[Qemu-devel] Windows build crash problem

[Qemu-devel] Windows build crash problem

[武田 俊也]

Hi all.

I encountered the crash problem with windows build on Core-Duo host.
Qemu crashes in cpu_interrupt() called from host_alarm_handler() in vl.c
because CPUState *env = next_cpu is null.

I tried the patch to check env is not null in host_alarm_handler()
and now qemu does not crash.

        CPUState *env = next_cpu;
+	if(env) {
        /* stop the currently executing cpu because a timer occured */
        cpu_interrupt(env, CPU_INTERRUPT_EXIT);
#ifdef USE_KQEMU
        if (env->kqemu_enabled) {
            kqemu_cpu_interrupt(env);
        }
#endif
+	}
        event_pending = 1;
}

This is very temporary patch and I need to investigate
why next_cpu is null but sorry now I dont have enough time.

Thanks,
TAKEDA, toshiya

Re: [Qemu-devel] Windows build crash problem

[Filip Navara]

[-- Attachment #1: Type: text/plain, Size: 1041 bytes --]

Actually  somebody has already described the problem on the mailing list and
Paul Brook replied to that message. Hopefully you can find it in the
archives.

F.

On Dec 12, 2007 5:25 PM, 武田 俊也 <t-takeda@m1.interq.or.jp> wrote:

> Hi all.
>
> I encountered the crash problem with windows build on Core-Duo host.
> Qemu crashes in cpu_interrupt() called from host_alarm_handler() in vl.c
> because CPUState *env = next_cpu is null.
>
> I tried the patch to check env is not null in host_alarm_handler()
> and now qemu does not crash.
>
>        CPUState *env = next_cpu;
> +       if(env) {
>        /* stop the currently executing cpu because a timer occured */
>        cpu_interrupt(env, CPU_INTERRUPT_EXIT);
> #ifdef USE_KQEMU
>        if (env->kqemu_enabled) {
>            kqemu_cpu_interrupt(env);
>        }
> #endif
> +       }
>        event_pending = 1;
> }
>
> This is very temporary patch and I need to investigate
> why next_cpu is null but sorry now I dont have enough time.
>
> Thanks,
> TAKEDA, toshiya
>
>
>
>

[-- Attachment #2: Type: text/html, Size: 1521 bytes --]

Re: [Qemu-devel] Windows build crash problem

[andrzej zaborowski]

Hi,

On 12/12/2007, 武田 俊也 <t-takeda@m1.interq.or.jp> wrote:
> I encountered the crash problem with windows build on Core-Duo host.
> Qemu crashes in cpu_interrupt() called from host_alarm_handler() in vl.c
> because CPUState *env = next_cpu is null.
>
> I tried the patch to check env is not null in host_alarm_handler()
> and now qemu does not crash.
>
>         CPUState *env = next_cpu;
> +       if(env) {
>         /* stop the currently executing cpu because a timer occured */
>         cpu_interrupt(env, CPU_INTERRUPT_EXIT);
> #ifdef USE_KQEMU
>         if (env->kqemu_enabled) {
>             kqemu_cpu_interrupt(env);
>         }
> #endif
> +       }
>         event_pending = 1;
> }
>
> This is very temporary patch and I need to investigate
> why next_cpu is null but sorry now I dont have enough time.

It is NULL before main_loop starts. I thought the alarm wouldn't occur
before main_loop but apparently it happens. A better solution might be
not scheduling SIG_ALRM if next_cpu is unset, but that would be more
messy. _rearm must execute at least as many times as
host_alarm_handler so a check there would be less optimal. I committed
your patch.
Cheers