Re: [win-pv-devel] Windows 10 domU is unstable with PV Bus Driver

[Paul Durrant <Paul.Durrant@xxxxxxxxxx>]

> -----Original Message-----
> From: Michel D'HOOGE [mailto:michel.dhooge@xxxxxxxxx]
> Sent: 18 December 2017 09:10
> To: Paul Durrant <Paul.Durrant@xxxxxxxxxx>
> Cc: win-pv-devel@xxxxxxxxxxxxxxxxxxxx
> Subject: Re: [win-pv-devel] Windows 10 domU is unstable with PV Bus Driver
> 
> Hi Paul,
> 
> > Cool. Probably best to check back after a while and make sure it hasn't
> gone up from 174 after a reasonable amount of use to try to spot any
> potential leak before it actually takes your VM down.
> 
> I have 2 "tail -f" to track xenstored-access.log & qemu-dm-win10.log.
> They didn't show anything else after the 174 grant request was issued.
> 
> However, I managed to have the whole computer (i.e. including dom0) to
> reboot!

Oh dear. What version of Xen and dom0 are you running? Did you get any sort of 
crash dump out of dom0. Also, do you have the Xen watchdog turned on? I wonder 
whether it might have been that that triggered the reboot.

> I was locally connected (through a vswitch & through rdesktop to the
> domU. After a while, it stopped responding. I tried to connect through
> the VNC interface but Windows denied the access saying the user was
> already involved in an RDP connection.
> Before the reboot, the dom0 was behaving normally and nothing was
> added into the logs.
> 
> It also happened twice before I increased the gnttab_max_frames value
> (and I'm quite sure I was also rdesktop-ing to the domU).
> 
> The command I use is:
> rdesktop -u $USER -x 0x80 -g workarea -r disk:share=$HOME/rdp-share -r
> sound:off $XENIP
> 
> 
> 
> > Fair enough. 174 frames is a massive number of grants; I'm surprised the
> VM wants to use that many. How many PV network and storage interfaces
> so you have (and how many vcpus in your guest)?
> 
> A single PV network and 4 vcpus.
> 
> Here is my config:

Thanks Michel. That's a pretty modest config so I really don't see why your 
guest ends up so crazily hungry for grant references. It's most likely the 
storage rather than the network which is the hog. I'll try to provoke the 
behaviour you're seeing on my own rig.

Cheers,

  Paul

> 
> boot = "c"
> builder="hvm"
> device_model_args=[ "-trace", "enable=xen_platform_log" ]
> disk = [ '/dev/vg_sda/lv_win10,raw,hda,w' ]
> keymap = "fr"
> localtime=1
> memory = 3800
> name = "win10"
> on_crash="preserve"
> on_reboot="preserve"
> on_soft_reset="preserve"
> on_watchdog="preserve"
> usb = 1
> usbctrl = [ 'type=devicemodel,version=3' ]
> usbdevice = [ 'tablet' ]
> vcpus = 4
> vif = [ 'script=vif-openvswitch,bridge=ovsbr0,mac=00:16:3e:35:fd:24' ]
> vnc = 1
_______________________________________________
win-pv-devel mailing list
win-pv-devel@xxxxxxxxxxxxxxxxxxxx
https://lists.xenproject.org/mailman/listinfo/win-pv-devel