Xen project Mailing List

On Sun, Jul 15, 2018 at 8:19 AM, Kevin Wang <kjw@xxxxxxxxxxxx> wrote:

now, it still occasionally hangs during shutdown (initiated by xl trigger $HVM power), requiring an "xl destroy". The symptoms are slightly different; it sits there consuming about .2-.3 cpu seconds per clock second (1 cpu per hvm, so no parallelism at work), even if left overnight it never finished shutting down. But it's only like 1 in 10 boot cycles now, which is much more manageable than 100% of "clean" shutdowns aren't.

I find it so odd that the simple act of booting is so difficult 8<

- Kevin

On Sat, Jul 14, 2018 at 8:11 PM, Kevin Wang <kjw@xxxxxxxxxxxx> wrote:
Update, since the windows driver didn't want to support my old version of xen, I upgraded to ubuntu server 18 LTS. Now running Xen 4.9:
ii xen-hypervisor-4.9-amd64 4.9.2-0ubuntu1 amd64 Xen Hypervisor on AMD64

I still have the exact same behaviour.
but booting now feels like it takes twice as long, about 120s instead of about 45-60s. sigh, but I can live with it.

I can use xl trigger nmi to force windows to crash dump, but I haven't any idea what to do with it.

looked at vif; setting mac= didn't make any sense. ditto for bridge=. nor gatewaydev=. type=ioemu is the default, but I'm using hvm, so I still need this.

huh. setting model=e1000 seems to have stopped the crashing. It's not satisfying (no actual diagnosis of the failure) but it at least unblocks me from pushing this further into production.

- Kevin

On Thu, Jul 12, 2018 at 10:45 PM, Kevin Wang <kjw@xxxxxxxxxxxx> wrote:
Oh, one other odd thing is that start-> reboot? That doesn't trigger this behavior. I can reboot like 10 times in a row and it works fine. Only xl create has this effect. Will experiment more tomorrow, thanks for the info - kjw

On Thu, Jul 12, 2018, 20:49 Adi Pircalabu <adi@xxxxxxxxxxx> wrote:
On 12-07-2018 23:40, Kevin Wang wrote:
> Ubuntu Server 16, latest apt-get upgrade Specifically:
>
>> ii xen-hypervisor-4.6-amd64 4.6.5-0ubuntu1.4 amd64 Xen
>> Hypervisor on AMD64
>
> Sample config:
>
>> # cat /etc/xen/pos21.hvm
>> builder = "hvm"
>> name = "pos21.hvm"
>> memory = "4096"
>> vcpus = 1
>> vif = ['']
>> disk = ['phy:/dev/lvg/pos21-root,hda,w' ]
>> vnc = 1
>> vncdisplay = 1
>> vnclisten = '10.10.22.15'
>> usb = 1
>> usbdevice = ['tablet'] # fix mouse tracking problem
>> vncpasswd = 'elided'
>> boot = "dc"
>> stdvga = 1
>> videoram = 16
>> localtime = 1 # also need to manually set time inside windows to
>> make it stick
>
> lvg/lvm is just set up for raid 1
>
> Windows 10 Home is also up-to-date wrt Windows Updates. According to
> System Information:
>
>> Version 10.0.10240 Build 10240
>
>> BIOS Version/Date Xen 4.6.5 10/13/2017
>
>> A hypervisor has been detected...
>
> Hardware: Dell PowerEdge T30 ("desktop" form factor)Cpu:
>
>> model name : Intel(R) Xeon(R) CPU E3-1225 v5 @ 3.30GHz
>
> Symptoms:
> boot windows, and it hangs about 10 (cpu) seconds into booting. xl
> list shows that it's now taking up 100% of the cpu. waiting hours does
> nothing. xl destroy, then xl create and it boots fine but recreates
> the user profile as part of an automatic recovery something in
> windows. Start -> Shutdown, and it hangs again. repeat destroy|create,
> and it boots fine, but again recreates the user profile.
>
> The puzzling thing is that these are fresh windows installs. yes,
> plural. all 3 installations hang in the same way! I've tried dd'ing
> the disks as backups to test its repeatability, and it's 100%
> repeatable (out of about 5 tries across 3 VMs). Unfortunately it's
> slow; takes about an hour for each restore, so I haven't done more
> iterations of tests.
>
> I haven't the faintest idea what windows is dying on while booting.
> There's so far no software installed that should be starting at boot.
> (no office, nothing. just windows, maybe chrome. No virus software yet
> either. I glanced through Event Viewer, but nothing seems to stick
> out. I haven't the faintest idea what else to look for in Windows.

Can you replicate the crash after installing the Windows PV drivers?
https://www.xenproject.org/downloads/windows-pv-drivers/winpv-drivers-8/winpv-drivers-821.html

Other ideas:
- Juggle with vif type and model in the domU config
- Get rid of videoram and stdvga entries

If it helps here's a domU configuration template I'm using for Win10,
2012R2 and 2016 on various dom0 Xen versions (4.6.6, 4.8.3, 4.10.1), all
running on CentOS 6 and 7

builder = 'hvm'
memory = 3072
shadow_memory = '8'
cpu_weight = 256
name = 'win10'
vif = ['type=ioemu, bridge=br0, vifname=win10.0']
localtime = 1
acpi = 1
apic = 1
vnc = 1
vcpus = 2
vncdisplay = 1
vncconsole = 1
vnclisten = '0.0.0.0'
serial = 'pty'
disk = ['phy:/dev/vm/win10_img,hda,w']
boot = 'c'
sdl = 0
usb = 1
usbdevice = ['tablet']

HTH

---
Adi Pircalabu

Re: [Xen-users] New Win10 Home reliably crashes every other boot?!?