kvm-83

[diaolin]

I've seen in the new kvm-83 sources a new command line option:

-pcidevice host=.....

and a set of limitations:

• The host driver for the device, if any, needs to be removed first or assignment will fail.
• Currently, you cannot assign a device that shares irq with another host device
• The raw_io capability is needed for this to work
• This works only with the in-kernel irqchip method. In order to use the userspace irqchip, a kernel module (irqhook) and some extra changes are needed.

Is this included into the new version of PVE?

And, is it possible to have, in each kvm machine an input box that permits to add additional parameters to the kvm guest startup?

I think that having a cmd-line option like:

Add extra args:

can be a good solution for the temerary peoples....
withouth having to touch (modify) the QemuServer.pm

Diaolin

 

[dietmar]

Is this included into the new version of PVE?

V1.1 uses kvm-83, yes.

And, is it possible to have, in each kvm machine an input box that permits to add additional parameters to the kvm guest startup?

feel free to send me a patch. I do not want to have that on the web interface, but an additional line in the config file would be OK for me.

 

[littletiger]

V1.1 uses kvm-83, yes.

I think KVM's PCI Device Assignment requires Linux 2.6.28 kernel, so having kvm-83 might not be enough ?

http://www.linux-kvm.com/content/kvm-79-released-pci-device-assignment-pci-device-hot-plug

 

[diaolin]

hotplug or not hotplug

I think KVM's PCI Device Assignment requires Linux 2.6.28 kernel, so having kvm-83 might not be enough ?

http://www.linux-kvm.com/content/kvm-79-released-pci-device-assignment-pci-device-hot-plug

Mah, i hope that only hotplug does not work but....

 

[diaolin]

patch

V1.1 uses kvm-83, yes.



feel free to send me a patch. I do not want to have that on the web interface, but an additional line in the config file would be OK for me.

when tested with the new.... version

 

[dietmar]

I think KVM's PCI Device Assignment requires Linux 2.6.28 kernel, so having kvm-83 might not be enough ?

http://www.linux-kvm.com/content/kvm-79-released-pci-device-assignment-pci-device-hot-plug

Our kernel also includes the latest kvm kernel module.

 

[diaolin]

fantastico.

Our kernel also includes the latest kvm kernel module.

Good work, it works well, i've not probed the pci device assignment but the rest is very reliable.

A question: vga std or cirrus but vmware is not present in the interface nor in the code and if i set it into the code it reverts the cmd'line as cirrus.

Forgotten anything?

Just no problem with the vmtab${vmid}i${i}.
It works well and the shorewall configuration is a simple click and go.

Next week i'll send to you the files and my proposal of shorewall configuration.

Probed it into the course i-m talking with PVE and the result is
F A N T A S T I C

I'll send to you a screen with my proposal

Diaolin

 

[littletiger]

Our kernel also includes the latest kvm kernel module.

Great stuff! Now I can really make use of the VT-d feature in my computers ... Thanks very much for that.

 

[dietmar]

A question: vga std or cirrus but vmware is not present in the interface nor in the code and if i set it into the code it reverts the cmd'line as cirrus.

What is the advantage of 'vmware' - i did not find any docu about that?

 

[diaolin]

only a question

What is the advantage of 'vmware' - i did not find any docu about that?

i think that this parameter supports more resolutions like widescreen but...
don't know, it was only a simple question.

Do you have any plan to use 2.6.28 for enabling the -pcidevice ?????

I've seen that the manual of kvm does not provide any parameter like this.
In any case the source of kvm-83 includes the code (i hope)
grep -r pcidevice *
qemu/hw/device-assignment.c: * -pcidevice host=bus:dev.func[,dma=none][,name=Foo]
qemu/hw/device-assignment.c: * -pcidevice host=00:13.0,dma=pvdma
qemu/hw/device-assignment.c: fprintf(stderr, "pcidevice argument parse error; "
qemu/vl.c: "-pcidevice host=bus:dev.func[,dma=none][,name=string]\n"
qemu/vl.c: QEMU_OPTION_pcidevice,
qemu/vl.c: { "pcidevice", HAS_ARG, QEMU_OPTION_pcidevice },
qemu/vl.c: case QEMU_OPTION_pcidevice:

In order to use the "feature" is it possible to have a QemuServer.pm that reads from $vmid.conf:

xargs: ""

and QemuServer.pm

on line 1241

if ($conf->{xargs}) {
push @$cmd, $conf->{xargs};
}


But the really question: why use the cmdline?
into the management interface ->vmXXX->Options a simple
input box with name Additional args.

Tx. Diaolin

 

[dietmar]

Do you have any plan to use 2.6.28 for enabling the -pcidevice

It does not work with 2.6.24, and we cant use 2.6.28 because there is no OpenVZ patch.

But the really question: why use the cmdline?

Because the web interface must stay clean. Such experimental options are nothing for average users.

- Dietmar

 

[diaolin]

Ok, good

It does not work with 2.6.24, and we cant use 2.6.28 because there is no OpenVZ patch.



Because the web interface must stay clean. Such experimental options are nothing for average users.

- Dietmar

but if i write
xargs: myargs

and i modify the machine from interface the parameter remains as is, of course