Re: [Qemu-devel] Re: [PATCH v4 0/4] virtio: Use ioeventfd for virtqueue notify

[Stefan Hajnoczi]

On Wed, Nov 17, 2010 at 8:49 PM, Michael S. Tsirkin <address@hidden> wrote:
> On Wed, Nov 17, 2010 at 08:38:25PM +0000, Stefan Hajnoczi wrote:
>> On Wed, Nov 17, 2010 at 6:01 PM, Michael S. Tsirkin <address@hidden> wrote:
>> > On Wed, Nov 17, 2010 at 04:19:25PM +0000, Stefan Hajnoczi wrote:
>> >> The v4 version includes:
>> >>  * Simpler start/stop ioeventfd mechanism using bool ioeventfd_started 
>> >> state
>> >>  * Support for migration
>> >>  * Handle deassign race condition to avoid dropping a virtqueue kick
>> >>  * Add missing kvm_enabled() check to kvm_has_many_ioeventfds()
>> >>  * Documentation updates for qdev -device with ioeventfd=on|off
>> >
>> > Anthony, could you pls remind me what did you say
>> > about need to stop these threads on migration vmstop?
>> > Or am I confusing this with something else?
>>
>> Two points about the VM change state:
>> 1. It is used to bring up virtio-ioeventfd on the destination host
>> after migration.
>> 2. It handles the race condition where a virtqueue kick is dropped
>> because we deassign the ioeventfd on the source host.
>>
>> So I've implemented the VM change state to get correct migration behavior.
>>
>> The discussion you had with Anthony was more about events happening
>> while the VM is paused and how that could interfere with guest state
>> for migration IIRC.
>>
>> Stefan
>
> Exactly. Could your patches cause a situation where block virtio ring
> is used after migration stopped a VM?

No, when the VM is stopped we deassing the ioeventfd and remove its fd
handler.  Because the fd handler is removed no more events will be
processed.

Stefan