Re: [Xen-users] a question with gplpv

[Lu Fengqi <curiouslfq@xxxxxxxxx>]

Thank you very much. I confirmed that the backend is qdisk.

xen doc  just tell me the backendtype option normally shouldn't specified.
Can you tell me how to specify a backend type so that the guest's IO bypass qemu?

My current config file:
builder='hvm'
name = "pvhvm"
memory = 512
vcpus = 1
disk = [ 'file:/home/xen/win.img,hda,w' ]
vif = [ 'bridge=xenbr0' ]
boot="dc"
xen_platform_pci=1
_on_poweroff_ = 'destroy'
on_reboot   = 'restart'
on_crash    = 'restart'
sdl=0
vnc = 1
vnclisten = '0.0.0.0'
vncdisplay = 0
vncconsole=1
usb=1
usbdevice='tablet'
device_model_version='qemu-xen'

2016-09-07 13:50 GMT+08:00 Juergen Gross <jgross@xxxxxxxx>:
> On 07/09/16 07:38, 陆丰麒 wrote:
>> Hello everyone.
>>
>> Thanks for reading my questions. I deployed a pvhvm guest(windows xp
>> with gplpv) on xen 4.1.2 and debug corresponding  qemu-dm's process with
>> gdb, I found that the guest's IO seems to go through qemu-dm.In
>> addition, I try to run Fedora24 livecd with the pvhvm, the guest's IO
>> seems to go through qemu too.
>> In my opinion that the pvhvm guest's IO should bypass the qemu. Has
>> anyone can tell me  that how the pvhvm works?
>
> Xen is using qemu for some of the pv backends, too. I believe your
> guests are using the qdisk backend. This is the reason you are seeing
> xen_disk.c in your backtraces: pure hardware emulation being used for
> the I/Os wouldn't have any Xen related components involved.
>
> You can check which backends are being used by calling
>
> xenstore-ls
>
> in dom0. The output should contain the backend type for each guest
> device.
>
>
> Juergen



--
-------------------------------------------------------------
Thanks,
Lu

_______________________________________________
Xen-users mailing list
Xen-users@xxxxxxxxxxxxx
https://lists.xen.org/xen-users