X-Git-Url: http://git.onelab.eu/?a=blobdiff_plain;f=lib%2Fnetdev-vport.c;h=330854493a1903e7afeaedcb134710cf736e7ed2;hb=85b20fd6ee585f462e012fbcc7f966a81edab2ed;hp=e11cb2a8d685156d4757a38d4e8e07a14beeb21f;hpb=007948177581f3b3dad188221593d0e4bdca6ba0;p=sliver-openvswitch.git diff --git a/lib/netdev-vport.c b/lib/netdev-vport.c index e11cb2a8d..330854493 100644 --- a/lib/netdev-vport.c +++ b/lib/netdev-vport.c @@ -1,5 +1,5 @@ /* - * Copyright (c) 2010, 2011 Nicira Networks. + * Copyright (c) 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. @@ -21,6 +21,7 @@ #include #include #include +#include #include #include #include @@ -35,59 +36,51 @@ #include "netdev-linux.h" #include "netdev-provider.h" #include "netlink.h" +#include "netlink-notifier.h" #include "netlink-socket.h" #include "ofpbuf.h" -#include "openvswitch/datapath-protocol.h" #include "openvswitch/tunnel.h" #include "packets.h" #include "route-table.h" -#include "rtnetlink.h" #include "shash.h" #include "socket-util.h" +#include "unaligned.h" #include "vlog.h" VLOG_DEFINE_THIS_MODULE(netdev_vport); -struct netdev_vport_notifier { - struct netdev_notifier notifier; - struct list list_node; - struct shash_node *shash_node; -}; +/* Default to the OTV port, per the VXLAN IETF draft. */ +#define VXLAN_DST_PORT 8472 + +#define DEFAULT_TTL 64 struct netdev_dev_vport { struct netdev_dev netdev_dev; struct ofpbuf *options; - int dp_ifindex; /* -1 if unknown. */ - uint32_t port_no; /* UINT32_MAX if unknown. */ -}; - -struct netdev_vport { - struct netdev netdev; + unsigned int change_seq; + uint8_t etheraddr[ETH_ADDR_LEN]; + struct netdev_tunnel_config tnl_cfg; }; struct vport_class { - enum odp_vport_type type; + enum ovs_vport_type type; struct netdev_class netdev_class; int (*parse_config)(const char *name, const char *type, - const struct shash *args, struct ofpbuf *options); + const struct smap *args, struct ofpbuf *options, + struct netdev_tunnel_config *tnl_cfg); int (*unparse_config)(const char *name, const char *type, const struct nlattr *options, size_t options_len, - struct shash *args); + struct smap *args); }; -static struct shash netdev_vport_notifiers = - SHASH_INITIALIZER(&netdev_vport_notifiers); - static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(5, 20); static int netdev_vport_create(const struct netdev_class *, const char *, - const struct shash *, struct netdev_dev **); + struct netdev_dev **); static void netdev_vport_poll_notify(const struct netdev *); static int tnl_port_config_from_nlattr(const struct nlattr *options, size_t options_len, - struct nlattr *a[ODP_TUNNEL_ATTR_MAX + 1]); - -static const char *netdev_vport_get_tnl_iface(const struct netdev *netdev); + struct nlattr *a[OVS_TUNNEL_ATTR_MAX + 1]); static bool is_vport_class(const struct netdev_class *class) @@ -109,16 +102,20 @@ netdev_dev_vport_cast(const struct netdev_dev *netdev_dev) return CONTAINER_OF(netdev_dev, struct netdev_dev_vport, netdev_dev); } -static struct netdev_vport * -netdev_vport_cast(const struct netdev *netdev) +static struct netdev_dev_vport * +netdev_vport_get_dev(const struct netdev *netdev) { - struct netdev_dev *netdev_dev = netdev_get_dev(netdev); - assert(is_vport_class(netdev_dev_get_class(netdev_dev))); - return CONTAINER_OF(netdev, struct netdev_vport, netdev); + return netdev_dev_vport_cast(netdev_get_dev(netdev)); +} + +static const struct netdev_tunnel_config * +get_netdev_tunnel_config(const struct netdev_dev *netdev_dev) +{ + return &netdev_dev_vport_cast(netdev_dev)->tnl_cfg; } /* If 'netdev' is a vport netdev, returns an ofpbuf that contains Netlink - * options to include in ODP_VPORT_ATTR_OPTIONS for configuring that vport. + * options to include in OVS_VPORT_ATTR_OPTIONS for configuring that vport. * Otherwise returns NULL. */ const struct ofpbuf * netdev_vport_get_options(const struct netdev *netdev) @@ -130,48 +127,67 @@ netdev_vport_get_options(const struct netdev *netdev) : NULL); } -enum odp_vport_type +enum ovs_vport_type netdev_vport_get_vport_type(const struct netdev *netdev) { const struct netdev_dev *dev = netdev_get_dev(netdev); const struct netdev_class *class = netdev_dev_get_class(dev); return (is_vport_class(class) ? vport_class_cast(class)->type - : class == &netdev_internal_class ? ODP_VPORT_TYPE_INTERNAL - : class == &netdev_linux_class ? ODP_VPORT_TYPE_NETDEV - : ODP_VPORT_TYPE_UNSPEC); + : class == &netdev_internal_class ? OVS_VPORT_TYPE_INTERNAL + : (class == &netdev_linux_class || + class == &netdev_tap_class) ? OVS_VPORT_TYPE_NETDEV + : OVS_VPORT_TYPE_UNSPEC); +} + +static uint32_t +get_u32_or_zero(const struct nlattr *a) +{ + return a ? nl_attr_get_u32(a) : 0; } const char * netdev_vport_get_netdev_type(const struct dpif_linux_vport *vport) { - struct nlattr *a[ODP_TUNNEL_ATTR_MAX + 1]; + struct nlattr *a[OVS_TUNNEL_ATTR_MAX + 1]; switch (vport->type) { - case ODP_VPORT_TYPE_UNSPEC: + case OVS_VPORT_TYPE_UNSPEC: break; - case ODP_VPORT_TYPE_NETDEV: + case OVS_VPORT_TYPE_NETDEV: return "system"; - case ODP_VPORT_TYPE_INTERNAL: + case OVS_VPORT_TYPE_INTERNAL: return "internal"; - case ODP_VPORT_TYPE_PATCH: + case OVS_VPORT_TYPE_PATCH: return "patch"; - case ODP_VPORT_TYPE_GRE: + case OVS_VPORT_TYPE_GRE: if (tnl_port_config_from_nlattr(vport->options, vport->options_len, a)) { break; } - return (nl_attr_get_u32(a[ODP_TUNNEL_ATTR_FLAGS]) & TNL_F_IPSEC + return (get_u32_or_zero(a[OVS_TUNNEL_ATTR_FLAGS]) & TNL_F_IPSEC ? "ipsec_gre" : "gre"); - case ODP_VPORT_TYPE_CAPWAP: + case OVS_VPORT_TYPE_GRE64: + if (tnl_port_config_from_nlattr(vport->options, vport->options_len, + a)) { + break; + } + return (get_u32_or_zero(a[OVS_TUNNEL_ATTR_FLAGS]) & TNL_F_IPSEC + ? "ipsec_gre64" : "gre64"); + + case OVS_VPORT_TYPE_CAPWAP: return "capwap"; - case __ODP_VPORT_TYPE_MAX: + case OVS_VPORT_TYPE_VXLAN: + return "vxlan"; + + case OVS_VPORT_TYPE_FT_GRE: + case __OVS_VPORT_TYPE_MAX: break; } @@ -182,76 +198,19 @@ netdev_vport_get_netdev_type(const struct dpif_linux_vport *vport) static int netdev_vport_create(const struct netdev_class *netdev_class, const char *name, - const struct shash *args, struct netdev_dev **netdev_devp) { - const struct vport_class *vport_class = vport_class_cast(netdev_class); - struct ofpbuf *options = NULL; - struct shash fetched_args; - int dp_ifindex; - uint32_t port_no; - int error; + struct netdev_dev_vport *dev; - shash_init(&fetched_args); + dev = xzalloc(sizeof *dev); + netdev_dev_init(&dev->netdev_dev, name, netdev_class); + dev->change_seq = 1; + eth_addr_random(dev->etheraddr); - dp_ifindex = -1; - port_no = UINT32_MAX; - if (!shash_is_empty(args)) { - /* Parse the provided configuration. */ - options = ofpbuf_new(64); - error = vport_class->parse_config(name, netdev_class->type, - args, options); - } else { - /* Fetch an existing configuration from the kernel. - * - * This case could be ambiguous with initializing a new vport with an - * empty configuration, but none of the existing vport classes accept - * an empty configuration. */ - struct dpif_linux_vport reply; - struct ofpbuf *buf; + *netdev_devp = &dev->netdev_dev; + route_table_register(); - error = dpif_linux_vport_get(name, &reply, &buf); - if (!error) { - /* XXX verify correct type */ - error = vport_class->unparse_config(name, netdev_class->type, - reply.options, - reply.options_len, - &fetched_args); - if (error) { - VLOG_ERR_RL(&rl, "%s: failed to parse kernel config (%s)", - name, strerror(error)); - } else { - options = ofpbuf_clone_data(reply.options, reply.options_len); - dp_ifindex = reply.dp_ifindex; - port_no = reply.port_no; - } - ofpbuf_delete(buf); - } else { - VLOG_ERR_RL(&rl, "%s: vport query failed (%s)", - name, strerror(error)); - } - } - - if (!error) { - struct netdev_dev_vport *dev; - - dev = xmalloc(sizeof *dev); - netdev_dev_init(&dev->netdev_dev, name, - shash_is_empty(&fetched_args) ? args : &fetched_args, - netdev_class); - dev->options = options; - dev->dp_ifindex = dp_ifindex; - dev->port_no = port_no; - - *netdev_devp = &dev->netdev_dev; - route_table_register(); - } else { - ofpbuf_delete(options); - } - - shash_destroy(&fetched_args); - - return error; + return 0; } static void @@ -259,50 +218,82 @@ netdev_vport_destroy(struct netdev_dev *netdev_dev_) { struct netdev_dev_vport *netdev_dev = netdev_dev_vport_cast(netdev_dev_); + ofpbuf_delete(netdev_dev->options); route_table_unregister(); free(netdev_dev); } static int -netdev_vport_open(struct netdev_dev *netdev_dev_, int ethertype OVS_UNUSED, - struct netdev **netdevp) +netdev_vport_open(struct netdev_dev *netdev_dev, struct netdev **netdevp) { - struct netdev_vport *netdev; - - netdev = xmalloc(sizeof *netdev); - netdev_init(&netdev->netdev, netdev_dev_); - - *netdevp = &netdev->netdev; + *netdevp = xmalloc(sizeof **netdevp); + netdev_init(*netdevp, netdev_dev); return 0; } static void -netdev_vport_close(struct netdev *netdev_) +netdev_vport_close(struct netdev *netdev) { - struct