Re: [PATCH] plugins: add two events for cpu_restore_state_from_tb() and cpu_io_recompile()

[Alex Bennée]

Richard Henderson <richard.henderson@linaro.org> writes:

> On 9/2/24 10:52, Alex Bennée wrote:
>> Pierrick Bouvier <pierrick.bouvier@linaro.org> writes:
>> 
>>> Hi Xingran,
>>>
>>> On 9/2/24 03:42, Alex Bennée wrote:
>>>> Xingran Wang <wangxingran123456@outlook.com> writes:
>>>>
>>>>> Currently, the instruction count obtained by plugins using the translation
>>>>> block execution callback is larger than the actual value. Adding callbacks
>>>>> in cpu_restore_state_from_tb() and cpu_io_recompile() allows plugins to
>>>>> correct the instruction count when exiting a translation block
>>>>> mid-execution, properly subtracting the excess unexecuted
>>>>> instructions.
>>>> This smells like exposing two much of the TCG internals to the
>>>> plugin
>>>> mechanism. You can already detect when we don't reach the end of a block
>>>> of instructions by instrumentation as I did in:
>>>>
>>>
>>> I agree that this is definitely a QEMU implementation "detail", and
>>> should not be a concern for end users.
>>>
<snip>
>>    /**
>>     * qemu_plugin_register_vcpu_tb_exec_end_cb() - register execution 
>> callback at end of TB
>>     * @tb: the opaque qemu_plugin_tb handle for the translation
>>     * @cb: callback function
>>     * @flags: does the plugin read or write the CPU's registers?
>>     * @userdata: any plugin data to pass to the @cb?
>>     *
>>     * The @cb function is called every time a translated unit executes.
>>     */
>>    QEMU_PLUGIN_API
>>    void qemu_plugin_register_vcpu_tb_exec_end_cb(struct qemu_plugin_tb *tb,
>>                                                   
>> qemu_plugin_vcpu_tb_end_cb_t cb,
>>                                                   enum qemu_plugin_cb_flags 
>> flags,
>>                                                   void *userdata);
>> I think the tricky bit would be getting TCG to emit the callback
>> code
>> for the last instruction before the
>> tcg_gen_exit_tb/tcg_gen_lookup_and_goto_ptr bits but after whatever else
>> it has done to execute the instruction.
>> I don't think we could easily support inline ops at tb end though.
>> Richard,
>> What do you think?
> I think this will miss all exceptions raised in the middle of the block.
> I don't think it will be reliable at all.

Ahh yes - I guess we can't fixup as we go through cpu_loop_exit() and
the restore code is only called from helpers. Oh well I think we can
make do with what we currently have.

>
>
> r~

-- 
Alex Bennée
Virtualisation Tech Lead @ Linaro