On Wed, Jul 09, 2025 at 02:21:57PM -0700, Hal Murray wrote:
> I'm interested in the case where the underlying socket has done
> setsockopt(sock, SOL_SOCKET, SO_RCVTIMEO, ...
> (and/or SO_SNDTIMEO)
I wouldn't expect anything special to happen happen here, if a read
times out, that's just a read error, so shouldn't look different from a
TCP reset, or similar transport/network layer problem.
FWIW, Postfix handles this differently, the socket is set non-blocking
and retries of the SSL handshake or read/write operations, after
WANTREAD or WANTWRITE, happen only after the socket polls ready, and if
those polls time out the (SMTP) application processes a timeout above
the TLS layer, abandoning the SSL connection.
--
Viktor. 🇺🇦 Слава Україні!