> On 14 Jan 2021, at 12:56, Luc Michel <luc.mic...@greensocs.com> wrote:
>
> ... But yes the goal here is to rely on QEMU's already existing device models
> to dynamically create a machine.
I did not check the current QEMU device models, so I might be wrong, but I
think that the scope might be slightly different.
In my fork I read the JSON with the detailed peripherals definitions, which
include not only memory address, but registers, sizes, alignment, bit-fields,
access right, reset values, etc, and from these definitions I automatically
create the QEMU objects.
The result is a device with all registers available, you can already read/write
all of them, just that there are no actions associated yet, and these actions
need to be added manually.
There is also a tool that generates the C code for this, so basically I can
turn a CMSIS SVD definition into a complete QEMU device, partly functional
(without actions).
I plan to further improve this mechanism, but it'll probably take some time.
Regards,
Liviu