resolved: transaction - exponentially increase retry timeouts

Rather than fixing this to 5s for unicast DNS and 1s for LLMNR, start
at a tenth of those values and increase exponentially until the old
values are reached. For LLMNR the recommended timeout for IEEE802
networks (which basically means all of the ones we care about) is 100ms,
so that should be uncontroversial. For unicast DNS I have found no
recommended value. However, it seems vastly more likely that hitting a
500ms timeout is casued by a packet loss, rather than the RTT genuinely
being greater than 500ms, so taking this as a startnig value seems
reasonable to me.

In the common case this greatly reduces the latency due to normal packet
loss. Moreover, once we get support for probing for features, this means
that we can send more packets before degrading the feature level whilst
still allowing us to settle on the correct feature level in a reasonable
timeframe.

The timeouts are tracked per server (or per scope for the multicast
protocols), and once a server (or scope) receives a successfull package
the timeout is reset. We also track the largest RTT for the given
server/scope, and always start our timouts at twice the largest
observed RTT.

1 files changed, 56 insertions, 4 deletions

diff --git a/src/resolve/resolved-dns-transaction.c b/src/resolve/resolved-dns-transaction.c
index 8a93b265c6..487b2c5162 100644
--- a/src/resolve/resolved-dns-transaction.c
+++ b/src/resolve/resolved-dns-transaction.c
@@ -319,11 +319,14 @@ static void dns_transaction_next_dns_server(DnsTransaction *t) {
 }
 
 void dns_transaction_process_reply(DnsTransaction *t, DnsPacket *p) {
+        usec_t ts;
         int r;
 
         assert(t);
         assert(p);
         assert(t->state == DNS_TRANSACTION_PENDING);
+        assert(t->scope);
+        assert(t->scope->manager);
 
         /* Note that this call might invalidate the query. Callers
          * should hence not attempt to access the query or transaction
@@ -369,6 +372,26 @@ void dns_transaction_process_reply(DnsTransaction *t, DnsPacket *p) {
                 }
         }
 
+        r = sd_event_now(t->scope->manager->event, clock_boottime_or_monotonic(), &ts);
+        if (r < 0)
+                ts = now(clock_boottime_or_monotonic());
+
+        switch (t->scope->protocol) {
+        case DNS_PROTOCOL_DNS:
+                assert(t->server);
+
+                dns_server_packet_received(t->server, ts - t->start_usec);
+
+                break;
+        case DNS_PROTOCOL_LLMNR:
+        case DNS_PROTOCOL_MDNS:
+                dns_scope_packet_received(t->scope, ts - t->start_usec);
+
+                break;
+        default:
+                assert_not_reached("Invalid DNS protocol.");
+        }
+
         if (DNS_PACKET_TC(p)) {
                 /* Response was truncated, let's try again with good old TCP */
                 r = dns_transaction_open_tcp(t);
@@ -434,9 +457,9 @@ static int on_dns_packet(sd_event_source *s, int fd, uint32_t revents, void *use
                 return r;
 
         if (dns_packet_validate_reply(p) > 0 &&
-            DNS_PACKET_ID(p) == t->id) {
+            DNS_PACKET_ID(p) == t->id)
                 dns_transaction_process_reply(t, p);
-        } else
+        else
                 log_debug("Invalid DNS packet.");
 
         return 0;
@@ -481,6 +504,12 @@ static int on_transaction_timeout(sd_event_source *s, usec_t usec, void *userdat
         /* Timeout reached? Try again, with a new server */
         dns_transaction_next_dns_server(t);
 
+        /* ... and possibly increased timeout */
+        if (t->server)
+                dns_server_packet_lost(t->server, usec - t->start_usec);
+        else
+                dns_scope_packet_lost(t->scope, usec - t->start_usec);
+
         r = dns_transaction_go(t);
         if (r < 0)
                 dns_transaction_complete(t, DNS_TRANSACTION_RESOURCES);
@@ -528,8 +557,26 @@ static int dns_transaction_make_packet(DnsTransaction *t) {
         return 0;
 }
 
+static usec_t transaction_get_resend_timeout(DnsTransaction *t) {
+        assert(t);
+        assert(t->scope);
+
+        switch (t->scope->protocol) {
+        case DNS_PROTOCOL_DNS:
+                assert(t->server);
+
+                return t->server->resend_timeout;
+        case DNS_PROTOCOL_LLMNR:
+        case DNS_PROTOCOL_MDNS:
+                return t->scope->resend_timeout;
+        default:
+                assert_not_reached("Invalid DNS protocol.");
+        }
+}
+
 int dns_transaction_go(DnsTransaction *t) {
         bool had_stream;
+        usec_t ts;
         int r;
 
         assert(t);
@@ -555,7 +602,12 @@ int dns_transaction_go(DnsTransaction *t) {
                 return 0;
         }
 
+        r = sd_event_now(t->scope->manager->event, clock_boottime_or_monotonic(), &ts);
+        if (r < 0)
+                ts = now(clock_boottime_or_monotonic());
+
         t->n_attempts++;
+        t->start_usec = ts;
         t->received = dns_packet_unref(t->received);
         t->cached = dns_answer_unref(t->cached);
         t->cached_rcode = 0;
@@ -600,7 +652,7 @@ int dns_transaction_go(DnsTransaction *t) {
                                 t->scope->manager->event,
                                 &t->timeout_event_source,
                                 clock_boottime_or_monotonic(),
-                                now(clock_boottime_or_monotonic()) + jitter,
+                                ts + jitter,
                                 LLMNR_JITTER_INTERVAL_USEC,
                                 on_transaction_timeout, t);
                 if (r < 0)
@@ -660,7 +712,7 @@ int dns_transaction_go(DnsTransaction *t) {
                         t->scope->manager->event,
                         &t->timeout_event_source,
                         clock_boottime_or_monotonic(),
-                        now(clock_boottime_or_monotonic()) + TRANSACTION_TIMEOUT_USEC(t->scope->protocol), 0,
+                        ts + transaction_get_resend_timeout(t), 0,
                         on_transaction_timeout, t);
         if (r < 0)
                 return r;