ~VALID_ETHERADDR; + } + + /* Tap devices must be brought down before setting the address. */ + if (is_tap_netdev(netdev_)) { + netdev_turn_flags_off(netdev_, NETDEV_UP, &sf); + } + error = set_etheraddr(netdev_get_name(netdev_), mac); + if (!error || error == ENODEV) { + netdev->ether_addr_error = error; + netdev->cache_valid |= VALID_ETHERADDR; if (!error) { - netdev_dev->cache_valid |= VALID_ETHERADDR; - memcpy(netdev_dev->etheraddr, mac, ETH_ADDR_LEN); + memcpy(netdev->etheraddr, mac, ETH_ADDR_LEN); } - } else { - error = 0; } + + netdev_restore_flags(sf); + return error; } -/* Returns a pointer to 'netdev''s MAC address. The caller must not modify or - * free the returned buffer. */ +/* Copies 'netdev''s MAC address to 'mac' which is passed as param. */ static int netdev_linux_get_etheraddr(const struct netdev *netdev_, uint8_t mac[ETH_ADDR_LEN]) { - struct netdev_dev_linux *netdev_dev = - netdev_dev_linux_cast(netdev_get_dev(netdev_)); - if (!(netdev_dev->cache_valid & VALID_ETHERADDR)) { + struct netdev_linux *netdev = netdev_linux_cast(netdev_); + + if (!(netdev->cache_valid & VALID_ETHERADDR)) { int error = get_etheraddr(netdev_get_name(netdev_), - netdev_dev->etheraddr); - if (error) { - return error; - } - netdev_dev->cache_valid |= VALID_ETHERADDR; + netdev->etheraddr); + + netdev->ether_addr_error = error; + netdev->cache_valid |= VALID_ETHERADDR; } - memcpy(mac, netdev_dev->etheraddr, ETH_ADDR_LEN); - return 0; + + if (!netdev->ether_addr_error) { + memcpy(mac, netdev->etheraddr, ETH_ADDR_LEN); + } + + return netdev->ether_addr_error; } /* Returns the maximum size of transmitted (and received) packets on 'netdev', @@ -1004,22 +1013,23 @@ netdev_linux_get_etheraddr(const struct netdev *netdev_, static int netdev_linux_get_mtu(const struct netdev *netdev_, int *mtup) { - struct netdev_dev_linux *netdev_dev = - netdev_dev_linux_cast(netdev_get_dev(netdev_)); - if (!(netdev_dev->cache_valid & VALID_MTU)) { + struct netdev_linux *netdev = netdev_linux_cast(netdev_); + if (!(netdev->cache_valid & VALID_MTU)) { struct ifreq ifr; int error; error = netdev_linux_do_ioctl(netdev_get_name(netdev_), &ifr, SIOCGIFMTU, "SIOCGIFMTU"); - if (error) { - return error; - } - netdev_dev->mtu = ifr.ifr_mtu; - netdev_dev->cache_valid |= VALID_MTU; + + netdev->netdev_mtu_error = error; + netdev->mtu = ifr.ifr_mtu; + netdev->cache_valid |= VALID_MTU; } - *mtup = netdev_dev->mtu; - return 0; + + if (!netdev->netdev_mtu_error) { + *mtup = netdev->mtu; + } + return netdev->netdev_mtu_error; } /* Sets the maximum size of transmitted (MTU) for given device using linux @@ -1028,21 +1038,28 @@ netdev_linux_get_mtu(const struct netdev *netdev_, int *mtup) 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 netdev_linux *netdev = netdev_linux_cast(netdev_); struct ifreq ifr; int error; + if (netdev->cache_valid & VALID_MTU) { + if (netdev->netdev_mtu_error) { + return netdev->netdev_mtu_error; + } + if (netdev->mtu == mtu) { + return 0; + } + netdev->cache_valid &= ~VALID_MTU; + } ifr.ifr_mtu = mtu; error = netdev_linux_do_ioctl(netdev_get_name(netdev_), &ifr, SIOCSIFMTU, "SIOCSIFMTU"); - if (error) { - return error; + if (!error || error == ENODEV) { + netdev->netdev_mtu_error = error; + netdev->mtu = ifr.ifr_mtu; + netdev->cache_valid |= VALID_MTU; } - - netdev_dev->mtu = ifr.ifr_mtu; - netdev_dev->cache_valid |= VALID_MTU; - return 0; + return error; } /* Returns the ifindex of 'netdev', if successful, as a positive number. @@ -1059,13 +1076,12 @@ netdev_linux_get_ifindex(const struct netdev *netdev) static int netdev_linux_get_carrier(const struct netdev *netdev_, bool *carrier) { - struct netdev_dev_linux *netdev_dev = - netdev_dev_linux_cast(netdev_get_dev(netdev_)); + struct netdev_linux *netdev = netdev_linux_cast(netdev_); - if (netdev_dev->miimon_interval > 0) { - *carrier = netdev_dev->miimon; + if (netdev->miimon_interval > 0) { + *carrier = netdev->miimon; } else { - *carrier = netdev_dev->carrier; + *carrier = (netdev->ifi_flags & IFF_RUNNING) != 0; } return 0; @@ -1074,7 +1090,7 @@ netdev_linux_get_carrier(const struct netdev *netdev_, bool *carrier) static long long int netdev_linux_get_carrier_resets(const struct netdev *netdev) { - return netdev_dev_linux_cast(netdev_get_dev(netdev))->carrier_resets; + return netdev_linux_cast(netdev)->carrier_resets; } static int @@ -1119,6 +1135,7 @@ netdev_linux_get_miimon(const char *name, bool *miimon) VLOG_DBG_RL(&rl, "%s: failed to query MII, falling back to ethtool", name); + COVERAGE_INC(netdev_get_ethtool); memset(&ecmd, 0, sizeof ecmd); error = netdev_linux_do_ethtool(name, &ecmd, ETHTOOL_GLINK, "ETHTOOL_GLINK"); @@ -1139,14 +1156,12 @@ static int netdev_linux_set_miimon_interval(struct netdev *netdev_, long long int interval) { - struct netdev_dev_linux *netdev_dev; - - netdev_dev = netdev_dev_linux_cast(netdev_get_dev(netdev_)); + struct netdev_linux *netdev = netdev_linux_cast(netdev_); interval = interval > 0 ? MAX(interval, 100) : 0; - if (netdev_dev->miimon_interval != interval) { - netdev_dev->miimon_interval = interval; - timer_set_expired(&netdev_dev->miimon_timer); + if (netdev->miimon_interval != interval) { + netdev->miimon_interval = interval; + timer_set_expired(&netdev->miimon_timer); } return 0; @@ -1159,22 +1174,25 @@ netdev_linux_miimon_run(void) struct shash_node *node; shash_init(&device_shash); - netdev_dev_get_devices(&netdev_linux_class, &device_shash); + netdev_get_devices(&netdev_linux_class, &device_shash); SHASH_FOR_EACH (node, &device_shash) { - struct netdev_dev_linux *dev = node->data; + struct netdev *netdev = node->data; + struct netdev_linux *dev = netdev_linux_cast(netdev); bool miimon; if (dev->miimon_interval <= 0 || !timer_expired(&dev->miimon_timer)) { + netdev_close(netdev); continue; } - netdev_linux_get_miimon(dev->netdev_dev.name, &miimon); + netdev_linux_get_miimon(dev->up.name, &miimon); if (miimon != dev->miimon) { dev->miimon = miimon; - netdev_dev_linux_changed(dev); + netdev_linux_changed(dev, dev->ifi_flags, 0); } timer_set_duration(&dev->miimon_timer, dev->miimon_interval); + netdev_close(netdev); } shash_destroy(&device_shash); @@ -1187,13 +1205,15 @@ netdev_linux_miimon_wait(void) struct shash_node *node; shash_init(&device_shash); - netdev_dev_get_devices(&netdev_linux_class, &device_shash); + netdev_get_devices(&netdev_linux_class, &device_shash); SHASH_FOR_EACH (node, &device_shash) { - struct netdev_dev_linux *dev = node->data; + struct netdev *netdev = node->data; + struct netdev_linux *dev = netdev_linux_cast(netdev); if (dev->miimon_interval > 0) { timer_wait(&dev->miimon_timer); } + netdev_close(netdev); } shash_destroy(&device_shash); } @@ -1220,7 +1240,7 @@ check_for_working_netlink_stats(void) } else { VLOG_INFO("RTM_GETLINK failed (%s), obtaining netdev stats " "via proc (you are probably running a pre-2.6.19 " - "kernel)", strerror(error)); + "kernel)", ovs_strerror(error)); return false; } } @@ -1234,24 +1254,76 @@ swap_uint64(uint64_t *a, uint64_t *b) *b = tmp; } +/* Copies 'src' into 'dst', performing format conversion in the process. + * + * 'src' is allowed to be misaligned. */ +static void +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->multicast = 0; + dst->collisions = 0; + dst->rx_length_errors = 0; + dst->rx_over_errors = 0; + dst->rx_crc_errors = 0; + dst->rx_frame_errors = 0; + dst->rx_fifo_errors = 0; + dst->rx_missed_errors = 0; + dst->tx_aborted_errors = 0; + dst->tx_carrier_errors = 0; + dst->tx_fifo_errors = 0; + dst->tx_heartbeat_errors = 0; + dst->tx_window_errors = 0; +} + +static int +get_stats_via_vport__(const struct netdev *netdev, struct netdev_stats *stats) +{ + struct dpif_linux_vport reply; + struct ofpbuf *buf; + int error; + + error = dpif_linux_vport_get(netdev_get_name(netdev), &reply, &buf); + if (error) { + return error; + } else if (!reply.stats) { + ofpbuf_delete(buf); + return EOPNOTSUPP; + } + + netdev_stats_from_ovs_vport_stats(stats, reply.stats); + + ofpbuf_delete(buf); + + return 0; +} + 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_)); + struct netdev_linux *netdev = netdev_linux_cast(netdev_); - if (netdev_dev->have_vport_stats || - !