On 11 Jan 2021, at 18:58, Alex Bennée alex.bennee@linaro.org wrote:
Yang Zhang yang.zhang@96boards.org writes:
Hi Alex
I am in the rust-vmm camp so...
I think rust-vmm is certainly worth exploring (I'm slowly re-implementing virtio-rpmb while learning rust). However I think it will be some time before there is a feature equivalent rust based VMM that can do everything QEMU can.
Agreed. Hence agreeing on the benchmark proposal.
I should also note there has long been an interest expressed in a hybrid build of QEMU which uses Rust code for device models. As of yet no one has written such a thing though.
Having said that a slimmed down version of QEMU for arm also makes sense. Do you have any suggestions wrt the spec for benchmark purpose? Would you be proposing to look at virtio storage/network devices mainly?
Yes - similar to the x86 NEMU build. Something that just supports -M virt (or something like it) and the core virtio devices (net/block/rng) and without any TCG baggage. It would still support QMP/migration and the rest of the features useful for the cloud.
This slimmed down spec makes sense to me.
Thanks Yang
On Mon, 11 Jan 2021 at 15:49, Alex Bennée alex.bennee@linaro.org wrote:
Hi,
As Linaro Big Ideas aren't meant to hang around too long I wanted to do a call to see if there was any interest in taking this further. Feedback so far has been it look interesting and certainly looks useful from the upstream point of view but generally members have other plans (rust-vmm based solutions, kvmtool etc). From an engineering point of view I still think there is use in having a slimmed down build of QEMU for ARM if only to have something to benchmark against when developing the new shiny tools.
So what do people think?
Worth having a call and firming up an initiative for voting on or clsong the card due to lack of interest?
-- Alex Bennée
-- Alex Bennée