Re: [PATCH net 2/2] Revert "net: advertise TCP MSS from the configured MTU, not the learned PMTU"

[Esko Mobius]

On Thu, 29 Oct 2026, David Laight wrote:
> But that relies on something having gone the other way first.
> In general that won't be true.

Agreed, that is the same point Jiayuan made.

> It is almost worth implement something that retries with the shorter mtu
> to detect the PPPoE mtu blackhole at the sending end.

Linux already has this: net.ipv4.tcp_mtu_probing (RFC 4821). With it
set to 1, once the retransmission timeouts reach tcp_retries1 the
connection drops to tcp_base_mss and probes up from there. It is off
by default, and the same goes for Windows (EnablePMTUBHDetect); macOS
turns it on by default.

> Or even just use the shorter mtu for long RTT connections.

Both of these have to happen at the sending end, which in my case is
the client (it is an upload), so the server can't help. The only lever
the server has is the MSS it advertises in the SYN-ACK, and it has no
RTT sample at that point. For now the MSS is clamped on the path.

While reading this code I noticed there are no selftests for the black
hole detection or for MTU probing. I'd like to add packetdrill tests
for both once net-next reopens.

Thanks,
Yongjian