[Xen-users] Mouse in a DomU suddently dying during high CPU usage of Dom0/other DomUs

[Zir Blazer <zir_blazer@xxxxxxxxxxx>]

The problem is pretty much what is stated on the title. When there is heavy CPU 
Usage anywhere else but the DomU that I mainly use for gaming, and 
simultaneously, I do something CPU (And maybe I/O too) intensive in my DomU, 
the Mouse suddently dies, having no chance to recover but restarting the DomU.

I'm starting a home cloud hosted in my computer which has a fully working Xen 
install, and other home members may use a VM I'm providing so they can access 
it, install some heavy applications which are too slow in their computers, and 
do their work there. In practice, this is extremely disruptive for me since 
when they put their VM at full load, everytime that I'm also doing some duty 
here (Which may be playing a game), I have ocassional minifreezes. During the 
minifreezes, the Mouse cursor may get stuck in one direction (Notorious in 
First Person Shooter games like Doom, with the zDoom port) which after some 
seconds and moving the cursor around recovers, but, if the minifreeze last more 
than what appears to be a timeout, the Mouse simply dies in my DomU. I see the 
cursor at the last position, but it is totally unmovable. If I Ctrl + Alt to 
Dom0 X.org or another DomU with GUI and mouse, things are working fine there.
I have consistently noticed this behaviator since a year ago or so, as I recall 
that when compiling packages in Dom0 (With only one DomU running, my gaming 
one) I also had this same issue, just that I workarounded it by doing very 
gentle cursor movements when I loaded my DomU simultaneously, as it seems that 
simultaneous load and hard turning Mouse movements is what causes the Mouse to 
die.

After asking several times in IRC with no good response, I found no way to do a 
soft reset of the Mouse, forcing me to have to shut down and start again the 
DomU. This is pretty much the only reason why I have to restart that DomU at 
all.


Reelevant info:

HARDWARE
Processor: Xeon E3-1245V3 (4C/8T, but I have Hyper Threading disabled)
Storage: Seagate HDD.15 4 TB Hard Disk with LVM volumes for each DomU

SOFTWARE
Arch Linux as Dom0
Xen 4.5, Linux Kernel 4.0.1


Dom0 sees all 4 Cores but is limited to 2 GiB RAM. Most of the Mouse dying 
issues when using a single DomU but a loaded Dom0 started to appear after I 
enabled the compiler makeflags=-j4 to compile using 4 Threads. I don't recall 
it happened at all with only a single Thread.
Affected DomU runs WXP x64 SP2, but it happened too in WXP SP3 IIRC. It also 
has 4 Cores (vcpus = 4), and has affinity just to pin them (cpus = "0,1,2,3"). 
For interacting with it, I use sdl = 1. Oh, and also has VGA Passthrough, but I 
don't think than that is involved, even through I didn't tried in a test VM 
just to check if it happens on others. I'm NOT using usbdevice = "tablet".
The new DomU that generates this condition has only 2 Cores, pinned down (cpu = 
"2,3"). It also has sdl = 1, but also simultaneously has SPICE enabled, this 
way, I can see (And even interact if needed) with the VM while they remotely 
use them with a SPICE client.


Any ideas? Yes, I know that it would be better if I only use Core 0 and 1 in my 
VM while the other one is working, but since CPU hotplugging is not a Windows 
friendly feature, it means that I would have to restart it anyways, so I'm 
looking for a way to merely do a soft reset of the Mouse when things go wrong.  
                                         
_______________________________________________
Xen-users mailing list
Xen-users@xxxxxxxxxxxxx
http://lists.xen.org/xen-users