Note |
---|
This page will be updated during the month and is not final until after the 1st week of December. |
Summary of Month
...
Highlights
Prototype a kernel level driver for ivshmem that can exchange messages with QEMU’s virtio-msg-ivshmem mode
...
Held a sprint between AMD, Arm, and Linaro the first week of November
An initial implementation of virtio-msg via FFA for Xen DOMU to Xen DOM0
AMD has used the Linaro kernel prototypes on real HW for X86 + ARM via PCIe
Issues/Changes
1 week of conference and 1 week of vacation limited the amount of work in November
Jira Legacy | ||||||||||||
---|---|---|---|---|---|---|---|---|---|---|---|---|
|
...
(Intro slides on front page of project were updated)
Meeting notes
...
Follow up on mailing list virtio-msg@lists.linaro.org
Any issues?
Making the working spec available to all
Bertrand has published host first darft of the markdown format to the list
Can we host this markdown format in git somewhere? (w/ images etc)
Can we publish a snapshot of the PDF on the HVAC home page util another form is ready?
Updates from Bill
Updates from Edgar?
Updates from Lei?
Updates from others?
Actions:
Meeting notes October 24th
...
Notice: no meeting on Nov 7th
...
Updates from Bill
I have a kernel that can send a GET_STATUS msg to Edgar’s QEMU on another machine and receive the response.
Next steps: deferred RX processing and hooking to Viresh’s virtio-msg common code
Note: this code is ugly and prototype quality, Edgar and I need to agree on common shared memory format that will remove hard-coded values
Corresponding system level code is here. Each of these branches (kernel and demo) will be rebased as I go along.
Note: This will be the first time we interoperate between Authors. So far we have tested Edgar to Edgar and Viresh to Viresh
...
Updates from Edgar?
...
Updates from Lei?
...
Updates from Viresh:
Updated ffa driver Lei started, added memory sharing interface
Starting work on the interface to FFA DMA hal, tring to make it generic
Lorenzo may have some patches to integrate FFA indirect message support in the kernel
Starting to work with Bertrand’s Xen work which should allow to do any DOM to any DOM (including DOMU to DOMU)
No API for getting FFA messages to user space (needed on device side)
Need to discuss in Dublin
Discuss meeting timings, its getting very late for India (22:30 IST) from Nov, possible to move it two hours earlier ?
Updates from Armelle
...
Virtio-msg and OPTEE? Joakim/Sumit: need to discuss in Dublin
...
OPTEE as VM
Keep SOC vendor in real TZ
Allow SW provider to use OPTEE in a pKVM VM
[Joakim, Something about secure timer → Armelle uses protected storage for this, timer only needs slow update (10s)]
HW Crypto keeps track of key expiration
...
November 7
No meeting due to conflict with Linaro employee meetings
Meeting notes November 21
Updates from Bill
Any update on the notes from the Sprint?: WIP Document is here
Updates from Edgar?
Setup Dom0 on x86 w/ virtio-msg in kernel connecting ARM
Device side on Arm w/o Xen
Mostly working but loosing some interrupts are lost Versal to X86
Performance ~= 100MB Ethernet right now
Versal is mapping X86 memory as normal memory non-cached
Should work with Domu if versal PCI device is assigned to DomU
Will send an email w/ code pointers today or tomorrow
Versal side QEMU is using uio for shared memory and RX irq (send IRQ is devmem)
Updates from Lei?
Updates from Viresh?
Using Bertrahm Xen patches
Using patches for
Single QEMU system level w/ Xen
DOM0 backend in Rust for VSOCK
DomU kernel is using
/dev/virtio read & write messages (all messages including bus messages)
Bertrand: bus level messages should be processed in kernel
Alex
has setup a fork of the virtio spec on Linaro github w/ automated rendering
Updates from Armelle
Trusty is FFA compliant
Including Viresh work
Were missing UUID support but fixing that now and hope to enumerate devices
All new code in Trusty is Rust
Testing with Host not Guest VM
Arnaud
focused on co-processor
First working on kernel side, with Devicetree driven flow
Single message at a time in flat memory
No Zephyr code yet: maybe next week
Scaleable enumeration? : DT fragments but we don’t know the status
Updates from others?