On 3/18/25 17:42, Pierrick Bouvier wrote:
> On 3/18/25 14:32, Richard Henderson wrote:
>> Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
>> ---
>> include/exec/icount.h | 10 +++++++++-
>> 1 file changed, 9 insertions(+), 1 deletion(-)
>>
>> diff --git a/include/exec/icount.h b/include/exec/icount.h
>> index 4964987ae4..7a26b40084 100644
>> --- a/include/exec/icount.h
>> +++ b/include/exec/icount.h
>> @@ -22,13 +22,21 @@ typedef enum {
>> ICOUNT_ADAPTATIVE,
>> } ICountMode;
>> -#if defined(CONFIG_TCG) && !defined(CONFIG_USER_ONLY)
>> +#ifdef CONFIG_TCG
>> extern ICountMode use_icount;
>> #define icount_enabled() (use_icount)
>> #else
>> #define icount_enabled() ICOUNT_DISABLED
>> #endif
>> +/* Protect the CONFIG_USER_ONLY test vs poisoning. */
>> +#if defined(COMPILING_PER_TARGET) || defined(COMPILING_SYSTEM_VS_USER)
>> +# ifdef CONFIG_USER_ONLY
>> +# undef icount_enabled
>> +# define icount_enabled() ICOUNT_DISABLED
>> +# endif
>> +#endif
>> +
>> /*
>> * Update the icount with the executed instructions. Called by
>> * cpus-tcg vCPU thread so the main-loop can see time has moved forward.
>
> I understand the shortcut taken here, but I'm not sure we want to start having specifics
> for COMPILING_SYSTEM_VS_USER, out of the poison file.
>
> In this case, we can change icount_enabled() to a function, implement it in accel/tcg/
> icount-common.c (which is system only), and add a stub for user mode in accel/tcg/icount-
> user.c (or common-user/icount.c), returning ICOUNT_DISABLED, or 0, more simply.
(1) If I were to do anything it would not be adding a function, but
making use_icount a universal variable.
(2) There are a few uses where if (!icount_enabled()) needs to
evaluate to false for user-mode, so that dead code elimination happens.
Otherwise we then have to introduce extra stubs, etc, etc, etc.
r~