From 7f08d4b6fb5aa38f47fcb8612ea6d4de7eaa324e Mon Sep 17 00:00:00 2001
From: "Nicholas J. Kain" <nicholas@kain.us>
Date: Fri, 24 Feb 2017 07:39:14 -0500
Subject: [PATCH] arp: Handle initial gateway query asynchronously and retry
 failures.

The gateway/router MAC fingerprinting could perhaps be done more
robustly in the face of suspend or carrier loss, but the time window
in which things could get confused is very small and I would rather
just rely on supervisor respawn in that case.

Even this case I don't think I've ever seen.
---
 src/arp.c   | 18 +++++++++++++++++-
 src/arp.h   |  2 ++
 src/state.c |  7 ++++---
 3 files changed, 23 insertions(+), 4 deletions(-)

diff --git a/src/arp.c b/src/arp.c
index 30aa73d..3d1623d 100644
--- a/src/arp.c
+++ b/src/arp.c
@@ -63,7 +63,7 @@ int arp_probe_max = 2000;          // maximum delay until repeated probe (ms)
 #define DEFEND_INTERVAL 10000      // minimum interval between defensive ARPs
 
 static struct arp_data garp = {
-    .wake_ts = { -1, -1, -1, -1, -1, -1 },
+    .wake_ts = { -1, -1, -1, -1, -1, -1, -1 },
     .send_stats = {{0,0},{0,0},{0,0}},
     .last_conflict_ts = 0,
     .gw_check_initpings = 0,
@@ -555,14 +555,30 @@ int arp_collision_timeout(struct client_state_t cs[static 1], long long nowts)
 
 int arp_query_gateway(struct client_state_t cs[static 1])
 {
+    if (cs->sent_gw_query) {
+        garp.wake_ts[AS_QUERY_GW_SEND] = -1;
+        return ARPR_OK;
+    }
     if (arp_get_gw_hwaddr(cs) < 0) {
         log_warning("%s: (%s) Failed to send request to get gateway and agent hardware addresses: %s",
                     client_config.interface, __func__, strerror(errno));
+        garp.wake_ts[AS_QUERY_GW_SEND] = curms() + ARP_RETRANS_DELAY;
         return ARPR_FAIL;
     }
+    cs->sent_gw_query = true;
+    garp.wake_ts[AS_QUERY_GW_SEND] = -1;
     return ARPR_OK;
 }
 
+// 1 == not yet time, 0 == timed out, success, -1 == timed out, failure
+int arp_query_gateway_timeout(struct client_state_t cs[static 1], long long nowts)
+{
+    long long rtts = garp.wake_ts[AS_QUERY_GW_SEND];
+    if (rtts == -1) return 0;
+    if (nowts < rtts) return 1;
+    return arp_query_gateway(cs) == ARPR_OK ? 0 : -1;
+}
+
 int arp_announce(struct client_state_t cs[static 1])
 {
     if (cs->sent_first_announce && cs->sent_second_announce) {
diff --git a/src/arp.h b/src/arp.h
index 2279f55..409c3dc 100644
--- a/src/arp.h
+++ b/src/arp.h
@@ -63,6 +63,7 @@ typedef enum {
     AS_ANNOUNCE,        // Announcing our MAC/IP mapping to ethernet peers.
     AS_COLLISION_CHECK, // Checking to see if another host has our IP before
                         // accepting a new lease.
+    AS_QUERY_GW_SEND,   // Sending arp_ping to query the gateway MAC.
     AS_GW_CHECK,        // Seeing if the default GW still exists on the local
                         // segment after the hardware link was lost.
     AS_GW_QUERY,        // Finding the default GW MAC address.
@@ -120,6 +121,7 @@ int arp_do_collision_check(struct client_state_t cs[static 1]);
 int arp_collision_timeout(struct client_state_t cs[static 1], long long nowts);
 
 int arp_query_gateway(struct client_state_t cs[static 1]);
+int arp_query_gateway_timeout(struct client_state_t cs[static 1], long long nowts);
 
 int arp_announce(struct client_state_t cs[static 1]);
 int arp_announce_timeout(struct client_state_t cs[static 1], long long nowts);
diff --git a/src/state.c b/src/state.c
index 558039e..a32754d 100644
--- a/src/state.c
+++ b/src/state.c
@@ -537,8 +537,7 @@ int dhcp_handle(struct client_state_t cs[static 1], long long nowts,
         if (arp_timeout) {
             int r = arp_collision_timeout(cs, nowts);
             if (r == ARPR_FREE) {
-                if (arp_query_gateway(cs) == ARPR_OK)
-                    cs->sent_gw_query = true; // XXX: Handle the false case
+                arp_query_gateway(cs);
                 arp_announce(cs);
                 break;
             } else if (r == ARPR_OK) {
@@ -647,7 +646,9 @@ int dhcp_handle(struct client_state_t cs[static 1], long long nowts,
                 arp_defense_timeout(cs, nowts);
             else
                 arp_announce_timeout(cs, nowts);
-            if (!cs->got_router_arp || !cs->got_server_arp) {
+            if (!cs->sent_gw_query)
+                arp_query_gateway_timeout(cs, nowts);
+            else if (!cs->got_router_arp || !cs->got_server_arp) {
                 int r = arp_gw_query_timeout(cs, nowts);
                 if (r == ARPR_OK) {
                 } else if (r == ARPR_FAIL) {