On Mon, Jun 21, 2021 at 03:35:02PM -0700, Atish Patra wrote:
On Mon, Jun 21, 2021 at 1:53 PM Grant Likely grant.likely@arm.com wrote:
On 21/06/2021 18:35, Atish Patra wrote:
On Mon, Jun 21, 2021 at 10:19 AM Grant Likely grant.likely@arm.com wrote:
On 10/05/2021 18:37, Atish Patra wrote:
[...]
+UEFI Boot at S mode +^^^^^^^^^^^^^^^^^^^^^^
+Most systems are expected to boot UEFI at S mode as the hypervisor extension [RVHYPSPEC]_ is +still in draft state.
I just noticed this. If the hypervisor spec is still in draft, then why is it being referenced by EBBR? Shouldn't that spec be made final first, or at least commonly deployed?
The hypervisor spec is still in draft but the details of privilege modes will not change. In fact, there has been almost zero change in the last one year. It has been in draft stage for various other reasons but not related to the technical ones ;) It's already been implemented in an FPGA and a working KVM port is available [1]. We are hoping that it will be merged by the end of this year.
Draft specs are obviously an area where we should tread with caution.
That said I think there could be a case for making an exception:
* The surrounding ecosystem is relatively mature with multiple software implementatons (KVM and Xen) together simulators (plural) and an FPGA-based hardware implementation.
* Draft spec updates are infrequent since it is almost like the H extension spec is in a holding pattern waiting for additional features to land elsewhere (primarily a virtualizable external interrupt controller).
* The proposed language in EBBR is far away from the contentious/unconcluded parts of the spec and, conveniently is guidance (should) rather than mandatory (must).
* We can clearly reference a versioned spec even if is draft.
* The glacial progress of the H extension is notorious enough to make the news ;-)
Daniel.