On 8/7/21 8:16 AM, Greg Favor wrote:
On Fri, Aug 6, 2021 at 10:41 PM Anup Patel <anup.patel@... <mailto:anup.patel@...>> wrote:
I think it should be enough to require that the mstatus.MBE
field is set
to 0 and not hardwired.
[Anup] There was quite a bit of discussion in separate email thread
it was concluded to include the statement way it is. The grammar
do you have a link to the thread?
should be definitely fixed.
It now strikes me that requiring 2022-compliant platforms to hardware MBE to zero will result in future platform specs - that allow MBE to be writeable - to not be compatible with the 2022 platform spec. This I believe goes against the intent that a platform that is compliant with a given year's platform spec should inherently also be compatible with all earlier years of the platform spec (modulo any DEPRECATED features).
If this intent is correct, then it seems like we must not require MBE to be hardwired and instead need to say that it must be set to 0.
[Anup] For first release of RISC-V platform spec, only little-endian
is allowed. Future release of RISC-V platform spec will certainly allow
big-endian mode but we need software support first (i.e. big-endian
kernel and distro/rootfs).