netdev_vport *netdev = netdev_vport_cast(netdev_); free(netdev); } static int -netdev_vport_set_config(struct netdev_dev *dev_, const struct shash *args) +netdev_vport_get_config(struct netdev_dev *dev_, struct smap *args) { const struct netdev_class *netdev_class = netdev_dev_get_class(dev_); const struct vport_class *vport_class = vport_class_cast(netdev_class); struct netdev_dev_vport *dev = netdev_dev_vport_cast(dev_); const char *name = netdev_dev_get_name(dev_); + int error; + + if (!dev->options) { + struct dpif_linux_vport reply; + struct ofpbuf *buf; + + error = dpif_linux_vport_get(name, &reply, &buf); + if (error) { + VLOG_ERR_RL(&rl, "%s: vport query failed (%s)", + name, strerror(error)); + return error; + } + + dev->options = ofpbuf_clone_data(reply.options, reply.options_len); + ofpbuf_delete(buf); + } + + error = vport_class->unparse_config(name, netdev_class->type, + dev->options->data, + dev->options->size, + args); + if (error) { + VLOG_ERR_RL(&rl, "%s: failed to parse kernel config (%s)", + name, strerror(error)); + } + return error; +} + +static int +netdev_vport_set_config(struct netdev_dev *dev_, const struct smap *args) +{ + const struct netdev_class *netdev_class = netdev_dev_get_class(dev_); + const struct vport_class *vport_class = vport_class_cast(netdev_class); + struct netdev_dev_vport *dev = netdev_dev_vport_cast(dev_); + const char *name = netdev_dev_get_name(dev_); + struct netdev_tunnel_config tnl_cfg; struct ofpbuf *options; int error; options = ofpbuf_new(64); error = vport_class->parse_config(name, netdev_dev_get_type(dev_), - args, options); + args, options, &tnl_cfg); if (!error - && (options->size != dev->options->size + && (!dev->options + || options->size != dev->options->size || memcmp(options->data, dev->options->data, options->size))) { struct dpif_linux_vport vport; dpif_linux_vport_init(&vport); - vport.cmd = ODP_VPORT_CMD_SET; + vport.cmd = OVS_VPORT_CMD_SET; vport.name = name; vport.options = options->data; vport.options_len = options->size; @@ -313,6 +304,7 @@ netdev_vport_set_config(struct netdev_dev *dev_, const struct shash *args) * dpif_port_add()). */ ofpbuf_delete(dev->options); dev->options = options; + dev->tnl_cfg = tnl_cfg; options = NULL; error = 0; } @@ -322,84 +314,51 @@ netdev_vport_set_config(struct netdev_dev *dev_, const struct shash *args) return error; } -static int -netdev_vport_send(struct netdev *netdev, const void *data, size_t size) -{ - struct netdev_dev *dev_ = netdev_get_dev(netdev); - struct netdev_dev_vport *dev = netdev_dev_vport_cast(dev_); - - if (dev->dp_ifindex == -1) { - const char *name = netdev_get_name(netdev); - struct dpif_linux_vport reply; - struct ofpbuf *buf; - int error; - - error = dpif_linux_vport_get(name, &reply, &buf); - if (error) { - VLOG_ERR_RL(&rl, "%s: failed to query vport for send (%s)", - name, strerror(error)); - return error; - } - dev->dp_ifindex = reply.dp_ifindex; - dev->port_no = reply.port_no; - ofpbuf_delete(buf); - } - - return dpif_linux_vport_send(dev->dp_ifindex, dev->port_no, data, size); -} - static int netdev_vport_set_etheraddr(struct netdev *netdev, const uint8_t mac[ETH_ADDR_LEN]) { - struct dpif_linux_vport vport; - int error; - - dpif_linux_vport_init(&vport); - vport.cmd = ODP_VPORT_CMD_SET; - vport.name = netdev_get_name(netdev); - vport.address = mac; - - error = dpif_linux_vport_transact(&vport, NULL, NULL); - if (!error) { - netdev_vport_poll_notify(netdev); - } - return error; + memcpy(netdev_vport_get_dev(netdev)->etheraddr, mac, ETH_ADDR_LEN); + netdev_vport_poll_notify(netdev); + return 0; } static int netdev_vport_get_etheraddr(const struct netdev *netdev, uint8_t mac[ETH_ADDR_LEN]) { - struct dpif_linux_vport reply; - struct ofpbuf *buf; - int error; - - error = dpif_linux_vport_get(netdev_get_name(netdev), &reply, &buf); - if (!error) { - if (reply.address) { - memcpy(mac, reply.address, ETH_ADDR_LEN); - } else { - error = EOPNOTSUPP; - } - ofpbuf_delete(buf); - } - return error; + memcpy(mac, netdev_vport_get_dev(netdev)->etheraddr, ETH_ADDR_LEN); + return 0; } -static int -netdev_vport_get_mtu(const struct netdev *netdev, int *mtup) -{ - struct dpif_linux_vport reply; - struct ofpbuf *buf; - int error; - - error = dpif_linux_vport_get(netdev_get_name(netdev), &reply, &buf); - if (!error) { - *mtup = reply.mtu; - ofpbuf_delete(buf); - } - return error; +/* 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; } int @@ -417,54 +376,28 @@ netdev_vport_get_stats(const struct netdev *netdev, struct netdev_stats *stats) return EOPNOTSUPP; } - netdev_stats_from_rtnl_link_stats64(stats, reply.stats); + netdev_stats_from_ovs_vport_stats(stats, reply.stats); ofpbuf_delete(buf); return 0; } -int -netdev_vport_set_stats(struct netdev *netdev, const struct netdev_stats *stats) -{ - struct rtnl_link_stats64 rtnl_stats; - struct dpif_linux_vport vport; - int err; - - netdev_stats_to_rtnl_link_stats64(&rtnl_stats, stats); - - dpif_linux_vport_init(&vport); - vport.cmd = ODP_VPORT_CMD_SET; - vport.name = netdev_get_name(netdev); - vport.stats = &rtnl_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 int -netdev_vport_get_status(const struct netdev *netdev, struct shash *sh) +tunnel_get_status(const struct netdev *netdev, struct smap *smap) { - const char *iface = netdev_vport_get_tnl_iface(netdev); + static char iface[IFNAMSIZ]; + ovs_be32 route; - if (iface) { + route = netdev_vport_get_dev(netdev)->tnl_cfg.ip_dst; + if (route_table_get_name(route, iface)) { struct netdev *egress_netdev; - shash_add(sh, "tunnel_egress_iface", xstrdup(iface)); + smap_add(smap, "tunnel_egress_iface", iface); - if (!netdev_open_default(iface, &egress_netdev)) { - shash_add(sh, "tunnel_egress_iface_carrier", - xstrdup(netdev_get_carrier(egress_netdev) - ? "up" : "down")); + if (!netdev_open(iface, "system", &egress_netdev)) { + smap_add(smap, "tunnel_egress_iface_carrier", + netdev_get_carrier(egress_netdev) ? "up" : "down"); netdev_close(egress_netdev); } } @@ -485,57 +418,10 @@ netdev_vport_update_flags(struct netdev *netdev OVS_UNUSED, return 0; } -static char * -make_poll_name(const struct netdev *netdev) -{ - return xasprintf("%s:%s", netdev_get_type(netdev), netdev_get_name(netdev)); -} - -static int -netdev_vport_poll_add(struct netdev *netdev, - void (*cb)(struct netdev_notifier *), void *aux, - struct netdev_notifier **notifierp) -{ - char *poll_name = make_poll_name(netdev); - struct netdev_vport_notifier *notifier; - struct list *list; - struct shash_node *shash_node; - - shash_node = shash_find(&netdev_vport_notifiers, poll_name); - if (!shash_node) { - list = xmalloc(sizeof *list); - list_init(list); - shash_node = shash_add(&netdev_vport_notifiers, poll_name, list); - } else { - list = shash_node->data; - } - - notifier = xmalloc(sizeof *notifier); - netdev_notifier_init(¬ifier->notifier, netdev, cb, aux); - list_push_back(list, ¬ifier->list_node); - notifier->shash_node = shash_node; - - *notifierp = ¬ifier->notifier; - free(poll_name); - - return 0; -} - -static void -netdev_vport_poll_remove(struct netdev_notifier *notifier_) +static unsigned int +netdev_vport_change_seq(const struct netdev *netdev) { - struct netdev_vport_notifier *notifier = - CONTAINER_OF(notifier_, struct netdev_vport_notifier, notifier); - - struct list *list; - - list = list_remove(¬ifier->list_node); - if (list_is_empty(list)) { - shash_delete(&netdev_vport_notifiers, notifier->shash_node); - free(list); - } - - free(notifier); + return netdev_vport_get_dev(netdev)->change_seq; } static void @@ -550,144 +436,128 @@ netdev_vport_wait(void) route_table_wait(); } -/* get_tnl_iface() implementation. */ -static const char * -netdev_vport_get_tnl_iface(const struct netdev *netdev) -{ - struct nlattr *a[ODP_TUNNEL_ATTR_MAX + 1]; - uint32_t route; - struct netdev_dev_vport *ndv; - static char name[IFNAMSIZ]; - - ndv = netdev_dev_vport_cast(netdev_get_dev(netdev)); - if (tnl_port_config_from_nlattr(ndv->options->data, ndv->options->size, - a)) { - return NULL; - } - route = nl_attr_get_be32(a[ODP_TUNNEL_ATTR_DST_IPV4]); - - if (route_table_get_name(route, name)) { - return name; - } - - return NULL; -} - /* Helper functions. */ static void netdev_vport_poll_notify(const struct netdev *netdev) { - char *poll_name = make_poll_name(netdev); - struct list *list = shash_find_data(&netdev_vport_notifiers, - poll_name); - - if (list) { - struct netdev_vport_notifier *notifier; + struct netdev_dev_vport *ndv = netdev_vport_get_dev(netdev); - LIST_FOR_EACH (notifier, list_node, list) { - struct netdev_notifier *n = ¬ifier->notifier; - n->cb(n); - } + ndv->change_seq++; + if (!ndv->change_seq) { + ndv->change_seq++; } - - free(poll_name); } /* Code specific to individual vport types. */ -static void -set_key(const struct shash *args, const char *name, uint16_t type, - struct ofpbuf *options) +static ovs_be64 +parse_key(const struct smap *args, const char *name, + bool *present, bool *flow) { const char *s; - s = shash_find_data(args, name); + *present = false; + *flow = false; + + s = smap_get(args, name); if (!s) { - s = shash_find_data(args, "key"); + s = smap_get(args, "key"); if (!s) { - s = "0"; + return 0; } } + *present = true; + if (!strcmp(s, "flow")) { - /* This is the default if no attribute is present. */ + *flow = true; + return 0; } else { - nl_msg_put_be64(options, type, htonll(strtoull(s, NULL, 0))); + return htonll(strtoull(s, NULL, 0)); } } static int parse_tunnel_config(const char *name, const char *type, - const struct shash *args, struct ofpbuf *options) + const struct smap *args, struct ofpbuf *options, + struct netdev_tunnel_config *tnl_cfg_) { - bool is_gre = false; - bool is_ipsec = false; - struct shash_node *node; - bool ipsec_mech_set = false; - ovs_be32 daddr = htonl(0); - uint32_t flags; + bool ipsec_mech_set, needs_dst_port, has_csum; + struct netdev_tunnel_config tnl_cfg; + struct smap_node *node; + uint8_t flags; - flags = TNL_F_DF_DEFAULT | TNL_F_PMTUD | TNL_F_HDR_CACHE; - if (!strcmp(type, "gre")) { - is_gre = true; - } else if (!strcmp(type, "ipsec_gre")) { - is_gre = true; - is_ipsec = true; + flags = TNL_F_DF_DEFAULT; + has_csum = strstr(type, "gre"); + ipsec_mech_set = false; + memset(&tnl_cfg, 0, sizeof tnl_cfg); + + if (!strcmp(type, "capwap")) { + VLOG_WARN_ONCE("CAPWAP tunnel support is deprecated."); + } + + needs_dst_port = !strcmp(type, "vxlan"); + tnl_cfg.ipsec = strstr(type, "ipsec"); + if (tnl_cfg.ipsec) { flags |= TNL_F_IPSEC; - flags &= ~TNL_F_HDR_CACHE; } + tnl_cfg.dont_fragment = true; - SHASH_FOR_EACH (node, args) { - if (!strcmp(node->name, "remote_ip")) { + SMAP_FOR_EACH (node, args) { + if (!strcmp(node->key, "remote_ip")) { struct in_addr in_addr; - if (lookup_ip(node->data, &in_addr)) { + if (lookup_ip(node->value, &in_addr)) { VLOG_WARN("%s: bad %s 'remote_ip'", name, type); } else { - daddr = in_addr.s_addr; + tnl_cfg.ip_dst = in_addr.s_addr; } - } else if (!strcmp(node->name, "local_ip")) { + } else if (!strcmp(node->key, "local_ip")) { struct in_addr in_addr; - if (lookup_ip(node->data, &in_addr)) { + if (lookup_ip(node->value, &in_addr)) { VLOG_WARN("%s: bad %s 'local_ip'", name, type); } else { - nl_msg_put_be32(options, ODP_TUNNEL_ATTR_SRC_IPV4, - in_addr.s_addr); + tnl_cfg.ip_src = in_addr.s_addr; } - } else if (!strcmp(node->name, "tos")) { - if (!strcmp(node->data, "inherit")) { + } else if (!strcmp(node->key, "tos")) { + if (!strcmp(node->value, "inherit")) { flags |= TNL_F_TOS_INHERIT; + tnl_cfg.tos_inherit = true; } else { - nl_msg_put_u8(options, ODP_TUNNEL_ATTR_TOS, atoi(node->data)); + char *endptr; + int tos; + tos = strtol(node->value, &endptr, 0); + if (*endptr == '\0' && tos == (tos & IP_DSCP_MASK)) { + nl_msg_put_u8(options, OVS_TUNNEL_ATTR_TOS, tos); + tnl_cfg.tos = tos; + } else { + VLOG_WARN("%s: invalid TOS %s", name, node->value); + } } - } else if (!strcmp(node->name, "ttl")) { - if (!strcmp(node->data, "inherit")) { + } else if (!strcmp(node->key, "ttl")) { + if (!strcmp(node->value, "inherit")) { flags |= TNL_F_TTL_INHERIT; + tnl_cfg.ttl_inherit = true; } else { - nl_msg_put_u8(options, ODP_TUNNEL_ATTR_TTL, atoi(node->data)); + nl_msg_put_u8(options, OVS_TUNNEL_ATTR_TTL, atoi(node->value)); + tnl_cfg.ttl = atoi(node->value); } - } else if (!strcmp(node->name, "csum") && is_gre) { - if (!strcmp(node->data, "true")) { + } else if (!strcmp(node->key, "dst_port") && needs_dst_port) { + tnl_cfg.dst_port = htons(atoi(node->value)); + nl_msg_put_u16(options, OVS_TUNNEL_ATTR_DST_PORT, + atoi(node->value)); + } else if (!strcmp(node->key, "csum") && has_csum) { + if (!strcmp(node->value, "true")) { flags |= TNL_F_CSUM; + tnl_cfg.csum = true; } - } else if (!strcmp(node->name, "df_inherit")) { - if (!strcmp(node->data, "true")) { - flags |= TNL_F_DF_INHERIT; - } - } else if (!strcmp(node->name, "df_default")) { - if (!strcmp(node->data, "false")) { + } else if (!strcmp(node->key, "df_default")) { + if (!strcmp(node->value, "false")) { flags &= ~TNL_F_DF_DEFAULT; + tnl_cfg.dont_fragment = false; } - } else if (!strcmp(node->name, "pmtud")) { - if (!strcmp(node->data, "false")) { - flags &= ~TNL_F_PMTUD; - } - } else if (!strcmp(node->name, "header_cache")) { - if (!strcmp(node->data, "false")) { - flags &= ~TNL_F_HDR_CACHE; - } - } else if (!strcmp(node->name, "peer_cert") && is_ipsec) { - if (shash_find(args, "certificate")) { + } else if (!strcmp(node->key, "peer_cert") && tnl_cfg.ipsec) { + if (smap_get(args, "certificate")) { ipsec_mech_set = true; } else { const char *use_ssl_cert; @@ -698,7 +568,7 @@ parse_tunnel_config(const char *name, const char *type, * will like be removed when multiple SSL configurations * are supported by OVS. */ - use_ssl_cert = shash_find_data(args, "use_ssl_cert"); + use_ssl_cert = smap_get(args, "use_ssl_cert"); if (!use_ssl_cert || strcmp(use_ssl_cert, "true")) { VLOG_ERR("%s: 'peer_cert' requires 'certificate' argument", name); @@ -706,34 +576,44 @@ parse_tunnel_config(const char *name, const char *type, } ipsec_mech_set = true; } - } else if (!strcmp(node->name, "psk") && is_ipsec) { + } else if (!strcmp(node->key, "psk") && tnl_cfg.ipsec) { ipsec_mech_set = true; - } else if (is_ipsec - && (!strcmp(node->name, "certificate") - || !strcmp(node->name, "private_key") - || !strcmp(node->name, "use_ssl_cert"))) { + } else if (tnl_cfg.ipsec + && (!strcmp(node->key, "certificate") + || !strcmp(node->key, "private_key") + || !strcmp(node->key, "use_ssl_cert"))) { /* Ignore options not used by the netdev. */ - } else if (is_gre && (!strcmp(node->name, "key") || - !strcmp(node->name, "in_key") || - !strcmp(node->name, "out_key"))) { + } else if (!strcmp(node->key, "key") || + !strcmp(node->key, "in_key") || + !strcmp(node->key, "out_key")) { /* Handled separately below. */ } else { - VLOG_WARN("%s: unknown %s argument '%s'", name, type, node->name); + VLOG_WARN("%s: unknown %s argument '%s'", name, type, node->key); } } - if (is_ipsec) { - char *file_name = xasprintf("%s/%s", ovs_rundir(), - "ovs-monitor-ipsec.pid"); - pid_t pid = read_pidfile(file_name); - free(file_name); + /* Add a default destination port for VXLAN if none specified. */ + if (needs_dst_port && !tnl_cfg.dst_port) { + nl_msg_put_u16(options, OVS_TUNNEL_ATTR_DST_PORT, VXLAN_DST_PORT); + tnl_cfg.dst_port = htons(VXLAN_DST_PORT); + } + + if (tnl_cfg.ipsec) { + static pid_t pid = 0; + if (pid <= 0) { + char *file_name = xasprintf("%s/%s", ovs_rundir(), + "ovs-monitor-ipsec.pid"); + pid = read_pidfile(file_name); + free(file_name); + } + if (pid < 0) { VLOG_ERR("%s: IPsec requires the ovs-monitor-ipsec daemon", name); return EINVAL; } - if (shash_find(args, "peer_cert") && shash_find(args, "psk")) { + if (smap_get(args, "peer_cert") && smap_get(args, "psk")) { VLOG_ERR("%s: cannot define both 'peer_cert' and 'psk'", name); return EINVAL; } @@ -745,41 +625,65 @@ parse_tunnel_config(const char *name, const char *type, } } - if (is_gre) { - set_key(args, "in_key", ODP_TUNNEL_ATTR_IN_KEY, options); - set_key(args, "out_key", ODP_TUNNEL_ATTR_OUT_KEY, options); - } - - if (!daddr) { + if (!tnl_cfg.ip_dst) { VLOG_ERR("%s: %s type requires valid 'remote_ip' argument", name, type); return EINVAL; } - nl_msg_put_be32(options, ODP_TUNNEL_ATTR_DST_IPV4, daddr); + nl_msg_put_be32(options, OVS_TUNNEL_ATTR_DST_IPV4, tnl_cfg.ip_dst); + + if (tnl_cfg.ip_src) { + if (ip_is_multicast(tnl_cfg.ip_dst)) { + VLOG_WARN("%s: remote_ip is multicast, ignoring local_ip", name); + tnl_cfg.ip_src = 0; + } else { + nl_msg_put_be32(options, OVS_TUNNEL_ATTR_SRC_IPV4, tnl_cfg.ip_src); + } + } + + if (!tnl_cfg.ttl) { + tnl_cfg.ttl = DEFAULT_TTL; + } + + tnl_cfg.in_key = parse_key(args, "in_key", + &tnl_cfg.in_key_present, + &tnl_cfg.in_key_flow); + if (tnl_cfg.in_key_present && !tnl_cfg.in_key_flow) { + nl_msg_put_be64(options, OVS_TUNNEL_ATTR_IN_KEY, tnl_cfg.in_key); + } - nl_msg_put_u32(options, ODP_TUNNEL_ATTR_FLAGS, flags); + tnl_cfg.out_key = parse_key(args, "out_key", + &tnl_cfg.out_key_present, + &tnl_cfg.out_key_flow); + if (tnl_cfg.out_key_present && !tnl_cfg.out_key_flow) { + nl_msg_put_be64(options, OVS_TUNNEL_ATTR_OUT_KEY, tnl_cfg.out_key); + } + nl_msg_put_u32(options, OVS_TUNNEL_ATTR_FLAGS, flags); + + *tnl_cfg_ = tnl_cfg; return 0; } static int tnl_port_config_from_nlattr(const struct nlattr *options, size_t options_len, - struct nlattr *a[ODP_TUNNEL_ATTR_MAX + 1]) -{ - static const struct nl_policy odp_tunnel_policy[] = { - [ODP_TUNNEL_ATTR_FLAGS] = { .type = NL_A_U32 }, - [ODP_TUNNEL_ATTR_DST_IPV4] = { .type = NL_A_BE32 }, - [ODP_TUNNEL_ATTR_SRC_IPV4] = { .type = NL_A_BE32, .optional = true }, - [ODP_TUNNEL_ATTR_IN_KEY] = { .type = NL_A_BE64, .optional = true }, - [ODP_TUNNEL_ATTR_OUT_KEY] = { .type = NL_A_BE64, .optional = true }, - [ODP_TUNNEL_ATTR_TOS] = { .type = NL_A_U8, .optional = true }, - [ODP_TUNNEL_ATTR_TTL] = { .type = NL_A_U8, .optional = true }, + struct nlattr *a[OVS_TUNNEL_ATTR_MAX + 1]) +{ + static const struct nl_policy ovs_tunnel_policy[] = { + [OVS_TUNNEL_ATTR_FLAGS] = { .type = NL_A_U32, .optional = true }, + [OVS_TUNNEL_ATTR_DST_IPV4] = { .type = NL_A_BE32, .optional = true }, + [OVS_TUNNEL_ATTR_SRC_IPV4] = { .type = NL_A_BE32, .optional = true }, + [OVS_TUNNEL_ATTR_IN_KEY] = { .type = NL_A_BE64, .optional = true }, + [OVS_TUNNEL_ATTR_OUT_KEY] = { .type = NL_A_BE64, .optional = true }, + [OVS_TUNNEL_ATTR_TOS] = { .type = NL_A_U8, .optional = true }, + [OVS_TUNNEL_ATTR_TTL] = { .type = NL_A_U8, .optional = true }, + [OVS_TUNNEL_ATTR_DST_PORT] = { .type = NL_A_U16, .optional = true }, }; struct ofpbuf buf; ofpbuf_use_const(&buf, options, options_len); - if (!nl_policy_parse(&buf, 0, odp_tunnel_policy, - a, ARRAY_SIZE(odp_tunnel_policy))) { + if (!nl_policy_parse(&buf, 0, ovs_tunnel_policy, + a, ARRAY_SIZE(ovs_tunnel_policy))) { return EINVAL; } return 0; @@ -794,10 +698,9 @@ get_be64_or_zero(const struct nlattr *a) static int unparse_tunnel_config(const char *name OVS_UNUSED, const char *type OVS_UNUSED, const struct nlattr *options, size_t options_len, - struct shash *args) + struct smap *args) { - struct nlattr *a[ODP_TUNNEL_ATTR_MAX + 1]; - ovs_be32 daddr; + struct nlattr *a[OVS_TUNNEL_ATTR_MAX + 1]; uint32_t flags; int error; @@ -806,86 +709,93 @@ unparse_tunnel_config(const char *name OVS_UNUSED, const char *type OVS_UNUSED, return error; } - flags = nl_attr_get_u32(a[ODP_TUNNEL_ATTR_FLAGS]); - if (!(flags & TNL_F_HDR_CACHE) == !(flags & TNL_F_IPSEC)) { - smap_add(args, "header_cache", - flags & TNL_F_HDR_CACHE ? "true" : "false"); + if (a[OVS_TUNNEL_ATTR_DST_IPV4]) { + ovs_be32 daddr = nl_attr_get_be32(a[OVS_TUNNEL_ATTR_DST_IPV4]); + smap_add_format(args, "remote_ip", IP_FMT, IP_ARGS(daddr)); } - daddr = nl_attr_get_be32(a[ODP_TUNNEL_ATTR_DST_IPV4]); - shash_add(args, "remote_ip", xasprintf(IP_FMT, IP_ARGS(&daddr))); - - if (a[ODP_TUNNEL_ATTR_SRC_IPV4]) { - ovs_be32 saddr = nl_attr_get_be32(a[ODP_TUNNEL_ATTR_SRC_IPV4]); - shash_add(args, "local_ip", xasprintf(IP_FMT, IP_ARGS(&saddr))); + if (a[OVS_TUNNEL_ATTR_SRC_IPV4]) { + ovs_be32 saddr = nl_attr_get_be32(a[OVS_TUNNEL_ATTR_SRC_IPV4]); + smap_add_format(args, "local_ip", IP_FMT, IP_ARGS(saddr)); } - if (!a[ODP_TUNNEL_ATTR_IN_KEY] && !a[ODP_TUNNEL_ATTR_OUT_KEY]) { + if (!a[OVS_TUNNEL_ATTR_IN_KEY] && !a[OVS_TUNNEL_ATTR_OUT_KEY]) { smap_add(args, "key", "flow"); } else { - uint64_t in_key = get_be64_or_zero(a[ODP_TUNNEL_ATTR_IN_KEY]); - uint64_t out_key = get_be64_or_zero(a[ODP_TUNNEL_ATTR_OUT_KEY]); + uint64_t in_key = get_be64_or_zero(a[OVS_TUNNEL_ATTR_IN_KEY]); + uint64_t out_key = get_be64_or_zero(a[OVS_TUNNEL_ATTR_OUT_KEY]); if (in_key && in_key == out_key) { - shash_add(args, "key", xasprintf("%"PRIu64, in_key)); + smap_add_format(args, "key", "%"PRIu64, in_key); } else { - if (!a[ODP_TUNNEL_ATTR_IN_KEY]) { + if (!a[OVS_TUNNEL_ATTR_IN_KEY]) { smap_add(args, "in_key", "flow"); } else if (in_key) { - shash_add(args, "in_key", xasprintf("%"PRIu64, in_key)); + smap_add_format(args, "in_key", "%"PRIu64, in_key); } - if (!a[ODP_TUNNEL_ATTR_OUT_KEY]) { + if (!a[OVS_TUNNEL_ATTR_OUT_KEY]) { smap_add(args, "out_key", "flow"); } else if (out_key) { - shash_add(args, "out_key", xasprintf("%"PRIu64, out_key)); + smap_add_format(args, "out_key", "%"PRIu64, out_key); } } } + flags = get_u32_or_zero(a[OVS_TUNNEL_ATTR_FLAGS]); + if (flags & TNL_F_TTL_INHERIT) { - smap_add(args, "tos", "inherit"); - } else if (a[ODP_TUNNEL_ATTR_TTL]) { - int ttl = nl_attr_get_u8(a[ODP_TUNNEL_ATTR_TTL]); - shash_add(args, "tos", xasprintf("%d", ttl)); + smap_add(args, "ttl", "inherit"); + } else if (a[OVS_TUNNEL_ATTR_TTL]) { + int ttl = nl_attr_get_u8(a[OVS_TUNNEL_ATTR_TTL]); + smap_add_format(args, "ttl", "%d", ttl); } if (flags & TNL_F_TOS_INHERIT) { smap_add(args, "tos", "inherit"); - } else if (a[ODP_TUNNEL_ATTR_TOS]) { - int tos = nl_attr_get_u8(a[ODP_TUNNEL_ATTR_TOS]); - shash_add(args, "tos", xasprintf("%d", tos)); + } else if (a[OVS_TUNNEL_ATTR_TOS]) { + int tos = nl_attr_get_u8(a[OVS_TUNNEL_ATTR_TOS]); + smap_add_format(args, "tos", "0x%x", tos); + } + + if (a[OVS_TUNNEL_ATTR_DST_PORT]) { + uint16_t dst_port = nl_attr_get_u16(a[OVS_TUNNEL_ATTR_DST_PORT]); + if (dst_port != VXLAN_DST_PORT) { + smap_add_format(args, "dst_port", "%d", dst_port); + } } if (flags & TNL_F_CSUM) { smap_add(args, "csum", "true"); } if (flags & TNL_F_DF_INHERIT) { + /* Shouldn't happen as "df_inherit" is no longer supported. However, + * for completeness we report it if it's there. */ smap_add(args, "df_inherit", "true"); } if (!