Re: [PATCH 03/12] tests/plugin: add test plugin for inline operations

[Alex Bennée]

Pierrick Bouvier <pierrick.bouvier@linaro.org> writes:

> On 1/11/24 19:57, Philippe Mathieu-Daudé wrote:
>> Hi Pierrick,
>> On 11/1/24 15:23, Pierrick Bouvier wrote:
>>> For now, it simply performs instruction, bb and mem count, and ensure
>>> that inline vs callback versions have the same result. Later, we'll
>>> extend it when new inline operations are added.
>>>
>>> Use existing plugins to test everything works is a bit cumbersome, as
>>> different events are treated in different plugins. Thus, this new one.
>>>
>>> Signed-off-by: Pierrick Bouvier <pierrick.bouvier@linaro.org>
>>> ---
>>>    tests/plugin/inline.c    | 183 +++++++++++++++++++++++++++++++++++++++
>>>    tests/plugin/meson.build |   2 +-
>>>    2 files changed, 184 insertions(+), 1 deletion(-)
>>>    create mode 100644 tests/plugin/inline.c
>> 
>>> +#define MAX_CPUS 8
>> Where does this value come from?
>> 
>
> The plugin tests/plugin/insn.c had this constant, so I picked it up
> from here.
>
>> Should the pluggin API provide a helper to ask TCG how many
>> vCPUs are created?
>
> In user mode, we can't know how many simultaneous threads (and thus
> vcpu) will be triggered by advance. I'm not sure if additional cpus
> can be added in system mode.
>
> One problem though, is that when you register an inline op with a
> dynamic array, when you resize it (when detecting a new vcpu), you
> can't change it afterwards. So, you need a storage statically sized
> somewhere.
>
> Your question is good, and maybe we should define a MAX constant that
> plugins should rely on, instead of a random amount.

For user-mode it can be infinite. The existing plugins do this by
ensuring vcpu_index % max_vcpu. Perhaps we just ensure that for the
scoreboard as well? Of course that does introduce a trap for those using
user-mode...

-- 
Alex Bennée
Virtualisation Tech Lead @ Linaro