[vbox-dev] Issues with host networking stack on 5.0.17

[William Holroyd]

This ticket seems to be the closest to my issue -
https://www.virtualbox.org/ticket/15304

For example, the ability to ping a box when the problem occurs results in a
"GENERAL FAILURE" on each ping attempt. I've tried switching from NAT to
Bridged and the issue persists. From what I read, this looks to be an issue
with the new NDIS6 work for Windows 10 hosts. I did find that in one of the
Windows event logs, that every time the issue comes and goes, there is a
"network change" event. My guess is this filter ends up dieing somehow or
gets killed/reloaded, but there aren't any events saying so.

I have noticed since sending the prior email that if I have too many apps
with network connections open, this issue occurs more often. Simply closing
an app that I know persists a connection (or multiple) like HipChat or
Slack makes networking work again for a little while - and I can do this
inside the guest or on the host. It may seem like there is an artificial
limit that is somehow being hit in the NDIS adapter where it can't handle a
certain number of connections and fails for things that aren't already
established. I am certainly not doing anything excessive network wise, as
Windows can handle it just fine when I stop all of the VMs.

If there is a log somewhere I should check, or a method to start VirtualBox
to log issues with the NDIS adapter, please let me know. I'm more than
happy to help where I can.



On Thu, Apr 21, 2016 at 10:50 AM, William Holroyd <wholroyd at gmail.com>
wrote:

> Hi there,
>
> I didn't see a good place to provide feedback on test builds, so I'm
> trying the email list.
>
> I have a Windows 10 .14316 insider build for a host with a Fedora 23
> guest. The best way I can describe it is that the VirtualBox networking
> stack fails to accept new outgoing connections from time to time that
> affects both the host and the guest simultaneously. It appears that
> established connectivity fails, but anything new will fall on it's face for
> a period of a few seconds to a couple of minutes.
>
> Windows event logging for every log I can think of that would effect
> networking complains that connections couldn't be established or that DNS
> lookups timed out, but nothing about all up failures in drivers or services
> for when this occurs. Even the winsock ETL complains about connectivity
> failures, but not driver or service failures (everything complains as
> though the failure is elsewhere upstream).
>
> I have another Windows 10 .14316 box that doesn't have this issue, but it
> also doesn't have VirtualBox installed on it. I can get this troublesome
> machine to work properly for extended periods of time if I uninstall
> VirtualBox entirely. It seems to me something in the virtual switch (or
> similar mechanism) is having issues trying to handle traffic between the
> host and guests.
>
> Is there any method I can use that might be able to tell me what is going
> on in the virtual switch to see if it's having issues? It's the one thing I
> have no insight into and that everything appears to be pointing to.
>
> William
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://www.virtualbox.org/pipermail/vbox-dev/attachments/20160427/939b29b8/attachment.html>