Cc'ing Paolo
On 14/11/23 13:25, Philippe Mathieu-Daudé wrote:
> On 13/11/23 20:52, David Woodhouse wrote:
>> On Mon, 2023-11-13 at 16:21 +0100, Philippe Mathieu-Daudé wrote:
>>> "sysemu/xen.h" defines CONFIG_XEN_IS_POSSIBLE as a target-agnostic
>>> version of CONFIG_XEN. Use it in order to use "sysemu/xen-mapcache.h"
>>> in target-agnostic files.
>>>
>>> Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
>>
>> Noting that CONFIG_XEN_IS_POSSIBLE is for Xen accelerator support, and
>> may not be set in all cases when we're hosting Xen-compatible guests,
>
> As is CONFIG_XEN.
Actually we have:
- host Xen support
. CONFIG_XEN_BACKEND
- QEMU Xen accelerator
. CONFIG_XEN
- HW models
. CONFIG_XEN (old / generic models)
. CONFIG_XEN_BUS
. XEN_IGD_PASSTHROUGH (PCI stuff)
. CONFIG_XEN_EMU (Xen on KVM)
Paolo, David, is that correct?
When can we have CONFIG_XEN without CONFIG_XEN_BACKEND
(and vice-et-versa)?
So for clarity CONFIG_XEN could be split as:
- CONFIG_ACCEL_XEN
- CONFIG_XEN_MACHINE_FV (Fully-virtualized)
- CONFIG_XEN_MACHINE_PV (Para-virtualized)
> Maybe be worth renaming CONFIG_ACCEL_XEN if you think we need
> guest hw specific CONFIG_foo_XEN variables.
>
>> Reviewed-by: David Woodhouse <dwmw@amazon.co.uk>
>
> Thanks!