TCP RST problems, Linux guest, Windows host

[oeuftete]

OK, I'm not a TCP wizard by any stretch. I'll try to sketch out the problems I'm having minimally. People who know more can draw conclusions.

I have two VB 3.0.2 installations. Both running ubuntu intrepid guests. One is with a Vista host, one with XP.

Since upgrading to VB 3.0.2, I'm getting disconnected somewhat randomly from at least ssh and remote desktop connections (using rdesktop) initiated from the guest. I'm seeing this from both of my VB installations, to a variety of destinations.

Sometimes the disconnections happen after a few minutes of inactivity, but I've seen them happen in the span of seconds for both protocols. So this doesn't fit what I would think of as a "timeout".

If I connect to any of the destinations from my hosts (say, using putty for ssh, or Windows Remote Desktop Connection), I don't have the same problems.

If I use wireshark to look at the traffic around the disconnections, I see something roughly like:

Code: Select all

time/s    
0         guest -> dest   T.125 (or ssh)
0.000436  dest  -> guest  ACK
5.609626  guest -> dest   T.125
5.610107  dest  -> guest  RST

Any idea what's going on? As I've said, I'm having the same problem with two VB installations, using two different TCP protocols to a variety of destinations, and no problem when using the same protocols (although with different client software, of course) and destinations directly from the host. So VB is the only common thread I can find here.

[baf]

Try setting your guests to use the intel e1000 server network card type.

[oeuftete]

Try setting your guests to use the intel e1000 server network card type.

I'm afraid you need to be more specific for me. Also, I'd like to know why it might make a difference.

Edit: Maybe I can be more specific too. Are you talking about the other network adapters available in the UI? My current setting is PCnet-FAST III (NAT). Trying to choose any of the three "1000" variants gives me an error on startup about MAC addresses.

[baf]

There has been reported problems like yours with the pcnet adapters that 's why I said that.
To reset the mac-address remove /etc/udev/rules.d/70-persistent-net.rules and reboot the guest.

[oeuftete]

There has been reported problems like yours with the pcnet adapters that 's why I said that.
To reset the mac-address remove /etc/udev/rules.d/70-persistent-net.rules and reboot the guest.

Thank you for the clarification. The MAC address issue was actually during the initialization of the VM, and a little bit of extra searching pointed to stray port-forwarding pcnet extra data items that I needed to remove manually. The VM is running again... I'll wait to see if I see the same problems.

Is there a bug ticket for the issue I originally reported that you know of? I looked and couldn't find one.

[oeuftete]

Try setting your guests to use the intel e1000 server network card type.

So I have switched to using the Intel PRO/1000 MT Desktop adapter, with NAT. The problem still seems to exist exactly as before, unfortunately.

[slobo]

I have a similar problem.
Wireshark (on Windows XP Pro host) and tcpdump (on Ubuntu 10.04 guest) show that guest doesn't get any of the SYN packets, instead something on the host sends RST. Is there a way to inspect what is sending the reset packets? It is reproducible 100% on my setup:

I have a webapplication on port 3000, VBox nat forwards to that port (and few other ports as well):

Code: Select all

          <NAT network="192.168.244/24">
            <DNS pass-domain="true" use-proxy="false" use-host-resolver="false"/>
            <Alias logging="false" proxy-only="false" use-same-ports="false"/>
            <Forwarding name="gestweb" proto="1" hostport="3000" guestport="3000"/>
            <Forwarding name="gesthttps" proto="1" hostport="443" guestport="443"/>
            <Forwarding name="gestsnmp" proto="0" hostport="162" guestport="162"/>
            <Forwarding name="guesthttp" proto="1" hostport="80" guestport="80"/>
            <Forwarding name="guestssh" proto="1" hostport="22" guestport="22"/>
          </NAT>

Using any network adapter type (PCnet-PCI II & III, Intel Pro/1000 Desktop and Server x2, virtio), windows host will send RST packets when contacting the web application. Firefox starts 4 connections in quick succession, and the second two are killed with RST before reaching the guest. I'm running the client on another computer, locally this problem doesn't occur.

Any ideas? I am using Virtualbox 3.2.8.

[slobo]

Perhaps worth noting that I am also experiencing loss of network connection after some idle period. Like in Ticket 6559.