(netdev_dev->cache_valid & VALID_HAVE_VPORT_STATS)) { + if (!netdev->vport_stats_error || + !(netdev->cache_valid & VALID_VPORT_STAT_ERROR)) { int error; - error = netdev_vport_get_stats(netdev_, stats); - if (error) { + error = get_stats_via_vport__(netdev_, stats); + if (error && error != ENOENT) { VLOG_WARN_RL(&rl, "%s: obtaining netdev stats via vport failed " - "(%s)", netdev_get_name(netdev_), strerror(error)); + "(%s)", + netdev_get_name(netdev_), ovs_strerror(error)); } - netdev_dev->have_vport_stats = !error; - netdev_dev->cache_valid |= VALID_HAVE_VPORT_STATS; + netdev->vport_stats_error = error; + netdev->cache_valid |= VALID_VPORT_STAT_ERROR; } } @@ -1259,11 +1331,13 @@ static int netdev_linux_sys_get_stats(const struct netdev *netdev_, struct netdev_stats *stats) { - static int use_netlink_stats = -1; + static struct ovsthread_once once = OVSTHREAD_ONCE_INITIALIZER; + static int use_netlink_stats; int error; - if (use_netlink_stats < 0) { + if (ovsthread_once_start(&once)) { use_netlink_stats = check_for_working_netlink_stats(); + ovsthread_once_done(&once); } if (use_netlink_stats) { @@ -1290,8 +1364,7 @@ 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_linux *netdev = netdev_linux_cast(netdev_); struct netdev_stats dev_stats; int error; @@ -1300,14 +1373,14 @@ netdev_linux_get_stats(const struct netdev *netdev_, error = netdev_linux_sys_get_stats(netdev_, &dev_stats); if (error) { - if (!netdev_dev->have_vport_stats) { + if (netdev->vport_stats_error) { return error; } else { return 0; } } - if (!netdev_dev->have_vport_stats) { + if (netdev->vport_stats_error) { /* stats not available from OVS then use ioctl stats. */ *stats = dev_stats; } else { @@ -1335,11 +1408,9 @@ netdev_linux_get_stats(const struct netdev *netdev_, /* 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) +netdev_tap_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_linux *netdev = netdev_linux_cast(netdev_); struct netdev_stats dev_stats; int error; @@ -1347,7 +1418,7 @@ netdev_pseudo_get_stats(const struct netdev *netdev_, error = netdev_linux_sys_get_stats(netdev_, &dev_stats); if (error) { - if (!netdev_dev->have_vport_stats) { + if (netdev->vport_stats_error) { return error; } else { return 0; @@ -1360,7 +1431,7 @@ netdev_pseudo_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. */ - if (!netdev_dev->have_vport_stats) { + if (netdev->vport_stats_error) { *stats = dev_stats; swap_uint64(&stats->rx_packets, &stats->tx_packets); swap_uint64(&stats->rx_bytes, &stats->tx_bytes); @@ -1390,139 +1461,215 @@ netdev_pseudo_get_stats(const struct netdev *netdev_, return 0; } -/* Stores the features supported by 'netdev' into each of '*current', - * '*advertised', '*supported', and '*peer' that are non-null. Each value is a - * 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(const struct netdev *netdev, - uint32_t *current, uint32_t *advertised, - uint32_t *supported, uint32_t *peer) +netdev_internal_get_stats(const struct netdev *netdev_, + struct netdev_stats *stats) +{ + struct netdev_linux *netdev = netdev_linux_cast(netdev_); + + get_stats_via_vport(netdev_, stats); + return netdev->vport_stats_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) { struct ethtool_cmd ecmd; + uint32_t speed; int error; + if (netdev->cache_valid & VALID_FEATURES) { + return; + } + + COVERAGE_INC(netdev_get_ethtool); memset(&ecmd, 0, sizeof ecmd); - error = netdev_linux_do_ethtool(netdev_get_name(netdev), &ecmd, + error = netdev_linux_do_ethtool(netdev->up.name, &ecmd, ETHTOOL_GSET, "ETHTOOL_GSET"); if (error) { - return error; + goto out; } /* Supported features. */ - *supported = 0; + netdev->supported = 0; if (ecmd.supported & SUPPORTED_10baseT_Half) { - *supported |= OFPPF_10MB_HD; + netdev->supported |= NETDEV_F_10MB_HD; } if (ecmd.supported & SUPPORTED_10baseT_Full) { - *supported |= OFPPF_10MB_FD; + netdev->supported |= NETDEV_F_10MB_FD; } if (ecmd.supported & SUPPORTED_100baseT_Half) { - *supported |= OFPPF_100MB_HD; + netdev->supported |= NETDEV_F_100MB_HD; } if (ecmd.supported & SUPPORTED_100baseT_Full) { - *supported |= OFPPF_100MB_FD; + netdev->supported |= NETDEV_F_100MB_FD; } if (ecmd.supported & SUPPORTED_1000baseT_Half) { - *supported |= OFPPF_1GB_HD; + netdev->supported |= NETDEV_F_1GB_HD; } if (ecmd.supported & SUPPORTED_1000baseT_Full) { - *supported |= OFPPF_1GB_FD; + netdev->supported |= NETDEV_F_1GB_FD; } if (ecmd.supported & SUPPORTED_10000baseT_Full) { - *supported |= OFPPF_10GB_FD; + netdev->supported |= NETDEV_F_10GB_FD; } if (ecmd.supported & SUPPORTED_TP) { - *supported |= OFPPF_COPPER; + netdev->supported |= NETDEV_F_COPPER; } if (ecmd.supported & SUPPORTED_FIBRE) { - *supported |= OFPPF_FIBER; + netdev->supported |= NETDEV_F_FIBER; } if (ecmd.supported & SUPPORTED_Autoneg) { - *supported |= OFPPF_AUTONEG; + netdev->supported |= NETDEV_F_AUTONEG; } if (ecmd.supported & SUPPORTED_Pause) { - *supported |= OFPPF_PAUSE; + netdev->supported |= NETDEV_F_PAUSE; } if (ecmd.supported & SUPPORTED_Asym_Pause) { - *supported |= OFPPF_PAUSE_ASYM; + netdev->supported |= NETDEV_F_PAUSE_ASYM; } /* Advertised features. */ - *advertised = 0; + netdev->advertised = 0; if (ecmd.advertising & ADVERTISED_10baseT_Half) { - *advertised |= OFPPF_10MB_HD; + netdev->advertised |= NETDEV_F_10MB_HD; } if (ecmd.advertising & ADVERTISED_10baseT_Full) { - *advertised |= OFPPF_10MB_FD; + netdev->advertised |= NETDEV_F_10MB_FD; } if (ecmd.advertising & ADVERTISED_100baseT_Half) { - *advertised |= OFPPF_100MB_HD; + netdev->advertised |= NETDEV_F_100MB_HD; } if (ecmd.advertising & ADVERTISED_100baseT_Full) { - *advertised |= OFPPF_100MB_FD; + netdev->advertised |= NETDEV_F_100MB_FD; } if (ecmd.advertising & ADVERTISED_1000baseT_Half) { - *advertised |= OFPPF_1GB_HD; + netdev->advertised |= NETDEV_F_1GB_HD; } if (ecmd.advertising & ADVERTISED_1000baseT_Full) { - *advertised |= OFPPF_1GB_FD; + netdev->advertised |= NETDEV_F_1GB_FD; } if (ecmd.advertising & ADVERTISED_10000baseT_Full) { - *advertised |= OFPPF_10GB_FD; + netdev->advertised |= NETDEV_F_10GB_FD; } if (ecmd.advertising & ADVERTISED_TP) { - *advertised |= OFPPF_COPPER; + netdev->advertised |= NETDEV_F_COPPER; } if (ecmd.advertising & ADVERTISED_FIBRE) { - *advertised |= OFPPF_FIBER; + netdev->advertised |= NETDEV_F_FIBER; } if (ecmd.advertising & ADVERTISED_Autoneg) { - *advertised |= OFPPF_AUTONEG; + netdev->advertised |= NETDEV_F_AUTONEG; } if (ecmd.advertising & ADVERTISED_Pause) { - *advertised |= OFPPF_PAUSE; + netdev->advertised |= NETDEV_F_PAUSE; } if (ecmd.advertising & ADVERTISED_Asym_Pause) { - *advertised |= OFPPF_PAUSE_ASYM; + netdev->advertised |= NETDEV_F_PAUSE_ASYM; } /* Current settings. */ - if (ecmd.speed == SPEED_10) { - *current = ecmd.duplex ? OFPPF_10MB_FD : OFPPF_10MB_HD; - } else if (ecmd.speed == SPEED_100) { - *current = ecmd.duplex ? OFPPF_100MB_FD : OFPPF_100MB_HD; - } else if (ecmd.speed == SPEED_1000) { - *current = ecmd.duplex ? OFPPF_1GB_FD : OFPPF_1GB_HD; - } else if (ecmd.speed == SPEED_10000) { - *current = OFPPF_10GB_FD; + speed = ecmd.speed; + if (speed == SPEED_10) { + netdev->current = ecmd.duplex ? NETDEV_F_10MB_FD : NETDEV_F_10MB_HD; + } else if (speed == SPEED_100) { + netdev->current = ecmd.duplex ? NETDEV_F_100MB_FD : NETDEV_F_100MB_HD; + } else