Re: [Qemu-devel] [RFC 09/10] cpu: remove exit_request global.

[Frederic Konrad <fred.konrad@greensocs.com>]

On 29/01/2015 16:52, Peter Maydell wrote:
> On 16 January 2015 at 17:19,  <fred.konrad@greensocs.com> wrote:
>> From: KONRAD Frederic <fred.konrad@greensocs.com>
>>
>> This removes exit_request global and adds a variable in CPUState for this.
>> Only the flag for the first cpu is used for the moment as we are still with one
>> TCG thread.
>> --- a/cpus.c
>> +++ b/cpus.c
>> @@ -646,10 +646,14 @@ static void cpu_handle_guest_debug(CPUState *cpu)
>>
>>   static void cpu_signal(int sig)
>>   {
>> +    CPUState *cpu;
>>       if (current_cpu) {
>>           cpu_exit(current_cpu);
>>       }
>> -    exit_request = 1;
>> +
>> +    CPU_FOREACH(cpu) {
>> +        cpu->exit_loop_request = 1;
>> +    }
>>   }
> You can't do this -- this code is a signal handler so it could
> get run at any time including while the list of CPUs is being
> updated. (This is why we have the exit_request flag in the
> first place rather than just setting the exit_request flag in
> each CPU...)
>
> Possibly you want exit_request to be a per-thread variable,
> but I haven't thought much about it.
>
>> --- a/include/qom/cpu.h
>> +++ b/include/qom/cpu.h
>> @@ -249,6 +249,7 @@ struct CPUState {
>>       bool created;
>>       bool stop;
>>       bool stopped;
>> +    volatile sig_atomic_t exit_loop_request;
>>       volatile sig_atomic_t exit_request;
>>       uint32_t interrupt_request;
>>       int singlestep_enabled;
> This would duplicate the exit_request and
> exit_loop_request flags in the CPU, which is kind of odd.
>
> -- PMM
Actually, what we want to do is remove exit_requested global because 
when it exits
the loop in tcg_exec_all it does exit_requested = 0, and other vcpu 
doesn't exit.

This is not clear to me why we have both exit_requested global and 
exit_request in
CPUState.

Maybe we can just make exit_request thread local this might work.

Also, we need to be able to exit VCPU we want from a TCG thread.
eg: We want to flush all the tlb we need to exit all cpus but one..

Thanks,
Fred