Sending data that is greater then 1Mb causes permanent wait about fxssh HOT 14 CLOSED

Should the line ClientWindowSize -= packetSize; be present?

from fxssh.

Could please tell me how to re-produce this issue? What's client tool used? Please debug & watch the ClientInitialWindowSize and ClientMaxPacketSize on FxSsh.Services.Channel.cs

from fxssh.

Did you send large data (>ClientInitialWindowSize) at the message loop thread? In another word, did you send this data in service_CommandOpened event?

If yes. Yes , it is TURE make message blocking, going to dead lock.

from fxssh.

Hi, @xucito
Do you have more question?

from fxssh.

Hi @Aimeast any data sent via the SendData function will cause this deadlocking. I did send it via the service_CommandOpened event. I have fixed it by removing the ClientWindowSize packetsize subtraction.

xucito@8d3c0b9

from fxssh.

Hi @xucito
Please not do this. This will break flow-control mechanisms and may cause network more jam.
The right way is create new Threading or Task to send BIG message on service_CommandOpened event.

from fxssh.

Interesting, should it not close the session then? The behavior where it hangs seems incorrect. Can you link to me where you got the 1MB limitation or the Client Window Size limitation, I am using SSH based tools that hit the limitation quite easily so I am assuming that there is some incorrect behavior in the ssh implementation.

from fxssh.

In design, we can't fix it without redesign. Client Windows Size is send from Client, Server side must observe it.
A way to fix it shall re-design message loop, split message loop to message read thread and handle thread.

from fxssh.

Hi @xucito
Please try https://github.com/Aimeast/FxSsh/tree/fix24

from fxssh.

So that seems to work however I am encountering a new issue as a result of the fix you posted:
Seems that when you handle the invoke DataReceived, it is possible for the thread to continue without the full message being acknowledged by the client?

The loop below will sometimes not send the entire message before proceeding to send the closed

foreach (var section in splitResult)
{
          DataReceived?.Invoke(this, Encoding.ASCII.GetBytes(section));
}
CloseReceived?.Invoke(this, 0);

foreach (var section in splitResult)
{
          DataReceived?.Invoke(this, Encoding.ASCII.GetBytes(section));
          Thread.Sleep(500); //This line fixes it
}
CloseReceived?.Invoke(this, 0);

I do not know whether it is a coincidence but I did not have this issue before using my bad fix

from fxssh.

I have tested following code in program.cs service_CommandOpened, no issue found. Could you please tell how do you test and the scene. e.g. the length of section, length of splitResult, condition of network is fast or slow network?

            var rnd = new Random();
            for (int i = 0; i < 10000; i++)
            {
                e.Channel.SendData(Encoding.ASCII.GetBytes(rnd.Next().ToString()));
            }

from fxssh.

So this issue is observed when I have some latency (even if its not very much) between the server the client.

When I run both server and client on same machine, I don't encounter the issue, when I have the server and client on their own machines I get the following kinds of errors errors:

OpenSSL 1.0.2k-fips  26 Jan 2017\r\ndebug1: Reading configuration data /etc/ssh/ssh_config\r\ndebug1: /etc/ssh/ssh_config line 58: Applying options for *\r\ndebug1: auto-mux: Trying existing master\r\ndebug2: fd 3 setting O_NONBLOCK\r\ndebug2: mux_client_hello_exchange: master version 4\r\ndebug3: mux_client_forwards: request forwardings: 0 local, 0 remote\r\ndebug3: mux_client_request_session: entering\r\ndebug3: mux_client_request_alive: entering\r\ndebug3: mux_client_request_alive: done pid = 25163\r\ndebug3: mux_client_request_session: session request sent\r\ndebug1: mux_client_request_session: master session id: 2\r\ndebug3: mux_client_read_packet: read header failed: Broken pipe\r\ndebug2: Received exit status from master 0\r\nShared connection to 10.10.10.16 closed.\r\n

The errors indicate that the client was expecting the rest of the message but received the exit signal.

When I add the sleep, it no longer has this issue.

from fxssh.

I have updated https://github.com/Aimeast/FxSsh/tree/fix24 to test via My phone JuiceSSH client over 20+ times, the test goes to send a big random text. There are no any issues found.

var rnd = new Random();
for (int i = 0; i < 1000; i++)
{
    var str = "";
    for (int j = 0; j < 100; j++) // size tested from 100 to 3000
    {
        str += rnd.Next().ToString();
    }
    e.Channel.SendData(Encoding.ASCII.GetBytes(str));
}

So, can you exactly description how reproduce you issue. Including client name & version, server side simplified code, data size and so on.

from fxssh.

Let's close this for now, my client is actually a Ansible OpenSSL client. I will create a different issue when I write a test to reproduce and can reliably replicate this. Thanks for all the help

from fxssh.