This is an automated email from the ASF dual-hosted git repository.

acassis pushed a commit to branch master
in repository https://gitbox.apache.org/repos/asf/nuttx.git

commit 872bff9cae4e0d08133190408e26e51a06d0d091
Author: Marco Casaroli <[email protected]>
AuthorDate: Mon Nov 16 21:22:26 2026 +0100

    net/tcp: Double the retransmission timeout at the first expiry too.
    
    tcp_timer() set conn->rto = TCP_RTO << nrtx before it counted the
    expiry in nrtx, so the timeouts of one segment were 1.5 s, 1.5 s, 3 s,
    6 s, ...: the first back-off step came one expiry late.  RFC 6298,
    section 5.5: when the timer expires, back off the timer (RTO = RTO * 2).
    
    Now nrtx counts the expiry first: 1.5 s, 3 s, 6 s, ...  The limits
    (TCP_MAXRTX, TCP_MAXSYNRTX, the shift of at most 4) do not change.
    
    Assisted-by: Claude Code:claude-opus-5-5
    Signed-off-by: Marco Casaroli <[email protected]>
---
 net/tcp/tcp_timer.c | 8 +++++---
 1 file changed, 5 insertions(+), 3 deletions(-)

diff --git a/net/tcp/tcp_timer.c b/net/tcp/tcp_timer.c
index 3084006a6ed..9e17d944311 100644
--- a/net/tcp/tcp_timer.c
+++ b/net/tcp/tcp_timer.c
@@ -591,13 +591,15 @@ void tcp_timer(FAR struct net_driver_s *dev, FAR struct 
tcp_conn_s *conn)
                   goto done;
                 }
 
-              /* Exponential backoff. */
+              /* Exponential backoff: this timeout doubles the wait for
+               * the next one (RFC 6298, section 5.5).
+               */
 
+              conn->nrtx++;
 #ifndef CONFIG_NET_TCP_FIXED_RTO
-              conn->rto = TCP_RTO << (conn->nrtx > 4 ? 4: conn->nrtx);
+              conn->rto = TCP_RTO << (conn->nrtx > 4 ? 4 : conn->nrtx);
 #endif
               tcp_update_retrantimer(conn, conn->rto);
-              conn->nrtx++;
 
               /* Ok, so we need to retransmit. We do this differently
                * depending on which state we are in. In ESTABLISHED, we

Reply via email to