[Qemu-devel] Re: [PATCH v3] introduce on_vcpu

[Glauber Costa <glommer@redhat.com>]

On Thu, Aug 27, 2009 at 07:15:28PM +0200, Jan Kiszka wrote:
> Glauber Costa wrote:
> > on_vcpu is a qemu-kvm function that will make sure that a specific
> > piece of code will run on a requested cpu. We don't need that because
> > we're restricted to -smp 1 right now, but those days are likely to end soon.
> > 
> > So for the benefit of having qemu-kvm share more code with us, I'm
> > introducing our own version of on_vcpu(). Right now, we either run
> > a function on the current cpu, or abort the execution, because it would
> > mean something is seriously wrong.
> > 
> > As an example code, I "ported" kvm_update_guest_debug to use it,
> > with some slight differences from qemu-kvm.
> > 
> > This is probably 0.12 material
> > 
> > Signed-off-by: Glauber Costa <glommer@redhat.com>
> > CC: Jan Kiszka <jan.kiszka@siemens.com>
> > ---
> >  kvm-all.c |   35 +++++++++++++++++++++++++++++------
> >  1 files changed, 29 insertions(+), 6 deletions(-)
> > 
> > diff --git a/kvm-all.c b/kvm-all.c
> > index 61194b8..07a1cdb 100644
> > --- a/kvm-all.c
> > +++ b/kvm-all.c
> > @@ -155,6 +155,15 @@ static void kvm_reset_vcpu(void *opaque)
> >      }
> >  }
> >  
> > +static void on_vcpu(CPUState *env, void (*func)(void *data), void *data)
> > +{
> > +    if (env == cpu_single_env) {
> > +        func(data);
> > +        return;
> > +    }
> > +    abort();
> 
> Sorry, missed this before it went in: This abort fires already now when
> kvm_update_guest_debug is invoked by the gdbstub (where cpu_single_env
> is 0). This completely breaks guest debugging int kvm mode.
To begin with, I believe checking for cpu_single_env is totally evil.
It is unpredictable at best, since there are many places (like you just found out)
that will clear it.

qemu-kvm uses a TLS variable for that, to guarantee that we're in the same thread
as our calling context. I like this idea. if we have io-thread disabled, we're always
in the same thread, and will always execute the function directly as we used to do
before on_vcpu().

I do however remember anthony bending towards issuing a gettid() instead of using
a TLS var. I'm fine with both. Anthony, avi, you guys have a word here?