[Qemu-devel] Re: [RFC v2] queue_work proposal

[Glauber Costa]

On Thu, Sep 03, 2009 at 08:48:53PM +0200, Juan Quintela wrote:
> Glauber Costa <address@hidden> wrote:
> 
> Hi Glauber
> 
> > +int kvm_vcpu_ioctl(CPUState *env, int type, ...)
> > +{
> >      void *arg;
> >      va_list ap;
> > +    KVMIoctl data;
> >  
> >      va_start(ap, type);
> >      arg = va_arg(ap, void *);
> >      va_end(ap);
> >  
> > -    ret = ioctl(env->kvm_fd, type, arg);
> > -    if (ret == -1)
> > -        ret = -errno;
> > +    data.type = type;
> > +    data.data = arg;
> > +    data.fd = env->kvm_fd;
> >  
> > -    return ret;
> > +    qemu_queue_work(env, kvm_remote_ioctl, (void *)&data);
> > +
> > +    return data.ret;
> 
> Does this work?  giving the address of a local variable is a no-no.
Yes, because it synchronous. The stack will keep existing until queue_work 
returns.

Actually, this was one of the reasons Avi pushed so hardly against the async 
mechanism