Terminate SSHServerProcess when client sends ctrl+c (without Break for 0 msec error)

[LogFlames]

I wish to create a SSHServer which reads the stdin from a client and if it sends ctrl+c it should close the connection. I am able to sorta achieve this by using the following handle_client function:

async def handle_client(process: asyncssh.SSHServerProcess) -> None:
    process.stdout.write(f'information from server\n')
    async for line in process.stdin:
        print(line)

With this I can detect

        if line == 'exit':
            process.exit(0)
            break

for example which is okay, but I prefer to allow the user to exit with ctrl+c.

In this case ctrl+c does work on the client, however. Although it is because it causes an error on the server: Break for 0 msec. If I omit the async for line in process.stdin line everything that is done/written (text/ctrl+c/ctrl+z etc) is ignored and the connection has to be killed by killing the SSH process. If any reading of the input-stream is done (for example with process.stdin.read(1)/process.stdin.read(-1) etc) the same error is caused on ctrl+c.

Any help/pointers in either how to allow ctrl+c termination (without causing errors) or in figuring out why ctrl+c is causing the Break for 0 msec error would be very appreciated.

[ronf]

I'm going to assume that you're running with the defaults on the server, meaning the line_editor feature is enabled, and that you are using an interactive SSH client like OpenSSH, with it requesting a PTY (and sending terminal type information).

In such a setup, the line editor on the server takes responsibility for doing things like echoing the user's input and allowing them to edit it a line at a time before actually sending it to the server application. It also does things like convert a received Ctrl-C into a "break" (like what you'd get from OpenSSH by entering "~B") and converting a received Ctrl-D as a "soft" end-of-file, similar to what you get when a client half-closes a connection, where the server application will receive an empty string back from a read to indicate the EOF.

The way you deal with this is to put a try..except block round your read() call looking for BreakReceived, or other special exceptions like SignalReceived or TerminalSizeChanged. One such example is at https://asyncssh.readthedocs.io/en/latest/#simple-server-with-input, where you can see it looking for BreakReceived as one way of exiting the read loop.

If you don't want the line editing capability, you can set line_editor=False when starting up the server. However, if you want the user to see their input as they type it, it would become your server application's responsibility to handle that input as it comes in (generally a character at a time) and not receive complete lines of input.

Here's another example:

async def handle_client(process):
    process.stdout.write('Welcome to my SSH server, %s!\r\n\r\n' %
                         process.get_extra_info('username'))

    try:
        async for line in process.stdin:
            if not line:
                process.stdout.write('EOF received\r\n')
                break

            process.stdout.write(f'Line received: {repr(line)}\r\n')
    except asyncssh.BreakReceived:
        process.stdout.write('Break received\r\n')

    process.exit(0)

[LogFlames]

Thank you so much! Both for the snippet and the explanation!
This worked perfectly.

[ronf]

Happy to help!