On Tue 10. Dec 2024 at 10:21, Roman Bolshakov <rbolshakov@ddn.com
<mailto:rbolshakov@ddn.com>> wrote:
On 10.12.2024 04:22, Philippe Mathieu-Daudé wrote:
> On 9/12/24 21:36, phil@philjordan.eu <mailto:phil@philjordan.eu>
wrote:
>> From: Phil Dennis-Jordan <phil@philjordan.eu
<mailto:phil@philjordan.eu>>
>>
>> This seems to be entirely superfluous and is costly enough to
show up in
>
> So the pthread_kill(cpu->thread, SIG_IPI) is entirely superfluous?
>
>> profiling. hv_vcpu_interrupt() has been demonstrated to very
reliably
>> cause VM exits - even if the target vCPU isn't even running, it will
>> immediately exit on entry.
>>
>> Signed-off-by: Phil Dennis-Jordan <phil@philjordan.eu
<mailto:phil@philjordan.eu>>
>> ---
>> target/i386/hvf/hvf.c | 2 +-
>> 1 file changed, 1 insertion(+), 1 deletion(-)
>>
>> diff --git a/target/i386/hvf/hvf.c b/target/i386/hvf/hvf.c
>> index 3b6ee79fb2..936c31dbdd 100644
>> --- a/target/i386/hvf/hvf.c
>> +++ b/target/i386/hvf/hvf.c
>> @@ -214,7 +214,7 @@ static inline bool
>> apic_bus_freq_is_known(CPUX86State *env)
>> void hvf_kick_vcpu_thread(CPUState *cpu)
>> {
>> - cpus_kick_thread(cpu);
>> + cpu->thread_kicked = true;
>> hv_vcpu_interrupt(&cpu->accel->fd, 1);
>> }
>
SIG_IPI is macOS crutch handled in XNU kernel that was essential until
Phil submitted proper kick support with hv_vcpu_interrupt().
Ah yes, perhaps it allowed exit from hv_vcpu_run(). hv_vcpu_run_until()
definitely does not exit early upon receiving SIG_IPI (USR1).