Versions Compared

Old Version 54

changes.mady.by.user . Anmol

Saved on

compared with

New Version 55

changes.mady.by.user Jerry, Jiancong Zhao

Saved on

Key

  • This line was added.
  • This line was removed.
  • Formatting was changed.

...

  • Quick review of 1st workshop discussion*
      *For the 1st workshop meeting notes, please find at https://confluence.automotivelinux.org/display/VE/Meeting+Agenda#MeetingAgenda-Jan20,2021
  • Discussion on two methods to achieve zero-copy: dynamic vs dedicated heap
    • GPU/CPU cache coherency challenges 
    • Host allocation (share hostmem with guest) vs Guest allocation 
      • ChromeOS uses dynamic hostside allocation 
      • does host allocation and sharing introduce security concerns?
    • (over to Parag, OpenSynergy)
    • Dynamic vs Dedicated hostside allocation
      • where in host memory needs to be exposed to guest
      • host also has knowledge of where the real HW can see
    • Dedicated GPU heap (VRAM):
      • Idea: have a dedicated inter-vm shared mem area, guest requests buf params from hosts and allocates objects in shared memory and "sends" objects to host (paddrs)
      • What is needed to be done:
        • virtio gpu kernel driver adaptations → allocations from vram will be needed as well
        • MESA3d adaptations?
        • virgl adaptations
    • Can VFIO on Intel be applicable to the zero-copy case?
    • Type 1/2 HV: does the problem also apply to T2 HV?
      • There should be no difference between T1/2 HV, the same problem exists.
    • Xen HV:
      • QEmu - generic virtio backend for Xen
      • Virtio patches are being discussed
    • Updates from GPU workshop #1
      • Stride requirements were found (for RCar), hardcoded, but visual artifacts can be seen → activity postponed
  • Discuss on how we should move forward
  • Slides:
    • View file
      namevirtiogpu3d_workshop-f1.pdf
      height250


March 3, 2021

Attendees:

  • Jerry Zhao - Panasonic
  • Jan-Simon Moeller - Linux Foundation
  • Scott Murray - Konsulko
  • Mikhail Golubev - Open Synergy
  • Vasyl Vavrychuk - OpenSynergy
  • Andriy Tryshnivsky - OpenSynergy
  • Laurent Cremmer - Carmeq
  • Mark Silberberg - Volkswagen
  • Victor Duan - Linaro
  • Francis Ozog - Linaro
  • Masami Hiramatsu - Linaro
  • Alex Bennée (Stratos Tech Lead) - Linaro
  • Peter Griffin (Multimedia Tech Lead) - Linaro 
  • Tadao Tanikawa - Panasonic
  • Binghua - Qualcomm
  • Marius Vlad - Collabora
  • Tadao Tanikawa - Panasonic
  • Kenji Hosokawa - ADIT

Agenda: 

  • GSoC Proposal from Virt-EG
    • Link: https://wiki.linuxfoundation.org/gsoc/2021-gsoc-agl

    • If any EG member interested, please provide Project Name, Task Details, Exprience Level (Prerequisite Skill), Contact of Mentor by  

      Info

      • Experience Level: Description

      • Beginner:

        • Task #1: webcam demo app - write an app exposing the v4l webcam feed

          • your task will be to develop an QT/QML-based application to display the a webcam feed - talk to jsmoeller (at) linuxfoundation.org

      • Intermediate:

        • Task #2: demo app for the IIO service. Display sensor data from the iio-service on the display.

          • Your task will be to develop an QT/QML-based application to show data from sensors - talk to jsmoeller (at) linuxfoundation.org

      • Advanced:

        • Task #3: Platform Development - create or enhance the container image

          • Your task will be to either create or enhance a container-based setup using AGL - talk to jsmoeller (at) linuxfoundation.org

      • Advanced++:

        • Task #4 integration of ROS (Robot OS - meta-ros) - talk to jsmoeller (at) linuxfoundation.org

        • Task #5 extend the jailhouse support


  • VirtIO KK Release
  • All Member Meeting (Mar 16~17)
  • Follow-up of last workshop on VirtIO-gpu
    • Can VFIO on Intel be applicable to the zero-copy case?
      • Which address space are buffers in? Can we use IOMMU for that?
      • IOMMU is a must have to defeat against DMA attacks

      • Intel pioneered VFIO; Huawei did develop so called "warpdrive" technology

        Info
        titleMail from Kenneth (Huawei)

        Hello, Jerry,

        Nice meeting you. We are happy to discuss UADK with colleagues of the same profession. We are just afraid that we cannot give what you need;).  Because it seems you are looking for VFIO rather than UADK. We tried to merge our requirement to VFIO initially. But we gave up after the first RFC. To us, the major down side of VFIO is that it is a practical device. But we don’t take it a device. We take it a heterogeneous CPU. In UADK, the process attach to the accelerator, the accelerator share whole or part of the address space of the process. So the process can deliver task to the accelerator without warrying how to copy data to the accelerator. So there is no independent device is created, no DNA operation is provided. It is just for sharing memory between CPU and the accelerator. Because of this, we solve different set of problem. E.g. we have to take care of process-cloning to duplicate/stop the interface to child process.

        But it seem you are looking for a solution to isolate some GPU resource for particular VM. This is exactly what VFIO trying to do.

        So now it is you who may want to tell me. Are you guys really want to know the detail of UADK? J

        Cheers


  • Future Activity of EG
    • Welcome proposal from members