Re: [PATCH 0/2] specify the accelerator type using "-accel" parameter
ping
Hi!
A few week ago, i post a patchset about changing the way to specify the
accelerator type accordding to the QEMU maitainer's suggestion.
It would be convenient if Libvirt can support specifying acceleratoer
type in this way.Since more attibutes of accelerator can be specified
via Libvirt,such as kernel-irqchip，kvm-shadow-mem，tb-size，etc。
I think this series can make sense, and i would like to know whether
there is something wrong about this series, or other concern about this
series。I would be very thankful if letting me know.
Thanks!
Hyman
在 2021/1/21 10:21, Hyman 写道:
> Ping
Hi!
I think this series still make sense, even though the dirty ring
patch haven't been merged.
Using the -accel to specify accelerator is more graceful and more
feature such as kernel-irqchip，kvm-shadow-mem，tb-size，etc, can be
specified via domain xml if we continue doing this work, rather than
hard-writing code or using QEMU command line.
Thanks!
在 2021/1/13 23:06, huangy81(a)chinatelecom.cn 写道:
> From: Hyman <huangy81(a)chinatelecom.cn>
>
> This patchset aims to support dirty ring feature which has been
> introduced
> in kernel:
> https://lore.kernel.org/lkml/8b3f68dd-c61c-16a0-2077-0a5d3d57a357@redhat....
>
>
> The QEMU patchset is also being applied for merge, the review may be
> under
> the way:
> https://lore.kernel.org/qemu-devel/20210108164601.406146-1-peterx@redhat....
>
>
> Since QEMU enable the dirty ring feature by specifying the "-accel"
> sub-parameter.
> Libvirt use "-machine accel=xxx" option to specify the type of
> accelerator by
> default, which is conflict with QEMU. Either the Libvirt or QEMU may
> compromise,
>
> According to the Paolo Bonzini's patient and prompt explanation:
> https://lore.kernel.org/qemu-devel/3aa73987-40e8-3619-0723-9f17f73850bd@r...
>
> We'd like to have Libvirt switch to "-accel xxx" instead, cause the
> "-machine"
> options for accelerator are legacy and now there is a better mechanism