Re: [Qemu-devel] [PATCH v12 12/24] tcg: handle EXCP_ATOMIC exception for system emulation

[Richard Henderson]

On 02/13/2017 11:10 PM, Alex Bennée wrote:
> @@ -239,9 +240,16 @@ static void cpu_exec_step(CPUState *cpu)
>                       1 | CF_NOCACHE | CF_IGNORE_ICOUNT);
>      tb->orig_tb = NULL;
>      tb_unlock();
> -    /* execute the generated code */
> -    trace_exec_tb_nocache(tb, pc);
> -    cpu_tb_exec(cpu, tb);
> +
> +    cc->cpu_exec_enter(cpu);
> +
> +    if (sigsetjmp(cpu->jmp_env, 0) == 0) {
> +        /* execute the generated code */
> +        trace_exec_tb_nocache(tb, pc);
> +        cpu_tb_exec(cpu, tb);
> +    }

I don't understand this, since cpu_tb_exec has its own sigsetjmp.  Where is the 
exception supposed to come from that escapes?

> +                } else if (r == EXCP_ATOMIC) {
> +                    qemu_mutex_unlock_iothread();
> +                    cpu_exec_step_atomic(cpu);
> +                    qemu_mutex_lock_iothread();
...
> +            case EXCP_ATOMIC:
> +                qemu_mutex_unlock_iothread();
> +                cpu_exec_step_atomic(cpu);
> +                qemu_mutex_lock_iothread();


I just noticed this, but if you have to do a v13, it might be best to move 
these locks inside cpu_exec_step_atomic, as with tcg_cpu_exec.  Otherwise leave 
it for later.


r~