Hi Jesse, We aren't quite ready to share the Qemu patches just yet. They are currently very rough proof-of-concept patches and we are currently working on getting them cleaned up a bit and modified to match some preferences stated by the maintainers. As soon as they are in reasonable shape, we will share. Thanks, -Maran On 4/6/2018 12:03 PM, Larrew, Jesse wrote:
Hi Maran,
I'd like to give this a try. Are there qemu patches to go along with these?
Sincerely, Jesse
-----Original Message----- From: Boeuf, Sebastien [mailto:sebastien.boeuf@intel.com] Sent: Monday, April 2, 2018 3:46 PM To: Maran Wilson <maran.wilson@oracle.com>; kata- dev@lists.katacontainers.io Cc: jesse.butler@oracle.com Subject: Re: [kata-dev] Fast booting KVM guests w/o firmware and Qemu pc- lite
Thanks Maran,
This would indeed push Kata in the right direction regarding the optimizations needed for boot time.
Thanks, Sebastien ________________________________________ From: Maran Wilson [maran.wilson@oracle.com] Sent: Monday, April 02, 2018 1:40 PM To: kata-dev@lists.katacontainers.io Cc: jesse.butler@oracle.com Subject: [kata-dev] Fast booting KVM guests w/o firmware and Qemu pc-lite
There was some discussion about Qemu pc-lite vs upstream Qemu in the Architecture Committee meeting today. As Jesse Butler briefly mentioned on Slack, a few of us have been working on upstreaming a different approach to starting KVM guests without firmware that should be more acceptable for upstream maintainers. We are treating it as a more general purpose interface, but hopefully this framework will go a long way towards closing the gap between pc-lite and upstream Qemu in terms of guest startup time for solutions like Kata containers.
We have already posted a set of Linux kernel patches upstream that seem to be well received so far by the community. See https://lkml.org/lkml/2018/3/20/717 for the latest version. We are currently working on cleaning up the corresponding Qemu patches and getting them ready for upstream review as well. Based on the RFC discussions we have had with upstream maintainers regarding high level design, I'm cautiously optimistic that we won't run into any deal breakers when we do post the qemu patches.
The high level idea is that we could use this approach to achieve the same goal of booting directly to the uncompressed Linux kernel (without running firmware), but in a way that is acceptable to the upstream community since we aren't introducing any new Linux OS specific (zero-page) initialization in Qemu.
So for anyone interested in the Qemu/KVM side of Kata containers and how we might be able to start VMs more quickly, please take a look at the patches we have posted so far and feel free to join the discussion on the upstream mailing lists.
Thanks, -Maran
_______________________________________________ kata-dev mailing list kata-dev@lists.katacontainers.io http://lists.katacontainers.io/cgi-bin/mailman/listinfo/kata-dev
_______________________________________________ kata-dev mailing list kata-dev@lists.katacontainers.io http://lists.katacontainers.io/cgi-bin/mailman/listinfo/kata-dev