if (speed == SPEED_1000) { + netdev->current = ecmd.duplex ? NETDEV_F_1GB_FD : NETDEV_F_1GB_HD; + } else if (speed == SPEED_10000) { + netdev->current = NETDEV_F_10GB_FD; + } else if (speed == 40000) { + netdev->current = NETDEV_F_40GB_FD; + } else if (speed == 100000) { + netdev->current = NETDEV_F_100GB_FD; + } else if (speed == 1000000) { + netdev->current = NETDEV_F_1TB_FD; } else { - *current = 0; + netdev->current = 0; } if (ecmd.port == PORT_TP) { - *current |= OFPPF_COPPER; + netdev->current |= NETDEV_F_COPPER; } else if (ecmd.port == PORT_FIBRE) { - *current |= OFPPF_FIBER; + netdev->current |= NETDEV_F_FIBER; } if (ecmd.autoneg) { - *current |= OFPPF_AUTONEG; + netdev->current |= NETDEV_F_AUTONEG; } - /* Peer advertisements. */ - *peer = 0; /* XXX */ +out: + netdev->cache_valid |= VALID_FEATURES; + netdev->get_features_error = error; +} - return 0; +/* Stores the features supported by 'netdev' into of '*current', '*advertised', + * '*supported', and '*peer'. Each value is a bitmap of NETDEV_* bits. + * Returns 0 if successful, otherwise a positive errno value. */ +static int +netdev_linux_get_features(const struct netdev *netdev_, + enum netdev_features *current, + enum netdev_features *advertised, + enum netdev_features *supported, + enum netdev_features *peer) +{ + struct netdev_linux *netdev = netdev_linux_cast(netdev_); + + netdev_linux_read_features(netdev); + + if (!netdev->get_features_error) { + *current = netdev->current; + *advertised = netdev->advertised; + *supported = netdev->supported; + *peer = 0; /* XXX */ + } + return netdev->get_features_error; } /* Set the features advertised by 'netdev' to 'advertise'. */ static int -netdev_linux_set_advertisements(struct netdev *netdev, uint32_t advertise) +netdev_linux_set_advertisements(struct netdev *netdev, + enum netdev_features advertise) { struct ethtool_cmd ecmd; int error; + COVERAGE_INC(netdev_get_ethtool); memset(&ecmd, 0, sizeof ecmd); error = netdev_linux_do_ethtool(netdev_get_name(netdev), &ecmd, ETHTOOL_GSET, "ETHTOOL_GSET"); @@ -1531,42 +1678,43 @@ netdev_linux_set_advertisements(struct netdev *netdev, uint32_t advertise) } ecmd.advertising = 0; - if (advertise & OFPPF_10MB_HD) { + if (advertise & NETDEV_F_10MB_HD) { ecmd.advertising |= ADVERTISED_10baseT_Half; } - if (advertise & OFPPF_10MB_FD) { + if (advertise & NETDEV_F_10MB_FD) { ecmd.advertising |= ADVERTISED_10baseT_Full; } - if (advertise & OFPPF_100MB_HD) { + if (advertise & NETDEV_F_100MB_HD) { ecmd.advertising |= ADVERTISED_100baseT_Half; } - if (advertise & OFPPF_100MB_FD) { + if (advertise & NETDEV_F_100MB_FD) { ecmd.advertising |= ADVERTISED_100baseT_Full; } - if (advertise & OFPPF_1GB_HD) { + if (advertise & NETDEV_F_1GB_HD) { ecmd.advertising |= ADVERTISED_1000baseT_Half; } - if (advertise & OFPPF_1GB_FD) { + if (advertise & NETDEV_F_1GB_FD) { ecmd.advertising |= ADVERTISED_1000baseT_Full; } - if (advertise & OFPPF_10GB_FD) { + if (advertise & NETDEV_F_10GB_FD) { ecmd.advertising |= ADVERTISED_10000baseT_Full; } - if (advertise & OFPPF_COPPER) { + if (advertise & NETDEV_F_COPPER) { ecmd.advertising |= ADVERTISED_TP; } - if (advertise & OFPPF_FIBER) { + if (advertise & NETDEV_F_FIBER) { ecmd.advertising |= ADVERTISED_FIBRE; } - if (advertise & OFPPF_AUTONEG) { + if (advertise & NETDEV_F_AUTONEG) { ecmd.advertising |= ADVERTISED_Autoneg; } - if (advertise & OFPPF_PAUSE) { + if (advertise & NETDEV_F_PAUSE) { ecmd.advertising |= ADVERTISED_Pause; } - if (advertise & OFPPF_PAUSE_ASYM) { + if (advertise & NETDEV_F_PAUSE_ASYM) { ecmd.advertising |= ADVERTISED_Asym_Pause; } + COVERAGE_INC(netdev_set_ethtool); return netdev_linux_do_ethtool(netdev_get_name(netdev), &ecmd, ETHTOOL_SSET, "ETHTOOL_SSET"); } @@ -1574,63 +1722,72 @@ netdev_linux_set_advertisements(struct netdev *netdev, uint32_t advertise) /* 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, +netdev_linux_set_policing(struct netdev *netdev_, uint32_t kbits_rate, uint32_t kbits_burst) { - struct netdev_dev_linux *netdev_dev = - netdev_dev_linux_cast(netdev_get_dev(netdev)); - const char *netdev_name = netdev_get_name(netdev); + struct netdev_linux *netdev = netdev_linux_cast(netdev_); + const char *netdev_name = netdev_get_name(netdev_); int error; - COVERAGE_INC(netdev_set_policing); kbits_burst = (!kbits_rate ? 0 /* Force to 0 if no rate specified. */ : !kbits_burst ? 1000 /* Default to 1000 kbits if 0. */ : kbits_burst); /* Stick with user-specified value. */ - if (netdev_dev->cache_valid & VALID_POLICING - && netdev_dev->kbits_rate == kbits_rate - && netdev_dev->kbits_burst == kbits_burst) { - /* Assume that settings haven't changed since we last set them. */ - return 0; + if (netdev->cache_valid & VALID_POLICING) { + if (netdev->netdev_policing_error) { + return netdev->netdev_policing_error; + } + + if (netdev->kbits_rate == kbits_rate && + netdev->kbits_burst == kbits_burst) { + /* Assume that settings haven't changed since we last set them. */ + return 0; + } + netdev->cache_valid &= ~VALID_POLICING; } + COVERAGE_INC(netdev_set_policing); /* Remove any existing ingress qdisc. */ - error = tc_add_del_ingress_qdisc(netdev, false); + 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; + netdev_name, ovs_strerror(error)); + goto out; } if (kbits_rate) { - error = tc_add_del_ingress_qdisc(netdev, true); + 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; + netdev_name, ovs_strerror(error)); + goto out; } - error = tc_add_policer(netdev, kbits_rate, kbits_burst); + 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_name, ovs_strerror(error)); + goto out; } } - netdev_dev->kbits_rate = kbits_rate; - netdev_dev->kbits_burst = kbits_burst; - netdev_dev->cache_valid |= VALID_POLICING; + netdev->kbits_rate = kbits_rate; + netdev->kbits_burst = kbits_burst; - return 0; +out: + if (!error || error == ENODEV) { + netdev->netdev_policing_error = error; + netdev->cache_valid |= VALID_POLICING; + } + return error; } static int netdev_linux_get_qos_types(const struct netdev *netdev OVS_UNUSED, struct sset *types) { - const struct tc_ops **opsp; + const struct tc_ops *const *opsp; for (opsp = tcs; *opsp != NULL; opsp++) { const struct tc_ops *ops = *opsp; @@ -1644,7 +1801,7 @@ netdev_linux_get_qos_types(const struct netdev *netdev OVS_UNUSED, static const struct tc_ops * tc_lookup_ovs_name(const char *name) { - const struct tc_ops **opsp; + const struct tc_ops *const *opsp; for (opsp = tcs; *opsp != NULL; opsp++) { const struct tc_ops *ops = *opsp; @@ -1658,7 +1815,7 @@ tc_lookup_ovs_name(const char *name) static const struct tc_ops * tc_lookup_linux_name(const char *name) { - const struct tc_ops **opsp; + const struct tc_ops *const *opsp; for (opsp = tcs; *opsp != NULL; opsp++) { const struct tc_ops *ops = *opsp; @@ -1670,14 +1827,13 @@ tc_lookup_linux_name(const char *name) } static struct tc_queue * -tc_find_queue__(const struct netdev *netdev, unsigned int queue_id, +tc_find_queue__(const struct netdev *netdev_, unsigned int queue_id, size_t hash) { - struct netdev_dev_linux *netdev_dev = - netdev_dev_linux_cast(netdev_get_dev(netdev)); + struct netdev_linux *netdev = netdev_linux_cast(netdev_); struct tc_queue *queue; - HMAP_FOR_EACH_IN_BUCKET (queue, hmap_node, hash, &netdev_dev->tc->queues) { + HMAP_FOR_EACH_IN_BUCKET (queue, hmap_node, hash, &netdev->tc->queues) { if (queue->queue_id == queue_id) { return queue; } @@ -1705,30 +1861,28 @@ netdev_linux_get_qos_capabilities(const struct netdev *netdev OVS_UNUSED, } static int -netdev_linux_get_qos(const struct netdev *netdev, - const char **typep, struct shash *details) +netdev_linux_get_qos(const struct netdev *netdev_, + const char **typep, struct smap *details) { - struct netdev_dev_linux *netdev_dev = - netdev_dev_linux_cast(netdev_get_dev(netdev)); + struct netdev_linux *netdev = netdev_linux_cast(netdev_); int error; - error = tc_query_qdisc(netdev); + error = tc_query_qdisc(netdev_); if (error) { return error; } - *typep = netdev_dev->tc->ops->ovs_name; - return (netdev_dev->tc->ops->qdisc_get - ? netdev_dev->tc->ops->qdisc_get(netdev, details) + *typep = netdev->tc->ops->ovs_name; + return (netdev->tc->ops->qdisc_get + ? netdev->tc->ops->qdisc_get(netdev_, details) : 0); } static int -netdev_linux_set_qos(struct netdev *netdev, - const char *type, const struct shash *details) +netdev_linux_set_qos(struct netdev *netdev_, + const char *type, const struct smap *details) { - struct netdev_dev_linux *netdev_dev = - netdev_dev_linux_cast(netdev_get_dev(netdev)); + struct netdev_linux *netdev = netdev_linux_cast(netdev_); const struct tc_ops *new_ops; int error; @@ -1737,106 +1891,104 @@ netdev_linux_set_qos(struct netdev *netdev, return EOPNOTSUPP; } - error = tc_query_qdisc(netdev); + error = tc_query_qdisc(netdev_); if (error) { return error; } - if (new_ops == netdev_dev->tc->ops) { - return new_ops->qdisc_set ? new_ops->qdisc_set(netdev, details) : 0; + if (new_ops == netdev->tc->ops) { + return new_ops->qdisc_set ? new_ops->qdisc_set(netdev_, details) : 0; } else { /* Delete existing qdisc. */ - error = tc_del_qdisc(netdev); + error = tc_del_qdisc(netdev_); if (error) { return error; } - assert(netdev_dev->tc == NULL); + ovs_assert(netdev->tc == NULL); /* Install new qdisc. */ - error = new_ops->tc_install(netdev, details); - assert((error == 0) == (netdev_dev->tc != NULL)); + error = new_ops->tc_install(netdev_, details); + ovs_assert((error == 0) == (netdev->tc != NULL)); return error; } } static int -netdev_linux_get_queue(const struct netdev *netdev, - unsigned int queue_id, struct shash *details) +netdev_linux_get_queue(const struct netdev *netdev_, + unsigned int queue_id, struct smap *details) { - struct netdev_dev_linux *netdev_dev = - netdev_dev_linux_cast(netdev_get_dev(netdev)); + struct netdev_linux *netdev = netdev_linux_cast(netdev_); int error; - error = tc_query_qdisc(netdev); + error = tc_query_qdisc(netdev_); if (error) { return error; } else { - struct tc_queue *queue = tc_find_queue(netdev, queue_id); + struct tc_queue *queue = tc_find_queue(netdev_, queue_id); return (queue - ? netdev_dev->tc->ops->class_get(netdev, queue, details) + ? netdev->tc->ops->class_get(netdev_, queue, details) : ENOENT); } } static int -netdev_linux_set_queue(struct netdev *netdev, - unsigned int queue_id, const struct shash *details) +netdev_linux_set_queue(struct netdev *netdev_, + unsigned int queue_id, const struct smap *details) { - struct netdev_dev_linux *netdev_dev = - netdev_dev_linux_cast(netdev_get_dev(netdev)); + struct netdev_linux *netdev = netdev_linux_cast(netdev_); int error; - error = tc_query_qdisc(netdev); + error = tc_query_qdisc(netdev_); if (error) { return error; - } else if (queue_id >= netdev_dev->tc->ops->n_queues - || !netdev_dev->tc->ops->class_set) { + } else if (queue_id >= netdev->tc->ops->n_queues + || !netdev->tc->ops->class_set) { return EINVAL; } - return netdev_dev->tc->ops->class_set(netdev, queue_id, details); + return netdev->tc->ops->class_set(netdev_, queue_id, details); } static int -netdev_linux_delete_queue(struct netdev *netdev, unsigned int queue_id) +netdev_linux_delete_queue(struct netdev *netdev_, unsigned int queue_id) { - struct netdev_dev_linux *netdev_dev = - netdev_dev_linux_cast(netdev_get_dev(netdev)); + struct netdev_linux *netdev = netdev_linux_cast(netdev_); int error; - error = tc_query_qdisc(netdev); + error = tc_query_qdisc(netdev_); if (error) { return error; - } else if (!netdev_dev->tc->ops->class_delete) { + } else if (!netdev->tc->ops->class_delete) { return EINVAL; } else { - struct tc_queue *queue = tc_find_queue(netdev, queue_id); + struct tc_queue *queue = tc_find_queue(netdev_, queue_id); return (queue - ? netdev_dev->tc->ops->class_delete(netdev, queue) + ? netdev->tc->ops->class_delete(netdev_, queue) : ENOENT); } } static int -netdev_linux_get_queue_stats(const struct netdev *netdev, +netdev_linux_get_queue_stats(const struct netdev *netdev_, unsigned int queue_id, struct netdev_queue_stats *stats) { - struct netdev_dev_linux *netdev_dev = - netdev_dev_linux_cast(netdev_get_dev(netdev)); + struct netdev_linux *netdev = netdev_linux_cast(netdev_); int error; - error = tc_query_qdisc(netdev); + error = tc_query_qdisc(netdev_); if (error) { return error; - } else if (!netdev_dev->tc->ops->class_get_stats) { + } else if (!netdev->tc->ops->class_get_stats) { return EOPNOTSUPP; } else { - const struct tc_queue *queue = tc_find_queue(netdev, queue_id); - return (queue - ? netdev_dev->tc->ops->class_get_stats(netdev, queue, stats) - : ENOENT); + const struct tc_queue *queue = tc_find_queue(netdev_, queue_id); + if (!queue) { + return ENOENT; + } + stats->created = queue->created; + return netdev->tc->ops->class_get_stats(netdev_, queue, stats); } } @@ -1851,70 +2003,69 @@ start_queue_dump(const struct netdev *netdev, struct nl_dump *dump) return false; } tcmsg->tcm_parent = 0; - nl_dump_start(dump, rtnl_sock, &request); + nl_dump_start(dump, NETLINK_ROUTE, &request); ofpbuf_uninit(&request); return true; } static int -netdev_linux_dump_queues(const struct netdev *netdev, +netdev_linux_dump_queues(const struct netdev *netdev_, netdev_dump_queues_cb *cb, void *aux) { - struct netdev_dev_linux *netdev_dev = - netdev_dev_linux_cast(netdev_get_dev(netdev)); - struct tc_queue *queue; - struct shash details; + struct netdev_linux *netdev = netdev_linux_cast(netdev_); + struct tc_queue *queue, *next_queue; + struct smap details; int last_error; int error; - error = tc_query_qdisc(netdev); + error = tc_query_qdisc(netdev_); if (error) { return error; - } else if (!netdev_dev->tc->ops->class_get) { + } else if (!netdev->tc->ops->class_get) { return EOPNOTSUPP; } last_error = 0; - shash_init(&details); - HMAP_FOR_EACH (queue, hmap_node, &netdev_dev->tc->queues) { - shash_clear(&details); + smap_init(&details); + HMAP_FOR_EACH_SAFE (queue, next_queue, hmap_node, + &netdev->tc->queues) { + smap_clear(&details); - error = netdev_dev->tc->ops->class_get(netdev, queue, &details); + error = netdev->tc->ops->class_get(netdev_, queue, &details); if (!error) { (*cb)(queue->queue_id, &details, aux); } else { last_error = error; } } - shash_destroy(&details); + smap_destroy(&details); return last_error; } static int -netdev_linux_dump_queue_stats(const struct netdev *netdev, +netdev_linux_dump_queue_stats(const struct netdev *netdev_, netdev_dump_queue_stats_cb *cb, void *aux) { - struct netdev_dev_linux *netdev_dev = - netdev_dev_linux_cast(netdev_get_dev(netdev)); + struct netdev_linux *netdev = netdev_linux_cast(netdev_); struct nl_dump dump; struct ofpbuf msg; int last_error; int error; - error = tc_query_qdisc(netdev); + error = tc_query_qdisc(netdev_); if (error) { return error; - } else if (!netdev_dev->tc->ops->class_dump_stats) { + } else if (!netdev->tc->ops->class_dump_stats) { return EOPNOTSUPP; } last_error = 0; - if (!start_queue_dump(netdev, &dump)) { + if (!start_queue_dump(netdev_, &dump)) { return ENODEV; } while (nl_dump_next(&dump, &msg)) { - error = netdev_dev->tc->ops->class_dump_stats(netdev, &msg, cb, aux); + error = netdev->tc->ops->class_dump_stats(netdev_, &msg, cb, aux); if (error) { last_error = error; } @@ -1928,28 +2079,27 @@ static int netdev_linux_get_in4(const struct netdev *netdev_, struct in_addr *address, struct in_addr *netmask) { - struct netdev_dev_linux *netdev_dev = - netdev_dev_linux_cast(netdev_get_dev(netdev_)); + struct netdev_linux *netdev = netdev_linux_cast(netdev_); - if (!