[Rust-VMM] Add Xen support to vmm-reference? (was Re: Call for topics and skipping next weeks sync)

Fri Jul 15 10:59:42 UTC 2022

+ Oleksandr Tyshchenko

-- Artem
________________________________
From: Alex Bennée <alex.bennee at linaro.org>
Sent: Wednesday, July 13, 2022 6:39:23 PM
To: Trilok Soni <tsoni at quicinc.com>
Cc: Stratos Mailing List <stratos-dev at op-lists.linaro.org>; Viresh Kumar <viresh.kumar at linaro.org>; Mathieu Poirier <mathieu.poirier at linaro.org>; Mike Holmes <mike.holmes at linaro.org>; Azzedine Touzni <atouzni at qti.qualcomm.com>; Stefano Stabellini <stefano.stabellini at xilinx.com>; Christopher Clark <christopher.w.clark at gmail.com>; Arnd Bergmann <arnd.bergmann at linaro.org>; Peter Griffin <peter.griffin at linaro.org>; AKASHI Takahiro <takahiro.akashi at linaro.org>; Artem Mygaiev <Artem_Mygaiev at epam.com>; Leonardo Garcia <leonardo.garcia at linaro.org>; Randy Linnell <randy.linnell at linaro.org>; don.harbin at linaro.org <don.harbin at linaro.org>; Sumit Semwal <sumit.semwal at linaro.org>; xen-devel at lists.xenproject.org <xen-devel at lists.xenproject.org>; Wei Liu <wl at xen.org>; Florescu, Andreea <fandree at amazon.com>; rust-vmm at lists.opendev.org <rust-vmm at lists.opendev.org>
Subject: Add Xen support to vmm-reference? (was Re: Call for topics and skipping next weeks sync)

(added xen/rust-vmm to CC)

Trilok Soni <tsoni at quicinc.com> writes:

> Hello Alex,
>
> It would be good to check if there is a enough interest to make
> "vmm-reference" extended to work w/ Type-1 Hypervisor like Xen. This
> will clean-up vmm-reference w/ KVM references, and it will be a good
> independent tool to test the new interfaces additions in the rust-vmm.
>
> https://urldefense.com/v3/__https://github.com/rust-vmm/vmm-reference__;!!GF_29dbcQIUBPA!0uj77Sl-hZVdR69XuAaDLI87ksTktSzE3-q2uL70xlReH_QbkDrjnvdm9fM0E5PeeK5iisOpQTs9Qf79mJWtnJ2Z$ [github[.]com]

We have certainly talked about it before and the we've put in some of
the groundwork with the xen-sys crate
(https://urldefense.com/v3/__https://github.com/rust-vmm/xen-sys__;!!GF_29dbcQIUBPA!0uj77Sl-hZVdR69XuAaDLI87ksTktSzE3-q2uL70xlReH_QbkDrjnvdm9fM0E5PeeK5iisOpQTs9Qf79mJRsjZzU$ [github[.]com]) which we are using for the
xen-vhost-master work. Off the top of my head we would need to work out:

  - Which additional APIs need implementing for VMM lifecycle management
  - How the run loop would look
    - AIUI Hyper-V munges this for rust-vmm by providing a KVM-like
      ioctl API (POC via CrosVM: https://urldefense.com/v3/__https://github.com/petrutlucian94/crosvm/tree/windows__;!!GF_29dbcQIUBPA!0uj77Sl-hZVdR69XuAaDLI87ksTktSzE3-q2uL70xlReH_QbkDrjnvdm9fM0E5PeeK5iisOpQTs9Qf79mNH1mFFW$ [github[.]com])
    - See also https://urldefense.com/v3/__https://github.com/rust-vmm/community/issues/121__;!!GF_29dbcQIUBPA!0uj77Sl-hZVdR69XuAaDLI87ksTktSzE3-q2uL70xlReH_QbkDrjnvdm9fM0E5PeeK5iisOpQTs9Qf79mOE_Kgjx$ [github[.]com]
  - Would we aim for full independence from the exiting Xen userspace libs?
    - xen-vhost-master was brought up using them, now slowly being
      replaced with pure rust

