[Qemu-devel] Re: Hot-adding disk on VM

[Dustin Kirkland <kirkland@canonical.com>]

On Sun, Oct 11, 2009 at 4:51 PM, Daniel Bareiro <daniel-listas@gmx.net> wrote:
> Hi, all!
>
> On Saturday, 22 August 2009 10:59:47 -0300,
> Daniel Bareiro wrote:
>
>> According to I see in this document [1], is necessary that is loaded
>> two modules in the guest: acpiphp and pci_hotplug.
>>
>> The pci_hotplug module is loaded. Nevertheless, in spite of existing
>> the acpiphp module, cannot be loaded:
>>
>> chandra:~# lsmod|grep pci_hotplug
>> pci_hotplug            23460  0
>>
>> chandra:~# modprobe acpiphp
>> FATAL: Error inserting acpiphp
>> (/lib/modules/2.6.26-2-686/kernel/drivers/pci/hotplug/acpiphp.ko): No such device
>>
>> chandra:~# locate acpiphp | grep 686
>> /lib/modules/2.6.26-2-686/kernel/drivers/pci/hotplug/acpiphp.ko
>> /lib/modules/2.6.26-2-686/kernel/drivers/pci/hotplug/acpiphp_ibm.ko
>>
>>
>> Which could be being the problem?
>
> I was doing other tests this time with Linux 2.6.31.2 compiled by myself
> from the sources of kernel.org. This time with loading the acpiphp
> module, the pci_hotplug module was loaded automatically without I get
> some type of error message.
>
> central:~# lsmod | grep acpiphp
> central:~#
> central:~# modprobe acpiphp
> central:~#
> central:~# lsmod | grep acpiphp
> acpiphp                22140  0
> pci_hotplug            29024  1 acpiphp
>
>
> Then, I tried to add a disk of the following way in host:
>
> (qemu) pci_add auto storage file=/dev/vm/central-test,if=virtio
> OK domain 0, bus 0, slot 5, function 0
>
> In the guest:
>
> Oct 11 18:09:01 central kernel: [  833.403875] pci 0000:00:02.0: BAR 6: bogus alignment [0x0-0x0] flags 0x2
> Oct 11 18:09:01 central kernel: [  833.406337] decode_hpp: Could not get hotplug parameters. Use defaults
> Oct 11 18:09:01 central kernel: [  833.430492] virtio-pci 0000:00:05.0: enabling device (0000 -> 0001)
> Oct 11 18:09:01 central kernel: [  833.452268] ACPI: PCI Interrupt Link [LNKA] enabled at IRQ 10
> Oct 11 18:09:01 central kernel: [  833.454668] virtio-pci 0000:00:05.0: PCI INT A -> Link[LNKA] -> GSI 10 (level, high) -> IRQ 10
> Oct 11 18:09:01 central kernel: [  833.500051] Clocksource tsc unstable (delta = -148257950 ns)
> Oct 11 18:09:01 central kernel: [  833.566755]  vda: vda1
>
>
> Then I attempt to remove the device assuring previously that it is not
> being used in guest:
>
> (qemu) pci_del 5
>
>
> In the guest:
>
> Oct 11 18:13:32 central kernel: [ 1104.144179] virtio-pci 0000:00:05.0: PCI INT A disabled
>
>
> But if I want to add the device again, I get a segmentation fault of the
> process that handles this virtual machine:
>
> (qemu) pci_add auto storage file=/dev/vm/central-test,if=virtio
> Connection closed by foreign host.

I can actually confirm this behavior.  I've been struggling with this
for a couple of days now.  CC'ing QEMU, as I've discussed this a bit
with Anthony Liguori.  There's at least two related bugs we've been
tracking:
 * https://bugs.edge.launchpad.net/ubuntu/+source/qemu-kvm/+bug/432154
 * https://bugs.edge.launchpad.net/ubuntu/+source/qemu-kvm/+bug/419590

:-Dustin