Linux Virtualization - Apr 2023 - [PATCH v6 11/11] vhost: allow userspace to create workers

On 4/12/23 2:56 AM, Jason Wang wrote:
>> I can spin another patchset with the single ioctl design so we can
compare.
> So I'm fine with this approach. One last question, I see this:
> 
> /* By default, a device gets one vhost_worker that its virtqueues share.
This */
> 
> I'm wondering if it is better to have a vhost_get_worker() to return
> the worker_id of a specific queue. In the future, this might be useful
> for devices that have multiple workers by default?
Yeah, it would be useful. Just 2 questions. You mean an ioctl command to get
this info right? I had added one in one version a while back, but dropped it
because for some reason I thought we were going to do a generic vhost sysfs
type of interface. We are not doing sysfs right?

On Fri, Apr 14, 2023 at 6:36?AM Mike Christie
<michael.christie at oracle.com> wrote:
>
> On 4/12/23 2:56 AM, Jason Wang wrote:
> >> I can spin another patchset with the single ioctl design so we can
compare.
> > So I'm fine with this approach. One last question, I see this:
> >
> > /* By default, a device gets one vhost_worker that its virtqueues
share. This */
> >
> > I'm wondering if it is better to have a vhost_get_worker() to
return
> > the worker_id of a specific queue. In the future, this might be useful
> > for devices that have multiple workers by default?
>
> Yeah, it would be useful. Just 2 questions. You mean an ioctl command to
get
> this info right? I had added one in one version a while back, but dropped
it
> because for some reason I thought we were going to do a generic vhost sysfs
> type of interface. We are not doing sysfs right?
Probably, but we need to make sure the emulatorpin can work for libvirt at last.

Thanks
>