(netdev_dev->cache_valid & VALID_IN4)) { + if (!(netdev->cache_valid & VALID_IN4)) { int error; - error = netdev_linux_get_ipv4(netdev_, &netdev_dev->address, + error = netdev_linux_get_ipv4(netdev_, &netdev->address, SIOCGIFADDR, "SIOCGIFADDR"); if (error) { return error; } - error = netdev_linux_get_ipv4(netdev_, &netdev_dev->netmask, + error = netdev_linux_get_ipv4(netdev_, &netdev->netmask, SIOCGIFNETMASK, "SIOCGIFNETMASK"); if (error) { return error; } - netdev_dev->cache_valid |= VALID_IN4; + netdev->cache_valid |= VALID_IN4; } - *address = netdev_dev->address; - *netmask = netdev_dev->netmask; + *address = netdev->address; + *netmask = netdev->netmask; return address->s_addr == INADDR_ANY ? EADDRNOTAVAIL : 0; } @@ -1957,15 +2107,14 @@ static int netdev_linux_set_in4(struct netdev *netdev_, struct in_addr address, struct in_addr netmask) { - struct netdev_dev_linux *netdev_dev = - netdev_dev_linux_cast(netdev_get_dev(netdev_)); + struct netdev_linux *netdev = netdev_linux_cast(netdev_); int error; error = do_set_addr(netdev_, SIOCSIFADDR, "SIOCSIFADDR", address); if (!error) { - netdev_dev->cache_valid |= VALID_IN4; - netdev_dev->address = address; - netdev_dev->netmask = netmask; + netdev->cache_valid |= VALID_IN4; + netdev->address = address; + netdev->netmask = netmask; if (address.s_addr != INADDR_ANY) { error = do_set_addr(netdev_, SIOCSIFNETMASK, "SIOCSIFNETMASK", netmask); @@ -1995,13 +2144,12 @@ parse_if_inet6_line(const char *line, static int netdev_linux_get_in6(const struct netdev *netdev_, struct in6_addr *in6) { - struct netdev_dev_linux *netdev_dev = - netdev_dev_linux_cast(netdev_get_dev(netdev_)); - if (!(netdev_dev->cache_valid & VALID_IN6)) { + struct netdev_linux *netdev = netdev_linux_cast(netdev_); + if (!(netdev->cache_valid & VALID_IN6)) { FILE *file; char line[128]; - netdev_dev->in6 = in6addr_any; + netdev->in6 = in6addr_any; file = fopen("/proc/net/if_inet6", "r"); if (file != NULL) { @@ -2012,15 +2160,15 @@ netdev_linux_get_in6(const struct netdev *netdev_, struct in6_addr *in6) if (parse_if_inet6_line(line, &in6_tmp, ifname) && !strcmp(name, ifname)) { - netdev_dev->in6 = in6_tmp; + netdev->in6 = in6_tmp; break; } } fclose(file); } - netdev_dev->cache_valid |= VALID_IN6; + netdev->cache_valid |= VALID_IN6; } - *in6 = netdev_dev->in6; + *in6 = netdev->in6; return 0; } @@ -2064,7 +2212,7 @@ netdev_linux_add_router(struct netdev *netdev OVS_UNUSED, struct in_addr router) rt.rt_flags = RTF_UP | RTF_GATEWAY; error = ioctl(af_inet_sock, SIOCADDRT, &rt) < 0 ? errno : 0; if (error) { - VLOG_WARN("ioctl(SIOCADDRT): %s", strerror(error)); + VLOG_WARN("ioctl(SIOCADDRT): %s", ovs_strerror(error)); } return error; } @@ -2081,7 +2229,7 @@ netdev_linux_get_next_hop(const struct in_addr *host, struct in_addr *next_hop, *netdev_name = NULL; stream = fopen(fn, "r"); if (stream == NULL) { - VLOG_WARN_RL(&rl, "%s: open failed: %s", fn, strerror(errno)); + VLOG_WARN_RL(&rl, "%s: open failed: %s", fn, ovs_strerror(errno)); return errno; } @@ -2131,25 +2279,41 @@ netdev_linux_get_next_hop(const struct in_addr *host, struct in_addr *next_hop, } static int -netdev_linux_get_status(const struct netdev *netdev, struct shash *sh) +netdev_linux_get_status(const struct netdev *netdev_, struct smap *smap) { - struct ethtool_drvinfo drvinfo; - int error; + struct netdev_linux *netdev = netdev_linux_cast(netdev_); + int error = 0; - 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)); + if (!(netdev->cache_valid & VALID_DRVINFO)) { + struct ethtool_cmd *cmd = (struct ethtool_cmd *) &netdev->drvinfo; + + COVERAGE_INC(netdev_get_ethtool); + memset(&netdev->drvinfo, 0, sizeof netdev->drvinfo); + error = netdev_linux_do_ethtool(netdev->up.name, + cmd, + ETHTOOL_GDRVINFO, + "ETHTOOL_GDRVINFO"); + if (!error) { + netdev->cache_valid |= VALID_DRVINFO; + } } + if (!error) { + smap_add(smap, "driver_name", netdev->drvinfo.driver); + smap_add(smap, "driver_version", netdev->drvinfo.version); + smap_add(smap, "firmware_version", netdev->drvinfo.fw_version); + } return error; } +static int +netdev_internal_get_status(const struct netdev *netdev OVS_UNUSED, + struct smap *smap) +{ + smap_add(smap, "driver_name", "openvswitch"); + return 0; +} + /* 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, @@ -2177,7 +2341,8 @@ netdev_linux_arp_lookup(const struct netdev *netdev, memcpy(mac, r.arp_ha.sa_data, ETH_ADDR_LEN); } else if (retval != ENXIO) { VLOG_WARN_RL(&rl, "%s: could not look up ARP entry for "IP_FMT": %s", - netdev_get_name(netdev), IP_ARGS(&ip), strerror(retval)); + netdev_get_name(netdev), IP_ARGS(ip), + ovs_strerror(retval)); } return retval; } @@ -2209,19 +2374,19 @@ iff_to_nd_flags(int iff) } static int -netdev_linux_update_flags(struct netdev *netdev, enum netdev_flags off, +netdev_linux_update_flags(struct netdev *netdev_, enum netdev_flags off, enum netdev_flags on, enum netdev_flags *old_flagsp) { + struct netdev_linux *netdev = netdev_linux_cast(netdev_); int old_flags, new_flags; - int error; + int error = 0; - error = get_flags(netdev, &old_flags); - if (!error) { - *old_flagsp = iff_to_nd_flags(old_flags); - new_flags = (old_flags & ~nd_to_iff_flags(off)) | nd_to_iff_flags(on); - if (new_flags != old_flags) { - error = set_flags(netdev, new_flags); - } + old_flags = netdev->ifi_flags; + *old_flagsp = iff_to_nd_flags(old_flags); + new_flags = (old_flags & ~nd_to_iff_flags(off)) | nd_to_iff_flags(on); + if (new_flags != old_flags) { + error = set_flags(netdev_get_name(netdev_), new_flags); + get_flags(netdev_, &netdev->ifi_flags); } return error; } @@ -2229,10 +2394,11 @@ netdev_linux_update_flags(struct netdev *netdev, enum netdev_flags off, static unsigned int netdev_linux_change_seq(const struct netdev *netdev) { - return netdev_dev_linux_cast(netdev_get_dev(netdev))->change_seq; + return netdev_linux_cast(netdev)->change_seq; } -#define NETDEV_LINUX_CLASS(NAME, CREATE, GET_STATS, SET_STATS) \ +#define NETDEV_LINUX_CLASS(NAME, CREATE, GET_STATS, SET_STATS, \ + GET_FEATURES, GET_STATUS) \ { \ NAME, \ \ @@ -2244,14 +2410,9 @@ netdev_linux_change_seq(const struct netdev *netdev) netdev_linux_destroy, \ NULL, /* get_config */ \ NULL, /* set_config */ \ + NULL, /* get_tunnel_config */ \ \ - netdev_linux_open, \ - netdev_linux_close, \ - \ - netdev_linux_listen, \ - netdev_linux_recv, \ - netdev_linux_recv_wait, \ - netdev_linux_drain, \ + netdev_linux_rx_open, \ \ netdev_linux_send, \ netdev_linux_send_wait, \ @@ -2267,7 +2428,7 @@ netdev_linux_change_seq(const struct netdev *netdev) GET_STATS, \ SET_STATS, \ \ - netdev_linux_get_features, \ + GET_FEATURES, \ netdev_linux_set_advertisements, \ \ netdev_linux_set_policing, \ @@ -2287,7 +2448,7 @@ netdev_linux_change_seq(const struct netdev *netdev) netdev_linux_get_in6, \ netdev_linux_add_router, \ netdev_linux_get_next_hop, \ - netdev_linux_get_status, \ + GET_STATUS, \ netdev_linux_arp_lookup, \ \ netdev_linux_update_flags, \ @@ -2300,21 +2461,34 @@ const struct netdev_class netdev_linux_class = "system", netdev_linux_create, netdev_linux_get_stats, - NULL); /* set_stats */ + NULL, /* set_stats */ + netdev_linux_get_features, + netdev_linux_get_status); const struct netdev_class netdev_tap_class = NETDEV_LINUX_CLASS( "tap", netdev_linux_create_tap, - netdev_pseudo_get_stats, - NULL); /* set_stats */ + netdev_tap_get_stats, + NULL, /* set_stats */ + netdev_linux_get_features, + netdev_linux_get_status); const struct netdev_class netdev_internal_class = NETDEV_LINUX_CLASS( "internal", netdev_linux_create, - netdev_pseudo_get_stats, - netdev_vport_set_stats); + netdev_internal_get_stats, + netdev_internal_set_stats, + NULL, /* get_features */ + netdev_internal_get_status); + +static const struct