From 864cf1812819836284d12030ce553ee743ca10f0 Mon Sep 17 00:00:00 2001 From: smh Date: Tue, 15 Dec 2015 16:02:11 +0000 Subject: Fix lagg failover due to missing notifications When using lagg failover mode neither Gratuitous ARP (IPv4) or Unsolicited Neighbour Advertisements (IPv6) are sent to notify other nodes that the address may have moved. This results is slow failover, dropped packets and network outages for the lagg interface when the primary link goes down. We now use the new if_link_state_change_cond with the force param set to allow lagg to force through link state changes and hence fire a ifnet_link_event which are now monitored by rip and nd6. Upon receiving these events each protocol trigger the relevant notifications: * inet4 => Gratuitous ARP * inet6 => Unsolicited Neighbour Announce This also fixes the carp IPv6 NA's that stopped working after r251584 which added the ipv6_route__llma route. The new behavour can be controlled using the sysctls: * net.link.ether.inet.arp_on_link * net.inet6.icmp6.nd6_on_link Also removed unused param from lagg_port_state and added descriptions for the sysctls while here. PR: 156226 MFC after: 1 month Sponsored by: Multiplay Differential Revision: https://reviews.freebsd.org/D4111 --- sys/netinet/if_ether.c | 93 ++++++++++++++++++++++++++++++++++++++++++-------- 1 file changed, 78 insertions(+), 15 deletions(-) (limited to 'sys/netinet/if_ether.c') diff --git a/sys/netinet/if_ether.c b/sys/netinet/if_ether.c index 72c2a60..2214542 100644 --- a/sys/netinet/if_ether.c +++ b/sys/netinet/if_ether.c @@ -107,6 +107,7 @@ VNET_PCPUSTAT_SYSUNINIT(arpstat); #endif /* VIMAGE */ static VNET_DEFINE(int, arp_maxhold) = 1; +static VNET_DEFINE(int, arp_on_link) = 1; #define V_arpt_keep VNET(arpt_keep) #define V_arpt_down VNET(arpt_down) @@ -114,6 +115,7 @@ static VNET_DEFINE(int, arp_maxhold) = 1; #define V_arp_maxtries VNET(arp_maxtries) #define V_arp_proxyall VNET(arp_proxyall) #define V_arp_maxhold VNET(arp_maxhold) +#define V_arp_on_link VNET(arp_on_link) SYSCTL_INT(_net_link_ether_inet, OID_AUTO, max_age, CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(arpt_keep), 0, @@ -136,6 +138,9 @@ SYSCTL_INT(_net_link_ether_inet, OID_AUTO, max_log_per_second, CTLFLAG_RW, &arp_maxpps, 0, "Maximum number of remotely triggered ARP messages that can be " "logged per second"); +SYSCTL_INT(_net_link_ether_inet, OID_AUTO, arp_on_link, CTLFLAG_VNET | CTLFLAG_RW, + &VNET_NAME(arp_on_link), 0, + "Send gratuitous ARP's on interface link up events"); #define ARP_LOG(pri, ...) do { \ if (ppsratecheck(&arp_lastlog, &arp_curpps, arp_maxpps)) \ @@ -156,6 +161,7 @@ static void arp_mark_lle_reachable(struct llentry *la); static void arp_iflladdr(void *arg __unused, struct ifnet *ifp); static eventhandler_tag iflladdr_tag; +static eventhandler_tag ifnet_link_event_tag; static const struct netisr_handler arp_nh = { .nh_name = "arp", @@ -1176,43 +1182,96 @@ arp_ifinit(struct ifnet *ifp, struct ifaddr *ifa) if (ntohl(dst_in->sin_addr.s_addr) == INADDR_ANY) return; - arp_announce_ifaddr(ifp, dst_in->sin_addr, IF_LLADDR(ifp)); + arp_announce_addr(ifp, &dst_in->sin_addr, IF_LLADDR(ifp)); arp_add_ifa_lle(ifp, dst); } -void -arp_announce_ifaddr(struct ifnet *ifp, struct in_addr addr, u_char *enaddr) +void __noinline +arp_announce_addr(struct ifnet *ifp, const struct in_addr *addr, u_char *enaddr) { - if (ntohl(addr.s_addr) != INADDR_ANY) - arprequest(ifp, &addr, &addr, enaddr); + if (ntohl(addr->s_addr) != INADDR_ANY) + arprequest(ifp, addr, addr, enaddr); } /* - * Sends gratuitous ARPs for each ifaddr to notify other - * nodes about the address change. + * Send gratuitous ARPs for all interfaces addresses to notify other nodes of + * changes. + * + * This is a noop if the interface isn't up or has been flagged for no ARP. */ -static __noinline void -arp_handle_ifllchange(struct ifnet *ifp) +void __noinline +arp_announce(struct ifnet *ifp) { + int i, cnt, entries; + u_char *lladdr; struct ifaddr *ifa; + struct in_addr *addr, *head; + + if (!(ifp->if_flags & IFF_UP) || (ifp->if_flags & IFF_NOARP)) + return; + + entries = 8; + cnt = 0; + head = malloc(sizeof(*addr) * entries, M_TEMP, M_NOWAIT); + if (head == NULL) { + log(LOG_INFO, "arp_announce: malloc %d entries failed\n", + entries); + return; + } + /* Take a copy then process to avoid locking issues. */ + IF_ADDR_RLOCK(ifp); TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { - if (ifa->ifa_addr->sa_family == AF_INET) - arp_ifinit(ifp, ifa); + if (ifa->ifa_addr->sa_family != AF_INET) + continue; + + if (cnt == entries) { + addr = (struct in_addr *)realloc(head, sizeof(*addr) * + (entries + 8), M_TEMP, M_NOWAIT); + if (addr == NULL) { + log(LOG_INFO, "arp_announce: realloc to %d " + "entries failed\n", entries + 8); + /* Process what we have. */ + break; + } + entries += 8; + head = addr; + } + + addr = head + cnt; + bcopy(IFA_IN(ifa), addr, sizeof(*addr)); + cnt++; } + IF_ADDR_RUNLOCK(ifp); + + lladdr = IF_LLADDR(ifp); + for (i = 0; i < cnt; i++) { + arp_announce_addr(ifp, head + i, lladdr); + } + free(head, M_TEMP); +} + +/* + * A handler for interface linkstate change events. + */ +static void +arp_ifnet_link_event(void *arg __unused, struct ifnet *ifp, int linkstate) +{ + + if (linkstate == LINK_STATE_UP && V_arp_on_link) + arp_announce(ifp); } /* - * A handler for interface link layer address change event. + * A handler for interface link layer address change events. */ static __noinline void arp_iflladdr(void *arg __unused, struct ifnet *ifp) { - if ((ifp->if_flags & IFF_UP) != 0) - arp_handle_ifllchange(ifp); + arp_announce(ifp); } static void @@ -1220,8 +1279,12 @@ arp_init(void) { netisr_register(&arp_nh); - if (IS_DEFAULT_VNET(curvnet)) + + if (IS_DEFAULT_VNET(curvnet)) { iflladdr_tag = EVENTHANDLER_REGISTER(iflladdr_event, arp_iflladdr, NULL, EVENTHANDLER_PRI_ANY); + ifnet_link_event_tag = EVENTHANDLER_REGISTER(ifnet_link_event, + arp_ifnet_link_event, 0, EVENTHANDLER_PRI_ANY); + } } SYSINIT(arp, SI_SUB_PROTO_DOMAIN, SI_ORDER_ANY, arp_init, 0); -- cgit v1.1