[pbs-devel] [PATCH proxmox master v1] fix #6738: rest-server: improve TLS handshake error logging

[Max R. Carrara]

On Wed Sep 10, 2025 at 5:40 PM CEST, Max R. Carrara wrote:
> Right now, when a client connects to the server, we continuously peek
> into the TCP stream's buffer until we can perform the TLS handshake
> check. If the client closes the connection after sending < 5 bytes, we
> exit the peek loop. If the client instead keeps the connection open,
> the future times out if the handshake check isn't made.
>
> Instead of exiting with `ErrorKind::ConnectionAborted`, exit with
> `ErrorKind::Other` instead, as that variant isn't ever returned by
> stdlib functions. That way we can be sure we don't accidentally ignore
> an error.
>
> Reduce the amount of log noise by ignoring our `ErrorKind::Other`
> errors, as those are usually only returned when some monitoring
> solution opens and closes a TCP connection for its healthchecks [0].
>
> Log timeouts without their error chain (like before), as
> there isn't any additional context for them.
>
> Because any other kinds of errors should be sufficiently rare, log
> them with their error chain. The error chain is formatted to stay in
> one line, as the default formatting looks quite awkward otherwise.
>
> Make the context added to the `async_io()` coroutine call more
> general and informative, as not only the stream buffer peeking can
> fail. Shorten the general log message as well.
>
> [0]: https://forum.proxmox.com/threads/subject-intermittent-backup-errors-and-repeated-tls-handshake-failures.170089/post-795100
>
> Fixes: #6738
> Signed-off-by: Max R. Carrara <m.carrara at proxmox.com>
> ---
>
> Tested this interactively in my Python console by opening a raw TCP
> stream using the `socket` module.
>
>  proxmox-rest-server/src/connection.rs | 29 ++++++++++++++++++++++++---
>  1 file changed, 26 insertions(+), 3 deletions(-)
>

bump