netdev_rx_class netdev_rx_linux_class = { + netdev_rx_linux_destroy, + netdev_rx_linux_recv, + netdev_rx_linux_wait, + netdev_rx_linux_drain, +}; /* HTB traffic control class. */ @@ -2334,25 +2508,23 @@ struct htb_class { }; static struct htb * -htb_get__(const struct netdev *netdev) +htb_get__(const struct netdev *netdev_) { - struct netdev_dev_linux *netdev_dev = - netdev_dev_linux_cast(netdev_get_dev(netdev)); - return CONTAINER_OF(netdev_dev->tc, struct htb, tc); + struct netdev_linux *netdev = netdev_linux_cast(netdev_); + return CONTAINER_OF(netdev->tc, struct htb, tc); } static void -htb_install__(struct netdev *netdev, uint64_t max_rate) +htb_install__(struct netdev *netdev_, uint64_t max_rate) { - struct netdev_dev_linux *netdev_dev = - netdev_dev_linux_cast(netdev_get_dev(netdev)); + struct netdev_linux *netdev = netdev_linux_cast(netdev_); struct htb *htb; htb = xmalloc(sizeof *htb); tc_init(&htb->tc, &tc_ops_htb); htb->max_rate = max_rate; - netdev_dev->tc = &htb->tc; + netdev->tc = &htb->tc; } /* Create an HTB qdisc. @@ -2439,7 +2611,7 @@ htb_setup_class__(struct netdev *netdev, unsigned int handle, tc_get_major(handle), tc_get_minor(handle), tc_get_major(parent), tc_get_minor(parent), class->min_rate, class->max_rate, - class->burst, class->priority, strerror(error)); + class->burst, class->priority, ovs_strerror(error)); } return error; } @@ -2500,17 +2672,17 @@ htb_parse_tcmsg__(struct ofpbuf *tcmsg, unsigned int *queue_id, static void htb_parse_qdisc_details__(struct netdev *netdev, - const struct shash *details, struct htb_class *hc) + const struct smap *details, struct htb_class *hc) { const char *max_rate_s; - max_rate_s = shash_find_data(details, "max-rate"); + max_rate_s = smap_get(details, "max-rate"); hc->max_rate = max_rate_s ? strtoull(max_rate_s, NULL, 10) / 8 : 0; if (!hc->max_rate) { - uint32_t current; + enum netdev_features current; netdev_get_features(netdev, ¤t, NULL, NULL, NULL); - hc->max_rate = netdev_features_to_bps(current) / 8; + hc->max_rate = netdev_features_to_bps(current, 100 * 1000 * 1000) / 8; } hc->min_rate = hc->max_rate; hc->burst = 0; @@ -2519,13 +2691,13 @@ htb_parse_qdisc_details__(struct netdev *netdev, static int htb_parse_class_details__(struct netdev *netdev, - const struct shash *details, struct htb_class *hc) + const struct smap *details, struct htb_class *hc) { const struct htb *htb = htb_get__(netdev); - const char *min_rate_s = shash_find_data(details, "min-rate"); - 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"); + const char *min_rate_s = smap_get(details, "min-rate"); + const char *max_rate_s = smap_get(details, "max-rate"); + const char *burst_s = smap_get(details, "burst"); + const char *priority_s = smap_get(details, "priority"); int mtu, error; error = netdev_get_mtu(netdev, &mtu); @@ -2583,7 +2755,7 @@ htb_query_class__(const struct netdev *netdev, unsigned int handle, } static int -htb_tc_install(struct netdev *netdev, const struct shash *details) +htb_tc_install(struct netdev *netdev, const struct smap *details) { int error; @@ -2623,6 +2795,7 @@ htb_update_queue__(struct netdev *netdev, unsigned int queue_id, hcp = xmalloc(sizeof *hcp); queue = &hcp->tc_queue; queue->queue_id = queue_id; + queue->created = time_msec(); hmap_insert(&htb->tc.queues, &queue->hmap_node, hash); } @@ -2675,15 +2848,15 @@ htb_tc_destroy(struct tc *tc) } static int -htb_qdisc_get(const struct netdev *netdev, struct shash *details) +htb_qdisc_get(const struct netdev *netdev, struct smap *details) { const struct htb *htb = htb_get__(netdev); - shash_add(details, "max-rate", xasprintf("%llu", 8ULL * htb->max_rate)); + smap_add_format(details, "max-rate", "%llu", 8ULL * htb->max_rate); return 0; } static int -htb_qdisc_set(struct netdev *netdev, const struct shash *details) +htb_qdisc_set(struct netdev *netdev, const struct smap *details) { struct htb_class hc; int error; @@ -2699,24 +2872,24 @@ htb_qdisc_set(struct netdev *netdev, const struct shash *details) static int htb_class_get(const struct netdev *netdev OVS_UNUSED, - const struct tc_queue *queue, struct shash *details) + const struct tc_queue *queue, struct smap *details) { const struct htb_class *hc = htb_class_cast__(queue); - shash_add(details, "min-rate", xasprintf("%llu", 8ULL * hc->min_rate)); + smap_add_format(details, "min-rate", "%llu", 8ULL * hc->min_rate); if (hc->min_rate != hc->max_rate) { - shash_add(details, "max-rate", xasprintf("%llu", 8ULL * hc->max_rate)); + smap_add_format(details, "max-rate", "%llu", 8ULL * hc->max_rate); } - shash_add(details, "burst", xasprintf("%llu", 8ULL * hc->burst)); + smap_add_format(details, "burst", "%llu", 8ULL * hc->burst); if (hc->priority) { - shash_add(details, "priority", xasprintf("%u", hc->priority)); + smap_add_format(details, "priority", "%u", hc->priority); } return 0; } static int htb_class_set(struct netdev *netdev, unsigned int queue_id, - const struct shash *details) + const struct smap *details) { struct htb_class hc; int error; @@ -2813,11 +2986,10 @@ struct hfsc_class { }; static struct hfsc * -hfsc_get__(const struct netdev *netdev) +hfsc_get__(const struct netdev *netdev_) { - struct netdev_dev_linux *netdev_dev; - netdev_dev = netdev_dev_linux_cast(netdev_get_dev(netdev)); - return CONTAINER_OF(netdev_dev->tc, struct hfsc, tc); + struct netdev_linux *netdev = netdev_linux_cast(netdev_); + return CONTAINER_OF(netdev->tc, struct hfsc, tc); } static struct hfsc_class * @@ -2827,16 +2999,15 @@ hfsc_class_cast__(const struct tc_queue *queue) } static void -hfsc_install__(struct netdev *netdev, uint32_t max_rate) +hfsc_install__(struct netdev *netdev_, uint32_t max_rate) { - struct netdev_dev_linux * netdev_dev; + struct netdev_linux *netdev = netdev_linux_cast(netdev_); struct hfsc *hfsc; - netdev_dev = netdev_dev_linux_cast(netdev_get_dev(netdev)); hfsc = xmalloc(sizeof *hfsc); tc_init(&hfsc->tc, &tc_ops_hfsc); hfsc->max_rate = max_rate; - netdev_dev->tc = &hfsc->tc; + netdev->tc = &hfsc->tc; } static void @@ -2858,6 +3029,7 @@ hfsc_update_queue__(struct netdev *netdev, unsigned int queue_id, hcp = xmalloc(sizeof *hcp); queue = &hcp->tc_queue; queue->queue_id = queue_id; + queue->created = time_msec(); hmap_insert(&hfsc->tc.queues, &queue->hmap_node, hash); } @@ -2976,20 +3148,20 @@ hfsc_query_class__(const struct netdev *netdev, unsigned int handle, } static void -hfsc_parse_qdisc_details__(struct netdev *netdev, const struct shash *details, +hfsc_parse_qdisc_details__(struct netdev *netdev, const struct smap *details, struct hfsc_class *class) { uint32_t max_rate; const char *max_rate_s; - max_rate_s = shash_find_data(details, "max-rate"); + max_rate_s = smap_get(details, "max-rate"); max_rate = max_rate_s ? strtoull(max_rate_s, NULL, 10) / 8 : 0; if (!max_rate) { - uint32_t current; + enum netdev_features current; netdev_get_features(netdev, ¤t, NULL, NULL, NULL); - max_rate = netdev_features_to_bps(current) / 8; + max_rate = netdev_features_to_bps(current, 100 * 1000 * 1000) / 8; } class->min_rate = max_rate; @@ -2998,7 +3170,7 @@ hfsc_parse_qdisc_details__(struct netdev *netdev, const struct shash *details, static int hfsc_parse_class_details__(struct netdev *netdev, - const struct shash *details, + const struct smap *details, struct hfsc_class * class) { const struct hfsc *hfsc; @@ -3006,8 +3178,8 @@ hfsc_parse_class_details__(struct netdev *netdev, const char *min_rate_s, *max_rate_s; hfsc = hfsc_get__(netdev); - min_rate_s = shash_find_data(details, "min-rate"); - max_rate_s = shash_find_data(details, "max-rate"); + min_rate_s = smap_get(details, "min-rate"); + max_rate_s = smap_get(details, "max-rate"); min_rate = min_rate_s ? strtoull(min_rate_s, NULL, 10) / 8 : 0; min_rate = MAX(min_rate, 1); @@ -3101,14 +3273,14 @@ hfsc_setup_class__(struct netdev *netdev, unsigned int handle, netdev_get_name(netdev), tc_get_major(handle), tc_get_minor(handle), tc_get_major(parent), tc_get_minor(parent), - class->min_rate, class->max_rate, strerror(error)); + class->min_rate, class->max_rate, ovs_strerror(error)); } return error; } static int -hfsc_tc_install(struct netdev *netdev, const struct shash *details) +hfsc_tc_install(struct netdev *netdev, const struct smap *details) { int error; struct hfsc_class class; @@ -3176,16 +3348,16 @@ hfsc_tc_destroy(struct tc *tc) } static int -hfsc_qdisc_get(const struct netdev *netdev, struct shash *details) +hfsc_qdisc_get(const struct netdev *netdev, struct smap *details) { const struct hfsc *hfsc; hfsc = hfsc_get__(netdev); - shash_add(details, "max-rate", xasprintf("%llu", 8ULL * hfsc->max_rate)); + smap_add_format(details, "max-rate", "%llu", 8ULL * hfsc->max_rate); return 0; } static int -hfsc_qdisc_set(struct netdev *netdev, const struct shash *details) +hfsc_qdisc_set(struct netdev *netdev, const struct smap *details) { int error; struct hfsc_class class; @@ -3203,21 +3375,21 @@ hfsc_qdisc_set(struct netdev *netdev, const struct shash *details) static int hfsc_class_get(const struct netdev *netdev OVS_UNUSED, - const struct tc_queue *queue, struct shash *details) + const struct tc_queue *queue, struct smap *details) { const struct hfsc_class *hc; hc = hfsc_class_cast__(queue); - shash_add(details, "min-rate", xasprintf("%llu", 8ULL * hc->min_rate)); + smap_add_format(details, "min-rate", "%llu", 8ULL * hc->min_rate); if (hc->min_rate != hc->max_rate) { - shash_add(details, "max-rate", xasprintf("%llu", 8ULL * hc->max_rate)); + smap_add_format(details, "max-rate", "%llu", 8ULL * hc->max_rate); } return 0; } static int hfsc_class_set(struct netdev *netdev, unsigned int queue_id, - const struct shash *details) + const struct smap *details) { int error; struct hfsc_class class; @@ -3307,22 +3479,19 @@ static const struct tc_ops tc_ops_hfsc = { * the "" (empty string) QoS type in the OVS database. */ static void -default_install__(struct netdev *netdev) +default_install__(struct netdev *netdev_) { - struct netdev_dev_linux *netdev_dev = - netdev_dev_linux_cast(netdev_get_dev(netdev)); - static struct tc *tc; + struct netdev_linux *netdev = netdev_linux_cast(netdev_); + static const struct tc tc = TC_INITIALIZER(&tc, &tc_ops_default); - if (!tc) { - tc = xmalloc(sizeof *tc); - tc_init(tc, &tc_ops_default); - } - netdev_dev->tc = tc; + /* Nothing but a tc class implementation is allowed to write to a tc. This + * class never does that, so we can legitimately use a const tc object. */ + netdev->tc = CONST_CAST(struct tc *, &tc); } static int default_tc_install(struct netdev *netdev, - const struct shash *details OVS_UNUSED) + const struct smap *details OVS_UNUSED) { default_install__(netdev); return 0; @@ -3356,17 +3525,14 @@ static const struct tc_ops tc_ops_default = { * */ static int -other_tc_load(struct netdev *netdev, struct ofpbuf *nlmsg OVS_UNUSED) +other_tc_load(struct netdev *netdev_, struct ofpbuf *nlmsg OVS_UNUSED) { - struct netdev_dev_linux *netdev_dev = - netdev_dev_linux_cast(netdev_get_dev(netdev)); - static struct tc *tc; + struct netdev_linux *netdev = netdev_linux_cast(netdev_); + static const struct tc tc = TC_INITIALIZER(&tc, &tc_ops_other); - if (!tc) { - tc = xmalloc(sizeof *tc); - tc_init(tc, &tc_ops_other); - } - netdev_dev->tc = tc; + /* Nothing but a tc class implementation is allowed to write to a tc. This + * class never does that, so we can legitimately use a const tc object. */ + netdev->tc = CONST_CAST(struct tc *, &tc); return 0; } @@ -3458,7 +3624,7 @@ tc_make_request(const struct netdev *netdev, int type, unsigned int flags, static int tc_transact(struct ofpbuf *request, struct ofpbuf **replyp) { - int error = nl_sock_transact(rtnl_sock, request, replyp); + int error = nl_transact(NETLINK_ROUTE, request, replyp); ofpbuf_uninit(request); return error; } @@ -3535,7 +3701,7 @@ tc_add_policer(struct netdev *netdev, int kbits_rate, int kbits_burst) 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_fill_rate(&tc_police.rate, (kbits_rate * 1000)/8, mtu); tc_police.burst = tc_bytes_to_ticks(tc_police.rate.rate, kbits_burst * 1024); @@ -3603,30 +3769,35 @@ read_psched(void) * [5] 2.6.32.21.22 (approx.) from Ubuntu 10.04 on VMware Fusion * [6] 2.6.34 from kernel.org on KVM */ + static struct ovsthread_once once = OVSTHREAD_ONCE_INITIALIZER; static const char fn[] = "/proc/net/psched"; unsigned int a, b, c, d; FILE *stream; + if (!ovsthread_once_start(&once)) { + return; + } + ticks_per_s = 1.0; buffer_hz = 100; stream = fopen(fn, "r"); if (!stream) { - VLOG_WARN("%s: open failed: %s", fn, strerror(errno)); - return; + VLOG_WARN("%s: open failed: %s", fn, ovs_strerror(errno)); + goto exit; } if (fscanf(stream, "%x %x %x %x", &a, &b, &c, &d) != 4) { VLOG_WARN("%s: read failed", fn); fclose(stream); - return; + goto exit; } VLOG_DBG("%s: psched parameters are: %u %u %u %u", fn, a, b, c, d); fclose(stream); if (!a || !c) { VLOG_WARN("%s: invalid scheduler parameters", fn); - return; + goto exit; } ticks_per_s = (double) a * c / b; @@ -3637,6 +3808,9 @@ read_psched(void) fn, a, b, c, d); } VLOG_DBG("%s: ticks_per_s=%f buffer_hz=%u", fn, ticks_per_s, buffer_hz); + +exit: + ovsthread_once_done(&once); } /* Returns the number of bytes that can be transmitted in 'ticks' ticks at a @@ -3644,9 +3818,7 @@ read_psched(void) static unsigned int tc_ticks_to_bytes(unsigned int rate, unsigned int ticks) { - if (!buffer_hz) { - read_psched(); - } + read_psched(); return (rate * ticks) / ticks_per_s; } @@ -3655,9 +3827,7 @@ tc_ticks_to_bytes(unsigned int rate, unsigned int ticks) static unsigned int tc_bytes_to_ticks(unsigned int rate, unsigned int size) { - if (!buffer_hz) { - read_psched(); - } + read_psched(); return rate ? ((unsigned long long int) ticks_per_s * size) / rate : 0; } @@ -3666,9 +3836,7 @@ tc_bytes_to_ticks(unsigned int rate, unsigned int size) static unsigned int tc_buffer_per_jiffy(unsigned int rate) { - if (!buffer_hz) { - read_psched(); - } + read_psched(); return rate / buffer_hz; } @@ -3816,7 +3984,7 @@ tc_query_class(const struct netdev *netdev, netdev_get_name(netdev), tc_get_major(handle), tc_get_minor(handle), tc_get_major(parent), tc_get_minor(parent), - strerror(error)); + ovs_strerror(error)); } return error; } @@ -3841,22 +4009,21 @@ tc_delete_class(const struct netdev *netdev, unsigned int handle) VLOG_WARN_RL(&rl, "delete %s class %u:%u failed (%s)", netdev_get_name(netdev), tc_get_major(handle), tc_get_minor(handle), - strerror(error)); + ovs_strerror(error)); } return error; } /* Equivalent to "tc qdisc del dev root". */ static int -tc_del_qdisc(struct netdev *netdev) +tc_del_qdisc(struct netdev *netdev_) { - struct netdev_dev_linux *netdev_dev = - netdev_dev_linux_cast(netdev_get_dev(netdev)); + struct netdev_linux *netdev = netdev_linux_cast(netdev_); struct ofpbuf request; struct tcmsg *tcmsg; int error; - tcmsg = tc_make_request(netdev, RTM_DELQDISC, 0, &request); + tcmsg = tc_make_request(netdev_, RTM_DELQDISC, 0, &request); if (!tcmsg) { return ENODEV; } @@ -3869,11 +4036,11 @@ tc_del_qdisc(struct netdev *netdev) * case we've accomplished our purpose. */ error = 0; } - if (!error && netdev_dev->tc) { - if (netdev_dev->tc->ops->tc_destroy) { - netdev_dev->tc->ops->tc_destroy(netdev_dev->tc); + if (!error && netdev->tc) { + if (netdev->tc->ops->tc_destroy) { + netdev->tc->ops->tc_destroy(netdev->tc); } - netdev_dev->tc = NULL; + netdev->tc = NULL; } return error; } @@ -3882,17 +4049,16 @@ tc_del_qdisc(struct netdev *netdev) * kernel to determine what they are. Returns 