I've added it as a discussion topic for the sync call:

  https://urldefense.com/v3/__https://calendar.google.com/event?action=TEMPLATE&tmeid=NW83NXNsZ2Q1NmF0bTE0dGNkM3Q0YjFrcTlfMjAyMjA4MDNUMTUwMDAwWiBjX2o3bmdpMW84cmxvZmtwZWQ0cjVjaDk4bXZnQGc&tmsrc=c_j7ngi1o8rlofkped4r5ch98mvg*40group.calendar.google.com__;JQ!!GF_29dbcQIUBPA!0uj77Sl-hZVdR69XuAaDLI87ksTktSzE3-q2uL70xlReH_QbkDrjnvdm9fM0E5PeeK5iisOpQTs9Qf79mIAXRpBH$ [calendar[.]google[.]com]

It would be great is we could get any interested folks from the rust-vmm
and Xen communities to come along to the call.

>
> ---Trilok Soni
>
> -----Original Message-----
> From: Alex Bennée <alex.bennee at linaro.org>
> Sent: Wednesday, July 13, 2022 2:16 AM
> To: Stratos Mailing List <stratos-dev at op-lists.linaro.org>
> Cc: Viresh Kumar <viresh.kumar at linaro.org>; Mathieu Poirier
> <mathieu.poirier at linaro.org>; Mike Holmes <mike.holmes at linaro.org>;
> Azzedine Touzni <atouzni at qti.qualcomm.com>; Stefano Stabellini
> <stefano.stabellini at xilinx.com>; Christopher Clark
> <christopher.w.clark at gmail.com>; Arnd Bergmann
> <arnd.bergmann at linaro.org>; Peter Griffin <peter.griffin at linaro.org>;
> AKASHI Takahiro <takahiro.akashi at linaro.org>; Artem Mygaiev
> <Artem_Mygaiev at epam.com>; Leonardo Garcia
> <leonardo.garcia at linaro.org>; Trilok Soni <tsoni at quicinc.com>; Randy
> Linnell <randy.linnell at linaro.org>; don.harbin at linaro.org; Sumit
> Semwal <sumit.semwal at linaro.org>
> Subject: Call for topics and skipping next weeks sync
>
> WARNING: This email originated from outside of Qualcomm. Please be wary of any links or attachments, and do not enable macros.
>
> Hi All,
>
> I'll be on holiday (moving house) next week so I won't be able to
> chair the Stratos sync meeting. As it is the middle of summer I'm
> going to propose we skip next weeks sync and re-convene on the 3rd of
> August. Any objections?
>
> While on the subject of sync meetings are there any topics to discuss.
> We've had some discussions on next rust-vmm device to implement but
> beyond a vague "maybe virtio-gpu to help with demos" I don't think
> we've nailed it down. virtio-can also keeps getting mentioned and
> while useful I'm wary it's not pushing our exploration of the
> possibilities of virtio further.
>
> I did a talk at GST22 last week which was an overview of VirtIO and what we had done so far as well as discussing some future directions. You can see the talk at:
>
>   https://urldefense.com/v3/__https://huawei-events.de/en/gsts22-j83dco-vod.htm__;!!GF_29dbcQIUBPA!0uj77Sl-hZVdR69XuAaDLI87ksTktSzE3-q2uL70xlReH_QbkDrjnvdm9fM0E5PeeK5iisOpQTs9Qf79mAohQusR$ [huawei-events[.]de]
>   (Day 2 stream, Chapter 6/TS 05:13:00)
>
> In it potential future areas of exploration where:
>
> Improve Xen API
> ═══════════════
>
>   • More standard mmap
>   • direct irqfd/eventfd routing
>
> Memory Isolation
> ════════════════
>
>   • fat virtqueue
>   • iommu/grants vs regions
>   (x-over with pKVM/CCA?)
>
> Bare metal rust
> ═══════════════
>
>   • re-use exiting VirtIO logic
>   • but without POSIX layer
>
> I'd like to get a better steer on what we should focus on next after we've demoed our existing rust-vmm daemons and the Xen vhost-master work.

--
Alex Bennée
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.opendev.org/pipermail/rust-vmm/attachments/20220715/4f566b03/attachment-0001.htm>