Re: [PATCH v6 4/4] tcp: Update data retransmission timeout

On Fri, Oct 24, 2025 at 7:04 AM Stefano Brivio wrote:

On Fri, 17 Oct 2025 14:28:38 +0800 Yumei Huang wrote:

...

Use an exponential backoff timeout for data retransmission according to RFC 2988 and RFC 6298. Set the initial RTO to one second as discussed in Appendix A of RFC 6298.

Also combine the macros defining the initial RTO for both SYN and ACK.

Signed-off-by: Yumei Huang Reviewed-by: David Gibson --- tcp.c | 27 ++++++++++++--------------- 1 file changed, 12 insertions(+), 15 deletions(-)

diff --git a/tcp.c b/tcp.c index 9385132..dc0ec6c 100644 --- a/tcp.c +++ b/tcp.c @@ -179,16 +179,14 @@ * * Timeouts are implemented by means of timerfd timers, set based on flags: * - * - SYN_TIMEOUT_INIT: if no ACK is received from tap/guest during handshake - * (flag ACK_FROM_TAP_DUE without ESTABLISHED event) within this time, resend - * SYN. It's the starting timeout for the first SYN retry. If this persists - * for more than TCP_MAX_RETRIES or (tcp_syn_retries + - * tcp_syn_linear_timeouts) times in a row, reset the connection - * - * - ACK_TIMEOUT: if no ACK segment was received from tap/guest, after sending - * data (flag ACK_FROM_TAP_DUE with ESTABLISHED event), re-send data from the - * socket and reset sequence to what was acknowledged. If this persists for - * more than TCP_MAX_RETRIES times in a row, reset the connection + * - RTO_INIT: if no ACK segment was received from tap/guest, either during + * handshake (flag ACK_FROM_TAP_DUE without ESTABLISHED event) or after + * sending data (flag ACK_FROM_TAP_DUE with ESTABLISHED event), re-send data + * from the socket and reset sequence to what was acknowledged. This is the + * timeout for the first retry, in seconds. If this persists too many times + * in a row, reset the connection: TCP_MAX_RETRIES for established + * connections, or (tcp_syn_retries + tcp_syn_linear_timeouts) during the + * handshake. * * - FIN_TIMEOUT: if a FIN segment was sent to tap/guest (flag ACK_FROM_TAP_DUE * with TAP_FIN_SENT event), and no ACK is received within this time, reset @@ -342,8 +340,7 @@ enum { #define WINDOW_DEFAULT 14600 /* RFC 6928 */

#define ACK_INTERVAL 10 /* ms */ -#define SYN_TIMEOUT_INIT 1 /* s */ -#define ACK_TIMEOUT 2 +#define RTO_INIT 1 /* s, RFC 6298 */ #define FIN_TIMEOUT 60 #define ACT_TIMEOUT 7200

@@ -588,13 +585,13 @@ static void tcp_timer_ctl(const struct ctx *c, struct tcp_tap_conn *conn) } else if (conn->flags & ACK_FROM_TAP_DUE) { if (!(conn->events & ESTABLISHED)) { if (conn->retries < c->tcp.syn_linear_timeouts) - it.it_value.tv_sec = SYN_TIMEOUT_INIT; + it.it_value.tv_sec = RTO_INIT; else - it.it_value.tv_sec = SYN_TIMEOUT_INIT << + it.it_value.tv_sec = RTO_INIT << (conn->retries - c->tcp.syn_linear_timeouts); } else - it.it_value.tv_sec = ACK_TIMEOUT; + it.it_value.tv_sec = RTO_INIT << conn->retries;

Same as on 3/4, but here it's clearly more convenient: just assign RTO_INIT, and multiply as needed in the if / else clauses.

I guess we can't just assign RTO_INIT. Maybe assign it only when retries==0, otherwise multiply as it.it_value.tv_sec <<=1. But it seems more complicated. What do you think?

...

} else if (CONN_HAS(conn, SOCK_FIN_SENT | TAP_FIN_ACKED)) { it.it_value.tv_sec = FIN_TIMEOUT; } else {

The rest of the series looks good to me.

It might be slightly more practical to factor in directly the RTO clamp, and I don't think it's complicated now that you have the helper from 2/4, but it's not a strong preference from my side, as the series makes sense in any case.

Reading tcp_rto_max_ms can be easy with the helper. My concern is about the way we get the total time for retries. I used to do it like this in v2, https://archives.passt.top/passt-dev/20251010074700.22177-4-yuhuang@redhat.c...: +#define RETRY_ELAPSED(timeout_init, retries) \ + ((timeout_init) * ((1 << ((retries) + 1)) - 2)) Though the formula is not quite right, we could refine it as below: #define RETRY_ELAPSED(retries) ((RTO_INIT) * ((1 << ((retries) + 1)) - 1)) Does it make sense to get the time this way?

-- Stefano

-- Thanks, Yumei Huang