Re: making it so SSH server sends ext-info-s

On May 24, 2024, at 5:16 PM, Terra Frost <terrafrost%gmail.com@localhost> wrote:

I'm trying to test some changes to an SSH client relating to RFC8308 but am having some difficulty doing so.

I can make the SSH client send ext-info-c to SSH servers but none of them are sending ext-info-s back.

Does anyone know of any SSH server that implements ext-info-s and what one must do to configure it?

According to the bottom of https://terrapin-attack.com/patches.html the people who discovered the terrapin attack contacted the maintainers of all SSH implementations that, among other things, implemented RFC8308. As OpenSSH was among those contacted I assume that that means that OpenSSH implements it but it's unclear to me how to get OpenSSH (or any other SSH server for that matter) to send ext-info-s.

The Terrapin changes don’t require ext-info-s. They actually add other independent items to the kex algorithm list (kex-strict-c-v00%openssh.com@localhost and kex-strict-s-v00%openssh.com@localhost), similar to the ext-info-c and ext-info-s values described in RFC 8308.

The reason you don’t see many implementations sending ext-info-s is that a server is only supposed to send this if it plans on actually parsing something out of the client’s SSH_MSG_EXT_INFO message. The same goes in reverse for the client. However, there are popular extensions generated by the server and parsed by the client (such as "server-sig-algs") but not as many widely implemented in the opposite direction.

If you’re looking for an implementation that sends ext-info-s, AsyncSSH (https://asyncssh.readthedocs.io/en/latest/) is one example. For now, the only thing AsyncSSH sends in the client SSH_MSG_EXT_INFO is “global-requests-ok” (see https://datatracker.ietf.org/doc/html/draft-ssh-global-requests-ok-00 for details).

--
Ron Frederick
ronf%timeheart.net@localhost