0 if successful, otherwise a * positive errno value. */ static int -tc_query_qdisc(const struct netdev *netdev) +tc_query_qdisc(const struct netdev *netdev_) { - struct netdev_dev_linux *netdev_dev = - netdev_dev_linux_cast(netdev_get_dev(netdev)); + struct netdev_linux *netdev = netdev_linux_cast(netdev_); struct ofpbuf request, *qdisc; const struct tc_ops *ops; struct tcmsg *tcmsg; int load_error; int error; - if (netdev_dev->tc) { + if (netdev->tc) { return 0; } @@ -3910,7 +4076,7 @@ tc_query_qdisc(const struct netdev *netdev) * * We could check for Linux 2.6.35+ and use a more straightforward method * there. */ - tcmsg = tc_make_request(netdev, RTM_GETQDISC, NLM_F_ECHO, &request); + tcmsg = tc_make_request(netdev_, RTM_GETQDISC, NLM_F_ECHO, &request); if (!tcmsg) { return ENODEV; } @@ -3943,13 +4109,13 @@ tc_query_qdisc(const struct netdev *netdev) } else { /* Who knows? Maybe the device got deleted. */ VLOG_WARN_RL(&rl, "query %s qdisc failed (%s)", - netdev_get_name(netdev), strerror(error)); + netdev_get_name(netdev_), ovs_strerror(error)); ops = &tc_ops_other; } /* Instantiate it. */ - load_error = ops->tc_load((struct netdev *) netdev, qdisc); - assert((load_error == 0) == (netdev_dev->tc != NULL)); + load_error = ops->tc_load(CONST_CAST(struct netdev *, netdev_), qdisc); + ovs_assert((load_error == 0) == (netdev->tc != NULL)); ofpbuf_delete(qdisc); return error ? error : load_error; @@ -4045,6 +4211,7 @@ netdev_linux_ethtool_set_flag(struct netdev *netdev, uint32_t flag, uint32_t new_flags; int error; + COVERAGE_INC(netdev_get_ethtool); memset(&evalue, 0, sizeof evalue); error = netdev_linux_do_ethtool(netdev_name, (struct ethtool_cmd *)&evalue, @@ -4053,6 +4220,7 @@ netdev_linux_ethtool_set_flag(struct netdev *netdev, uint32_t flag, return error; } + COVERAGE_INC(netdev_set_ethtool); evalue.data = new_flags = (evalue.data & ~flag) | (enable ? flag : 0); error = netdev_linux_do_ethtool(netdev_name, (struct ethtool_cmd *)&evalue, @@ -4061,6 +4229,7 @@ netdev_linux_ethtool_set_flag(struct netdev *netdev, uint32_t flag, return error; } + COVERAGE_INC(netdev_get_ethtool); memset(&evalue, 0, sizeof evalue); error = netdev_linux_do_ethtool(netdev_name, (struct ethtool_cmd *)&evalue, @@ -4133,7 +4302,7 @@ get_stats_via_netlink(int ifindex, struct netdev_stats *stats) ifi = ofpbuf_put_zeros(&request, sizeof *ifi); ifi->ifi_family = PF_UNSPEC; ifi->ifi_index = ifindex; - error = nl_sock_transact(rtnl_sock, &request, &reply); + error = nl_transact(NETLINK_ROUTE, &request, &reply); ofpbuf_uninit(&request); if (error) { return error; @@ -4169,7 +4338,7 @@ get_stats_via_proc(const char *netdev_name, struct netdev_stats *stats) stream = fopen(fn, "r"); if (!stream) { - VLOG_WARN_RL(&rl, "%s: open failed: %s", fn, strerror(errno)); + VLOG_WARN_RL(&rl, "%s: open failed: %s", fn, ovs_strerror(errno)); return errno; } @@ -4217,77 +4386,27 @@ get_stats_via_proc(const char *netdev_name, struct netdev_stats *stats) } 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) +get_flags(const struct netdev *dev, unsigned int *flags) { struct ifreq ifr; int error; - error = netdev_linux_do_ioctl(netdev_get_name(netdev), &ifr, SIOCGIFFLAGS, + *flags = 0; + error = netdev_linux_do_ioctl(dev->name, &ifr, SIOCGIFFLAGS, "SIOCGIFFLAGS"); - *flags = ifr.ifr_flags; + if (!error) { + *flags = ifr.ifr_flags; + } return error; } static int -set_flags(struct netdev *netdev, int flags) +set_flags(const char *name, unsigned int flags) { struct ifreq ifr; ifr.ifr_flags = flags; - return netdev_linux_do_ioctl(netdev_get_name(netdev), &ifr, SIOCSIFFLAGS, - "SIOCSIFFLAGS"); + return netdev_linux_do_ioctl(name, &ifr, SIOCSIFFLAGS, "SIOCSIFFLAGS"); } static int @@ -4299,7 +4418,7 @@ do_get_ifindex(const char *netdev_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", - netdev_name, strerror(errno)); + netdev_name, ovs_strerror(errno)); return -errno; } return ifr.ifr_ifindex; @@ -4308,19 +4427,23 @@ do_get_ifindex(const char *netdev_name) static int get_ifindex(const struct netdev *netdev_, int *ifindexp) { - struct netdev_dev_linux *netdev_dev = - netdev_dev_linux_cast(netdev_get_dev(netdev_)); - *ifindexp = 0; - if (!(netdev_dev->cache_valid & VALID_IFINDEX)) { + struct netdev_linux *netdev = netdev_linux_cast(netdev_); + + if (!(netdev->cache_valid & VALID_IFINDEX)) { int ifindex = do_get_ifindex(netdev_get_name(netdev_)); + if (ifindex < 0) { - return -ifindex; + netdev->get_ifindex_error = -ifindex; + netdev->ifindex = 0; + } else { + netdev->get_ifindex_error = 0; + netdev->ifindex = ifindex; } - netdev_dev->cache_valid |= VALID_IFINDEX; - netdev_dev->ifindex = ifindex; + netdev->cache_valid |= VALID_IFINDEX; } - *ifindexp = netdev_dev->ifindex; - return 0; + + *ifindexp = netdev->ifindex; + return netdev->get_ifindex_error; } static int @@ -4338,7 +4461,7 @@ get_etheraddr(const char *netdev_name, uint8_t ea[ETH_ADDR_LEN]) * to INFO for that case. */ VLOG(errno == ENODEV ? VLL_INFO : VLL_ERR, "ioctl(SIOCGIFHWADDR) on %s device failed: %s", - netdev_name, strerror(errno)); + netdev_name, ovs_strerror(errno)); return errno; } hwaddr_family = ifr.ifr_hwaddr.sa_family; @@ -4351,19 +4474,19 @@ get_etheraddr(const char *netdev_name, uint8_t ea[ETH_ADDR_LEN]) } static int -set_etheraddr(const char *netdev_name, int hwaddr_family, +set_etheraddr(const char *netdev_name, const uint8_t mac[ETH_ADDR_LEN]) { struct ifreq ifr; memset(&ifr, 0, sizeof ifr); ovs_strzcpy(ifr.ifr_name, netdev_name, sizeof ifr.ifr_name); - ifr.ifr_hwaddr.sa_family = hwaddr_family; + ifr.ifr_hwaddr.sa_family = ARPHRD_ETHER; memcpy(ifr.ifr_hwaddr.sa_data, mac, ETH_ADDR_LEN); COVERAGE_INC(netdev_set_hwaddr); if (ioctl(af_inet_sock, SIOCSIFHWADDR, &ifr) < 0) { VLOG_ERR("ioctl(SIOCSIFHWADDR) on %s device failed: %s", - netdev_name, strerror(errno)); + netdev_name, ovs_strerror(errno)); return errno; } return 0; @@ -4380,13 +4503,12 @@ netdev_linux_do_ethtool(const char *name, struct ethtool_cmd *ecmd, ifr.ifr_data = (caddr_t) ecmd; ecmd->cmd = cmd; - COVERAGE_INC(netdev_ethtool); if (ioctl(af_inet_sock, SIOCETHTOOL, &ifr) == 0) { return 0; } else { if (errno != EOPNOTSUPP) { VLOG_WARN_RL(&rl, "ethtool command %s on network device %s " - "failed: %s", cmd_name, name, strerror(errno)); + "failed: %s", cmd_name, name, ovs_strerror(errno)); } else { /* The device doesn't support this operation. That's pretty * common, so there's no point in logging anything. */ @@ -4402,7 +4524,7 @@ netdev_linux_do_ioctl(const char *name, struct ifreq *ifr, int cmd, 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)); + ovs_strerror(errno)); return errno; } return 0; @@ -4418,7 +4540,8 @@ netdev_linux_get_ipv4(const struct netdev *netdev, struct in_addr *ip, ifr.ifr_addr.sa_family = AF_INET; error = netdev_linux_do_ioctl(netdev_get_name(netdev), &ifr, cmd, cmd_name); if (!error) { - const struct sockaddr_in *sin = (struct sockaddr_in *) &ifr.ifr_addr; + const struct sockaddr_in *sin = ALIGNED_CAST(struct sockaddr_in *, + &ifr.ifr_addr); *ip = sin->sin_addr; } return error; @@ -4428,16 +4551,23 @@ netdev_linux_get_ipv4(const struct netdev *netdev, struct in_addr *ip, static int af_packet_sock(void) { - static int sock = INT_MIN; + static struct ovsthread_once once = OVSTHREAD_ONCE_INITIALIZER; + static int sock; - if (sock == INT_MIN) { + if (ovsthread_once_start(&once)) { sock = socket(AF_PACKET, SOCK_RAW, 0); if (sock >= 0) { - set_nonblocking(sock); + int error = set_nonblocking(sock); + if (error) { + close(sock); + sock = -error; + } } else { sock = -errno; - VLOG_ERR("failed to create packet socket: %s", strerror(errno)); + VLOG_ERR("failed to create packet socket: %s", + ovs_strerror(errno)); } + ovsthread_once_done(&once); } return sock;