Re: [PATCH] socket: do not enable POLLOUT for empty out buffer

[Nikolay <nitro@xxxxxxxxxxx>]

Thank you, Aris.

> No problem, as I understand you're using revents as a kind of a flag to
> detect when the output buffer is totally empty. It's not exactly how it
> was designed :)

= No, I do understand that POLLOUT is either a connect-attempt-finished 
or outgoing-buffer-available event.
This is why I suppose the patch - in my mind we don't need to wait for 
POLLOUT if we have already passed all the data to send(). I do not quite 
understand the reason why we need to wait for POLLOUT unconditionally...
POLLOUT will only mean that there is space available in output buffer.

But I do not analyze revent value in socket callback function. I simply 
check if session is established and do ssh_channel_select() or 
ssh_connect()/ssh_auth...() otherwise.

> I'm not sure I fully understand the sequence of actions here. Are we
> somewhere in (backtrace)
> ssh_event_dopoll()
> ____ssh_socket_event_callback()
> ________ssh_channel_select()

= Yes, backtrace seems like the one I have in mind.

> Are the parameters to these functions SSH sessions? Because if so you
> may enter in reentrancy zone and you may have unexpected results.
= I only pass session socket to ssh_event_dopoll().
But you are possibly correct, as ssh_channel_select() seems to be 
working with the channels' sessions' sockets.

> The best way to handle the spurious POLLOUT is to add the current ssh
> session into an ssh_event
> LIBSSH_API int ssh_event_add_session(ssh_event event, ssh_session session);
> and then call ssh_event_dopoll() with timeout=0
>
> I'm not sure how it would fit into your architecture though. Maybe you
> should use the channel callbacks instead of ssh_channel_select() inside
> of global socket callbacks.

Yes, I did try to use ssh_event_add_session() and set some callbacks.
But I strongly need it all to work in non-blocking mode. But the 
documentation for ssh_event_add_session() says:
> remove the poll handle from session and assign them to a event, when 
used in *blocking mode*

I think it's the reason I failed with using ssh_event_add_session() and 
session callbacks. I did get some SIGSEGV during those experiments.

I'm connecting in the following way:

if (session_not_connected && ssh_connect() == SSH_AGAIN)
  return to eventloop and wait for event on session socket, then run 
ssh_connect() again

My thread should not block in waiting for particular 
socket/session/channel I/O. I should be able to connect to multiple ssh 
server in non-blocking async mode.

Best regards, Nikolay Karikh

Attachment: smime.p7s
Description: Криптографическая подпись S/MIME