[vfio-users] Steam blocking VMs in csgo?

Taiidan at gmx.com Taiidan at gmx.com
Tue Feb 28 01:22:43 UTC 2017


You guys really should ask valve :<

Something worth doing for better latency is an IOMMU attached networking 
device.
The best choice for performance/latency is an SR-IOV networking device 
set up with the new libvirt method (so the mac address stays the same)

On 02/27/2017 08:19 PM, Brandon Ganem wrote:
> I've got a pretty normal network setup where i'm bridged to my wired. I can
> reproduce every time I play, somewhere between 15 minutes and 45 minutes.
> It's happened 5 or so times now.
>
> On Mon, Feb 27, 2017 at 8:17 PM, Scott <shewless at unleashed-web.org> wrote:
>
>> Why not just nat?
>>
>> scott
>>
>> On Feb 27, 2017 8:10 PM, "Ethan Bugden" <ethan961 at gmail.com> wrote:
>>
>>> I got kicked after ~45 mins of DMing again, waited a couple minutes then
>>> it said "Your connection to matchmaking servers is not reliable" or similar
>>> attempting to find a game, while it let me join a local server. Repaired
>>> steam service and it let me on official servers yet again... wonder if it's
>>> due to networking differences. I use ARP proxy between my wireless card and
>>> a TAP device.
>>>
>>> On Mon, Feb 27, 2017 at 2:06 PM, Brandon Ganem <brandonganem at gmail.com>
>>> wrote:
>>>
>>>> Just got kicked again from a casual game. I had not been running it as
>>>> admin, but doing so just now was not enough to let me back into a game.
>>>>
>>>> I'll do my best to find some contact points with Valve and reference
>>>> this thread as well. Thanks again!
>>>>
>>>> On Sun, Feb 26, 2017 at 9:59 PM, Ethan Bugden <ethan961 at gmail.com>
>>>> wrote:
>>>>
>>>>> Just to clarify, you did repair the steam service in an elevated
>>>>> command prompt/as admin, not a regular one? I've never had this problem
>>>>> occur and have it not solved by a repair, at least it should work long
>>>>> enough to finish your match. It does sound like this is an actual issue
>>>>> with playing in VMs being flagged in the first place though. I just did
>>>>> another 45mins of Valve DM in my VM without issues, I'll keep playing in VM
>>>>> when I play to gather some more info. I call qemu through the CLI, my CPU
>>>>> line is: -cpu host,hv_time,hv_vapic,hv_relaxed,hv_spinlocks=0x1fff, I
>>>>> don't hide KVM. I think we'd probably get better results emailing the VAC
>>>>> and/or CSGO teams than submitting a ticket. We only want to use one
>>>>> instance of the game, so hopefully if they're at least aware of this use
>>>>> case they can keep it in mind during future developments. Or if they fixed
>>>>> the native client that would be nice....
>>>>>
>>>>> On Mon, Feb 27, 2017 at 1:43 PM, Brandon Ganem <brandonganem at gmail.com>
>>>>> wrote:
>>>>>
>>>>>> I'll give it a go again if I get kicked after these changes. Thanks
>>>>>> for your help!
>>>>>>
>>>>>> On Sun, Feb 26, 2017 at 9:41 PM, Taiidan at gmx.com <Taiidan at gmx.com>
>>>>>> wrote:
>>>>>>
>>>>>>> Damn, like I said you should submit a support ticket with valve.
>>>>>>>
>>>>>>> I am sure there is a way, no matter how hard the web 2.0 companies
>>>>>>> try to make it.
>>>>>>>
>>>>>>>
>>>>>>> On 02/26/2017 09:39 PM, Brandon Ganem wrote:
>>>>>>>
>>>>>>>> Updated to the following:
>>>>>>>>     <features>
>>>>>>>>       <acpi/>
>>>>>>>>       <hyperv>
>>>>>>>>         <relaxed state='on'/>
>>>>>>>>         <vapic state='on'/>
>>>>>>>>         <spinlocks state='on' retries='8191'/>
>>>>>>>>       </hyperv>
>>>>>>>>       <kvm>
>>>>>>>>         <hidden state='on'/>
>>>>>>>>       </kvm>
>>>>>>>>     </features>
>>>>>>>>     <cpu mode='host-passthrough'>
>>>>>>>>       <topology sockets='1' cores='6' threads='2'/>
>>>>>>>>     </cpu>
>>>>>>>>     <clock offset='localtime'>
>>>>>>>>       <timer name='hypervclock' present='yes'/>
>>>>>>>>     </clock>
>>>>>>>>
>>>>>>>> So far i'm in a game, but it typically takes anywhere from 30
>>>>>>>> minutes to an
>>>>>>>> hour to get VAC'd based on the last 3 attempts.
>>>>>>>>
>>>>>>>> On Sun, Feb 26, 2017 at 9:34 PM, Taiidan at gmx.com <Taiidan at gmx.com>
>>>>>>>> wrote:
>>>>>>>>
>>>>>>>> You need to use a newer version of libvirt that has the vendor tag
>>>>>>>>> workaround to be able to get past that, it should also give you
>>>>>>>>> better
>>>>>>>>> performance to use all the hyperv enhancements.
>>>>>>>>>
>>>>>>>>> This
>>>>>>>>> <vendor_id state='on' value='whatever'/>
>>>>>>>>> needs to be supported but I forgot which version added it
>>>>>>>>>
>>>>>>>>>
>>>>>>>>> On 02/26/2017 09:30 PM, Brandon Ganem wrote:
>>>>>>>>>
>>>>>>>>> Looks like I've got the opposite of all of that enabled, probably
>>>>>>>>>> was for
>>>>>>>>>> code 43 issues. I'll update to recommended hyper-v settings and
>>>>>>>>>> see.
>>>>>>>>>>
>>>>>>>>>> On Sun, Feb 26, 2017 at 9:12 PM, Taiidan at gmx.com <Taiidan at gmx.com>
>>>>>>>>>> wrote:
>>>>>>>>>>
>>>>>>>>>> Did you try the hyperV clock? it could be a timer issue.
>>>>>>>>>>
>>>>>>>>>>>
>>> _______________________________________________
>>> vfio-users mailing list
>>> vfio-users at redhat.com
>>> https://www.redhat.com/mailman/listinfo/vfio-users
>>>
>>>
>
>
> _______________________________________________
> vfio-users mailing list
> vfio-users at redhat.com
> https://www.redhat.com/mailman/listinfo/vfio-users

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://listman.redhat.com/archives/vfio-users/attachments/20170227/4633d447/attachment.htm>


More information about the vfio-users mailing list