Re: [PATCH 01/13] plugins: change signature of qemu_plugin_insn_haddr

[Richard Henderson]

On 12/16/24 19:06, Pierrick Bouvier wrote:
> It makes more sense to return the same type than qemu_plugin_insn_vaddr.
>
> Signed-off-by: Pierrick Bouvier <pierrick.bouvier@linaro.org>
> ---
>   include/qemu/qemu-plugin.h |  2 +-
>   plugins/api.c              | 12 ++++++------
>   2 files changed, 7 insertions(+), 7 deletions(-)

No, it does not.

qemu_plugin_insn_vaddr is returning a guest virtual address.
qemu_plugin_insn_haddr is returning a host address.

I'm not sure why we decided that returning a host pointer was a good idea.  Probably it 
was the easiest thing to retrieve from softmmu.

One could argue that we should be returning something else, the only question 
is what.

Perhaps guest physical address, which wasn't possible before, but which is now stored 
within CPUTLBEntryFull. Interpreting this requires you to know the physical address space 
to which it applies. In the case of Arm, the address space varies depending on Secure vs 
Non-Secure state.

Perhaps ram_addr_t, which is *not* a guest physical address because it is not associated 
with any address space. It is more of a globally unique token with which a RAMBlock may be 
found.  It's how we stitch together address spaces under the hood.  The plugin would have 
to treat it as an opaque unique identifier.

But if we're going to return a host address, then void* is the correct type.


r~