(flags & TNL_F_DF_DEFAULT)) { smap_add(args, "df_default", "false"); } - if (!(flags & TNL_F_PMTUD)) { - smap_add(args, "pmtud", "false"); - } return 0; } static int parse_patch_config(const char *name, const char *type OVS_UNUSED, - const struct shash *args, struct ofpbuf *options) + const struct smap *args, struct ofpbuf *options, + struct netdev_tunnel_config *tnl_cfg) { const char *peer; - peer = shash_find_data(args, "peer"); + memset(tnl_cfg, 0, sizeof *tnl_cfg); + + peer = smap_get(args, "peer"); if (!peer) { VLOG_ERR("%s: patch type requires valid 'peer' argument", name); return EINVAL; } - if (shash_count(args) > 1) { + if (smap_count(args) > 1) { VLOG_ERR("%s: patch type takes only a 'peer' argument", name); return EINVAL; } @@ -900,7 +810,7 @@ parse_patch_config(const char *name, const char *type OVS_UNUSED, return EINVAL; } - nl_msg_put_string(options, ODP_PATCH_ATTR_PEER, peer); + nl_msg_put_string(options, OVS_PATCH_ATTR_PEER, peer); return 0; } @@ -908,60 +818,62 @@ parse_patch_config(const char *name, const char *type OVS_UNUSED, static int unparse_patch_config(const char *name OVS_UNUSED, const char *type OVS_UNUSED, const struct nlattr *options, size_t options_len, - struct shash *args) + struct smap *args) { - static const struct nl_policy odp_patch_policy[] = { - [ODP_PATCH_ATTR_PEER] = { .type = NL_A_STRING, + static const struct nl_policy ovs_patch_policy[] = { + [OVS_PATCH_ATTR_PEER] = { .type = NL_A_STRING, .max_len = IFNAMSIZ, .optional = false } }; - struct nlattr *a[ARRAY_SIZE(odp_patch_policy)]; + struct nlattr *a[ARRAY_SIZE(ovs_patch_policy)]; struct ofpbuf buf; ofpbuf_use_const(&buf, options, options_len); - if (!nl_policy_parse(&buf, 0, odp_patch_policy, - a, ARRAY_SIZE(odp_patch_policy))) { + if (!nl_policy_parse(&buf, 0, ovs_patch_policy, + a, ARRAY_SIZE(ovs_patch_policy))) { return EINVAL; } - smap_add(args, "peer", nl_attr_get_string(a[ODP_PATCH_ATTR_PEER])); + smap_add(args, "peer", nl_attr_get_string(a[OVS_PATCH_ATTR_PEER])); return 0; } -#define VPORT_FUNCTIONS(GET_STATUS) \ +#define VPORT_FUNCTIONS(GET_TUNNEL_CONFIG, GET_STATUS) \ NULL, \ netdev_vport_run, \ netdev_vport_wait, \ \ netdev_vport_create, \ netdev_vport_destroy, \ + netdev_vport_get_config, \ netdev_vport_set_config, \ + GET_TUNNEL_CONFIG, \ \ netdev_vport_open, \ netdev_vport_close, \ \ - NULL, /* enumerate */ \ - \ + NULL, /* listen */ \ NULL, /* recv */ \ NULL, /* recv_wait */ \ NULL, /* drain */ \ \ - netdev_vport_send, /* send */ \ + NULL, /* send */ \ NULL, /* send_wait */ \ \ netdev_vport_set_etheraddr, \ netdev_vport_get_etheraddr, \ - netdev_vport_get_mtu, \ + NULL, /* get_mtu */ \ + NULL, /* set_mtu */ \ NULL, /* get_ifindex */ \ NULL, /* get_carrier */ \ + NULL, /* get_carrier_resets */ \ NULL, /* get_miimon */ \ netdev_vport_get_stats, \ - netdev_vport_set_stats, \ + NULL, /* set_stats */ \ \ NULL, /* get_features */ \ NULL, /* set_advertisements */ \ - NULL, /* get_vlan_vid */ \ \ NULL, /* set_policing */ \ NULL, /* get_qos_types */ \ @@ -985,27 +897,27 @@ unparse_patch_config(const char *name OVS_UNUSED, const char *type OVS_UNUSED, \ netdev_vport_update_flags, \ \ - netdev_vport_poll_add, \ - netdev_vport_poll_remove, + netdev_vport_change_seq + +#define TUNNEL_CLASS(NAME, VPORT_TYPE) \ + { VPORT_TYPE, \ + { NAME, VPORT_FUNCTIONS(get_netdev_tunnel_config, \ + tunnel_get_status) }, \ + parse_tunnel_config, unparse_tunnel_config } void netdev_vport_register(void) { static const struct vport_class vport_classes[] = { - { ODP_VPORT_TYPE_GRE, - { "gre", VPORT_FUNCTIONS(netdev_vport_get_status) }, - parse_tunnel_config, unparse_tunnel_config }, - - { ODP_VPORT_TYPE_GRE, - { "ipsec_gre", VPORT_FUNCTIONS(netdev_vport_get_status) }, - parse_tunnel_config, unparse_tunnel_config }, - - { ODP_VPORT_TYPE_CAPWAP, - { "capwap", VPORT_FUNCTIONS(netdev_vport_get_status) }, - parse_tunnel_config, unparse_tunnel_config }, - - { ODP_VPORT_TYPE_PATCH, - { "patch", VPORT_FUNCTIONS(NULL) }, + TUNNEL_CLASS("gre", OVS_VPORT_TYPE_GRE), + TUNNEL_CLASS("ipsec_gre", OVS_VPORT_TYPE_GRE), + TUNNEL_CLASS("gre64", OVS_VPORT_TYPE_GRE64), + TUNNEL_CLASS("ipsec_gre64", OVS_VPORT_TYPE_GRE64), + TUNNEL_CLASS("capwap", OVS_VPORT_TYPE_CAPWAP), + TUNNEL_CLASS("vxlan", OVS_VPORT_TYPE_VXLAN), + + { OVS_VPORT_TYPE_PATCH, + { "patch", VPORT_FUNCTIONS(NULL, NULL) }, parse_patch_config, unparse_patch_config } };