Handle unknown channel messages correctly

[mus65]

See discussion #1218 . Some servers send custom channel messages like 'keepalive@proftpd.org' as keep alive messages. This currently causes a NotSupportedException.

According to the spec https://datatracker.ietf.org/doc/html/rfc4254#section-5.4 :

"If the request is not recognized or is not
supported for the channel, SSH_MSG_CHANNEL_FAILURE is returned."

Send a failure message back instead of throwing an exception.

The test is mostly copy&paste from ChannelTest_OnSessionChannelRequestReceived_OnRequest_Exception. I also reproduced the issue and tested the fix with an actual ProFTPD server locally.

[Rob-Hague]

The whole paragraph is (emphasis mine):

 If 'want reply' is FALSE, no response will be sent to the request.
**Otherwise**, the recipient responds with either
SSH_MSG_CHANNEL_SUCCESS, SSH_MSG_CHANNEL_FAILURE, or request-specific
continuation messages. If the request is not recognized or is not
supported for the channel, SSH_MSG_CHANNEL_FAILURE is returned.

To me, that says we should only send SSH_MSG_CHANNEL_FAILURE if 'want reply' is true.

i.e.

 // Raise request specific event
OnRequest(requestInfo);
}
- else+ else if (e.Message.WantReply)
{
var reply = new ChannelFailureMessage(LocalChannelNumber);
SendMessage(reply);

Unfortunately, WantReply does not exist on ChannelRequestMessage, but instead on RequestInfo which is a wrapper for the "type-specific data" in the SSH_MSG_CHANNEL_REQUEST. I guess this was done because 'want reply' often depends on the type of the request when sending one. I see two options:

1. Plumb WantReply onto ChannelRequestMessage instead of RequestInfo
2. Add your UnknownRequestInfo into the production code in order to load the 'want reply' value; or
   2.2 Un-abstract RequestInfo for the same purpose

2 is probably easiest. What do you think?

[mus65]

2. Add your UnknownRequestInfo into the production code in order to load the 'want reply' value;

@Rob-Hague Pushed. Moved UnknownRequestInfo to production code and it checks WantReply now. The test still needs its own class to set WantReply=true .

[zybexXL]

This RFC clarification (draft) may be relevant here, though it's a slightly different issue arising from the same RFC paragraph:
http://www.watersprings.org/pub/id/draft-sgtatham-secsh-closure-race-00.html

This issue has been discussed before in other SSH packages:
https://libssh2.org/mail/libssh2-devel-archive-2012-01/0000.shtml

And this implementation also agrees with your interpretation of WantReply:
https://github.com/hudson/ganymed-ssh-2/blob/master/src/main/java/ch/ethz/ssh2/channel/ChannelManager.java#L1315

[Rob-Hague]

This RFC clarification (draft) may be relevant here

Thanks. The relevant text is:

Specifically, users of [RFC4254] should replace the first sentence of the third paragraph of section 5.4:
If 'want reply' is FALSE, no response will be sent to the request.
with the text:
If 'want reply' is FALSE, or if the sender has already sent SSH_MSG_CHANNEL_CLOSE for the channel, then the sender MUST NOT send a reply.

I think we are covered here because SendMessage has a check on IsOpen which is false when we have sent SSH_MSG_CHANNEL_CLOSE:

SSH.NET/src/Renci.SshNet/Channels/Channel.cs

Lines 479 to 488 in db3d7e8

	protectedvoidSendMessage(Messagemessage)
	{
	// Send channel messages only while channel is open
	if(!IsOpen)
	{
	return;
	}
	_session.SendMessage(message);
	}

SSH.NET/src/Renci.SshNet/Channels/Channel.cs

Lines 547 to 575 in db3d7e8

	if(!_closeMessageSent&&IsOpen&&IsConnected)
	{
	if(TrySendMessage(newChannelCloseMessage(RemoteChannelNumber)))
	{
	_closeMessageSent=true;
	// only wait for the channel to be closed by the server if we didn't send a
	// SSH_MSG_CHANNEL_CLOSE as response to a SSH_MSG_CHANNEL_CLOSE sent by the
	// server
	varcloseWaitResult=_session.TryWait(_channelClosedWaitHandle,ConnectionInfo.ChannelCloseTimeout);
	if(closeWaitResult!=WaitResult.Success)
	{
	DiagnosticAbstraction.Log(string.Format("Wait for channel close not successful: {0:G}.",closeWaitResult));
	}
	}
	}
	if(IsOpen)
	{
	// mark sure the channel is marked closed before we raise the Closed event
	// this also ensures don't raise the Closed event more than once
	IsOpen=false;
	if(_closeMessageReceived)
	{
	// raise event signaling that both ends of the channel have been closed
	Closed?.Invoke(this,newChannelEventArgs(LocalChannelNumber));
	}
	}

[Rob-Hague]

Thanks

[anandgmenon]

@Rob-Hague/ @WojciechNagorski Can we have a release with this fix please?

[Rob-Hague]

Probably some time in May

[anandgmenon]

@Rob-Hague can we please have a release for this?

[Rob-Hague]

Yes, soon

[anandgmenon]

Hey @Rob-Hague can you provide a tentative ETA on when you're planning to do the next release?