On 11/10/2018 19:12, Peter Maydell wrote: > On 11 October 2018 at 18:00, Philippe Mathieu-Daudé <phi...@redhat.com> wrote: >> On 11/10/2018 18:23, Peter Maydell wrote: >>> How many devices have a clock and nothing else that would cause >>> them to be non-user-creatable (ie no GPIOs, no IRQ lines, no >>> memory-mapped memory regions) ? >> >> I'm not sure I understood your question. > > Sorry, let me try rephrasing. > > We started with the question of whether devices with clocks should be > marked not user creatable. It's already the case that devices with any > of IRQs, GPIOs or memory-mapped registers can't be user created. > (In particular, any sysbus device is not user-creatable by default.) > So the question was intended to ask how often it matters whether > we can't wire up devices with clocks on the command line. > Are there any devices which would have clocks, but aren't *already* > ruled out of being user creatable for other reasons? If there > aren't any, it doesn't really matter much that we don't have > a mechanism for command line clock wiring.
Thank you for clearing this out. > > So I think the answer is: > * almost all users of this API framework are going to be > sysbus devices; those are already not user-creatable Yes. > * any device that uses this framework, and which is not > a sysbus device (but instead a plain old qdev device) > will need to mark itself as not-user-creatable by hand, > the same as if it uses the qdev gpio APIs OK, you answered my first question, it is now clearer to me :) Thank you! Phil.
