Re: [PATCH v3 5/8] Add memory hardware address read/write API

[Pierrick Bouvier <pierrick.bouvier@linaro.org>]

On 5/21/25 2:43 AM, Rowan Hart wrote:
> From: novafacing <rowanbhart@gmail.com>
> 
> Signed-off-by: novafacing <rowanbhart@gmail.com>
> Signed-off-by: Rowan Hart <rowanbhart@gmail.com>
> ---
>   include/qemu/qemu-plugin.h |  96 +++++++++++++++++++++++++++++++++++
>   plugins/api.c              | 100 +++++++++++++++++++++++++++++++++++++
>   2 files changed, 196 insertions(+)

Reading this patch, and patch 3 (Add address space API), I am not sure 
AddressSpace is something we want to leak in plugins interface.
It is a concept *very* internal to QEMU, and not reflecting directly 
something concerning the emulated architecture (it is related, but not 
officially described for it).

The same way qemu_plugin_write_memory_vaddr is only valid in the current 
page table setup, we could assume the same for current address space, 
and return an error if memory is not mapped with current AS.
Eventually, we could read/write a given hwaddr in all existing address 
spaces (starting with current mapped one), if it makes sense to do this, 
which I'm not sure about.

What are your thoughts on this?

qemu_plugin_translate_vaddr is fine for me.