quic: ack trace probes and timeout margin

firedancer-io · Feb 11, 2025 · 918742b · 918742b
1 parent 9402484
commit 918742b
Show file tree

Hide file tree

Showing 4 changed files with 9 additions and 4 deletions.
diff --git a/src/app/fdctl/config/default.toml b/src/app/fdctl/config/default.toml
@@ -945,7 +945,7 @@ dynamic_port_range = "8900-9000"
         idle_timeout_millis = 10000
 
         # Max delay for outgoing ACKs.
-        ack_delay_millis = 50
+        ack_delay_millis = 25
 
         # QUIC retry is a feature to combat new connection request
         # spamming.  See rfc9000 8.1.2 for more details.  This flag

diff --git a/src/waltz/quic/fd_quic.c b/src/waltz/quic/fd_quic.c
@@ -491,7 +491,7 @@ fd_quic_init( fd_quic_t * quic ) {
   }
   state->svc_delay[ FD_QUIC_SVC_INSTANT ] = 0UL;
   state->svc_delay[ FD_QUIC_SVC_ACK_TX  ] = quic->config.ack_delay;
-  state->svc_delay[ FD_QUIC_SVC_WAIT    ] = quic->config.idle_timeout;
+  state->svc_delay[ FD_QUIC_SVC_WAIT    ] = quic->config.idle_timeout * 10;
 
   /* Check TX AIO */
 
@@ -2706,7 +2706,8 @@ fd_quic_tls_cb_peer_params( void *        context,
 
   /* set the max_idle_timeout to the min of our and peer max_idle_timeout */
   if( peer_tp->max_idle_timeout ) {
-    conn->idle_timeout = fd_ulong_min( (ulong)(1e6) * peer_tp->max_idle_timeout, conn->idle_timeout );
+    /* 10 for margin of safety */
+    conn->idle_timeout = 10*fd_ulong_min( (ulong)(1e6) * peer_tp->max_idle_timeout, conn->idle_timeout );
   }
 
   /* set ack_delay_exponent so we can properly interpret peer's ack_delays

diff --git a/src/waltz/quic/fd_quic_ack_tx.c b/src/waltz/quic/fd_quic_ack_tx.c
@@ -55,11 +55,13 @@ fd_quic_ack_pkt( fd_quic_ack_gen_t * gen,
 
   /* Attempt to allocate another ACK queue entry */
   if( gen->head - gen->tail >= FD_QUIC_ACK_QUEUE_CNT ) {
+    FD_DTRACE_PROBE_3( fd_quic_ack_tx_overflow, enc_level, pkt_number, now );
     FD_DEBUG( FD_LOG_DEBUG(( "ACK queue overflow! (excessive reordering)" )); )
     return FD_QUIC_ACK_TX_ENOSPC;
   }
 
   /* Start new pending ACK */
+  FD_DTRACE_PROBE_3( fd_quic_ack_range_new, enc_level, pkt_number, now );
   FD_ACK_DEBUG( FD_LOG_DEBUG(( "gen=%p queue ACK for enc=%u pkt_num=%lu seq=%u",
     (void *)gen, enc_level, pkt_number, gen->head )); )
   fd_quic_ack_t * next_ack = fd_quic_ack_queue_ele( gen, gen->head );
@@ -105,6 +107,7 @@ fd_quic_gen_ack_frames( fd_quic_ack_gen_t * gen,
   for( ; gen->tail != gen->head; gen->tail++ ) {
     fd_quic_ack_t * ack = fd_quic_ack_queue_ele( gen, gen->tail );
     if( ack->enc_level != enc_level ) {
+      FD_DTRACE_PROBE_2( fd_quic_ack_tx_enc_level_break, ack->enc_level, enc_level );
       FD_ACK_DEBUG( FD_LOG_DEBUG(( "need encryption level %u for ACKs but have %u", ack->enc_level, enc_level )); )
       break;
     }
@@ -134,6 +137,7 @@ fd_quic_gen_ack_frames( fd_quic_ack_gen_t * gen,
   if( gen->head == gen->tail ) {
     gen->is_elicited = 0;
   } else {
+    FD_DTRACE_PROBE_1( fd_quic_ack_not_all_frames_flushed, gen->head - gen->tail );
     FD_ACK_DEBUG( FD_LOG_DEBUG(( "Not all ACK frames were flushed" )); )
   }
 

diff --git a/src/waltz/quic/tests/fuzz_quic_wire.c b/src/waltz/quic/tests/fuzz_quic_wire.c
@@ -198,7 +198,7 @@ LLVMFuzzerTestOneInput( uchar const * data,
 
   /* Simulate conn timeout */
   while( state->svc_queue[ FD_QUIC_SVC_WAIT ].head != UINT_MAX ) {
-    ulong idle_timeout_ts = conn->last_activity + quic->config.idle_timeout + 1UL;
+    ulong idle_timeout_ts = conn->last_activity + 10*quic->config.idle_timeout + 1UL;
     fd_quic_conn_t * conn = fd_quic_conn_at_idx( state, state->svc_queue[ FD_QUIC_SVC_WAIT ].head );
 
     /* Idle timeouts should not be scheduled significantly late */