[PATCH v2 0/3] bsp/xen: support for Xen hypervisor

Jeff Kubascik jeff.kubascik at dornerworks.com
Thu Dec 5 13:41:18 UTC 2019


On 12/5/2019 1:36 AM, Sebastian Huber wrote:
> Thanks, I check in the patch set.

Excellent, thank you!

> Should the BSP be renamed from "xen_virtual" to just "xen"?

I was on the fence with this. This specific BSP is designed to be purely
virtualized, with no dependence on the underlying hardware platform.
Theoretically, it should be completely portable across ARMv8 platforms with Xen.

Xen does allow the passthrough of peripherals with very little overhead, through
the 2-stage MMU, GIC virtualization extensions, and the SMMU of the ARMv8
architecture, which is great for embedded systems. My thought is that in the
future, additional BSPs could be created that include drivers for their
respective platforms, e.g. xen_zynqmp. However, it may make more sense to put
these types of BSPs in their respective platform directories instead.

I have been looking into using AArch64 qemu to simulate RTEMS on Xen and
realized that there is one big caveat to this portability, and that is GICv2
versus GICv3. The GICv2 implementation in qemu doesn't currently work with Xen,
but GICv3 does. The two are incompatible and require different drivers (the
exception is GICv3 does have optional features to allow for some backwards
compatibility, but this is not implemented on qemu). What I am thinking is the
current BSP should be instead renamed to xen_gicv2, with a xen_gicv3 BSP to be
implemented in the future.

On a related note, do you know what the current state is for a GICv3 driver in
RTEMS? Stew reached out on IRC a few days ago, and it sounded like there should
be a patch floating around on the mailing list. However, I have been unable to
locate it.

> --
> Sebastian Huber, embedded brains GmbH
> 
> Address : Dornierstr. 4, D-82178 Puchheim, Germany
> Phone   : +49 89 189 47 41-16
> Fax     : +49 89 189 47 41-09
> E-Mail  : sebastian.huber at embedded-brains.de
> PGP     : Public key available on request.
> 
> Diese Nachricht ist keine geschäftliche Mitteilung im Sinne des EHUG.
> 

Sincerely,
Jeff Kubascik


More information about the devel mailing list