Re: [PATCH net-next] tcp: document black hole detection for tcp_mtu_probing

From: Eric Dumazet

Date: Tue Nov 17 2026 - 03:18:15 EST


Le mar. 17 nov. 2026 à 09:09, Yongjian Wang via B4 Relay
<devnull+echoowall.gmail.com@xxxxxxxxxx> a écrit :
>
> From: Yongjian Wang <echoowall@xxxxxxxxx>
>
> ip-sysctl.rst says that tcp_mtu_probing=1 enables MTU probing once an
> ICMP black hole is detected, but not when that happens or what TCP
> does then.
>
> Describe both, as implemented in tcp_write_timeout() and
> tcp_mtu_probing(). Also add the missing defaults of tcp_mtu_probing
> and tcp_base_mss.
>
> Signed-off-by: Yongjian Wang <echoowall@xxxxxxxxx>
> ---
> This describes the behavior checked by the black hole detection tests
> added in commit 8a83f21ad75d ("selftests/net: packetdrill: add TCP black
> hole detection tests").
> ---
> Documentation/networking/ip-sysctl.rst | 16 ++++++++++++++++
> 1 file changed, 16 insertions(+)
>
> diff --git a/Documentation/networking/ip-sysctl.rst b/Documentation/networking/ip-sysctl.rst
> index 0b7bfbd98a92..a11baf04758c 100644
> --- a/Documentation/networking/ip-sysctl.rst
> +++ b/Documentation/networking/ip-sysctl.rst
> @@ -412,6 +412,8 @@ tcp_base_mss - INTEGER
> Path MTU discovery (MTU probing). If MTU probing is enabled,
> this is the initial MSS used by the connection.
>
> + Default: 1024
> +
> tcp_mtu_probe_floor - INTEGER
> If MTU probing is enabled this caps the minimum MSS used for search_low
> for the connection.
> @@ -711,6 +713,20 @@ tcp_mtu_probing - INTEGER
> - 1 - Disabled by default, enabled when an ICMP black hole detected
> - 2 - Always enabled, use initial MSS of tcp_base_mss.
>
> + A black hole is detected when the retransmission timer fires after
> + the handshake and retransmissions have stayed unacknowledged for as
> + long as tcp_retries1 retransmissions would take with the minimum
> + RTO, which is 3 seconds with the default tcp_retries1.

"minimum RTO" is ambiguous in this file: readers will think of
tcp_rto_min_us or the per-route rto_min. retransmits_timed_out() uses
the TCP_RTO_MIN constant here, so I would spell out 200 ms.

It might also be worth saying that this time is counted from the first
retransmission, and only checked when the timer fires. For a flow with
a 200 ms RTO, the detection happens at the fifth timeout, about
6 seconds after the original transmission.

(The 3 seconds also assume the default tcp_rto_max_ms, but I do not
think this needs to be documented.)

> +
> + With value 1, the first detection enables MTU probing and the MSS
> + falls back to tcp_base_mss. Each later detection, or any detection

The MSS is only lowered: tcp_sync_mss() uses
min(mss, search_low mss), so a connection already below tcp_base_mss
keeps its MSS.

"Each later detection" could be "Each further retransmission timeout",
since tcp_mtu_probing() is called on every RTO once the threshold has
been reached.

> + with value 2, lowers search_low to half its MSS, keeping it at most
> + tcp_base_mss and at least tcp_mtu_probe_floor and tcp_min_snd_mss.
> + Once MTU probing is enabled, the connection also probes for a larger
> + MSS, as with value 2.
> +
> + Default: 0
> +
> tcp_probe_interval - UNSIGNED INTEGER
> Controls how often to start TCP Packetization-Layer Path MTU
> Discovery reprobe. The default is reprobing every 10 minutes as
>
> ---
> base-commit: 6ff4c5942a301675eb12e5b49d53a81e604624a1
> change-id: 20261117-tcp-mtu-probing-doc-v1-351cb754fa7b
>
> Best regards,
> --
> Yongjian Wang <echoowall@xxxxxxxxx>
>

Reviewed-by: Eric Dumazet <edumazet@xxxxxxxxxx>