Re: [Qemu-devel] [PATCH] kvm: warn if num cpus is greater than num recommended

[Andrew Jones]

----- Original Message -----
> Il 22/08/2013 17:39, Andrew Jones ha scritto:
> > The comment in kvm_max_vcpus() states that it's using the recommended
> > procedure from the kernel API documentation to get the max number
> > of vcpus that kvm supports. It is, but by always returning the
> > maximum number supported. The maximum number should only be used
> > for development purposes. qemu should check KVM_CAP_NR_VCPUS for
> > the recommended number of vcpus. This patch adds a warning if a user
> > specifies a number of cpus between the recommended and max.
> > 
> > Signed-off-by: Andrew Jones <address@hidden>
> > ---
> >  kvm-all.c | 45 +++++++++++++++++++++++++++------------------
> >  1 file changed, 27 insertions(+), 18 deletions(-)
> > 
> > diff --git a/kvm-all.c b/kvm-all.c
> > index 716860f617455..9092e13ae60ea 100644
> > --- a/kvm-all.c
> > +++ b/kvm-all.c
> > @@ -1313,24 +1313,24 @@ static int kvm_irqchip_create(KVMState *s)
> >      return 0;
> >  }
> >  
> > -static int kvm_max_vcpus(KVMState *s)
> > +/* Find number of supported CPUs using the recommended
> > + * procedure from the kernel API documentation to cope with
> > + * older kernels that may be missing capabilities.
> > + */
> > +static int kvm_recommended_vcpus(KVMState *s)
> >  {
> >      int ret;
> >  
> > -    /* Find number of supported CPUs using the recommended
> > -     * procedure from the kernel API documentation to cope with
> > -     * older kernels that may be missing capabilities.
> > -     */
> > -    ret = kvm_check_extension(s, KVM_CAP_MAX_VCPUS);
> > -    if (ret) {
> > -        return ret;
> > -    }
> >      ret = kvm_check_extension(s, KVM_CAP_NR_VCPUS);
> > -    if (ret) {
> > -        return ret;
> > -    }
> > +    return (ret) ? ret : 4;
> > +}
> >  
> > -    return 4;
> > +static int kvm_max_vcpus(KVMState *s)
> > +{
> > +    int ret;
> > +
> > +    ret = kvm_check_extension(s, KVM_CAP_MAX_VCPUS);
> > +    return (ret) ? ret : kvm_recommended_vcpus(s);
> >  }
> >  
> >  int kvm_init(void)
> > @@ -1383,12 +1383,21 @@ int kvm_init(void)
> >          goto err;
> >      }
> >  
> > -    max_vcpus = kvm_max_vcpus(s);
> > +    max_vcpus = kvm_recommended_vcpus(s);
> >      if (smp_cpus > max_vcpus) {
> > -        ret = -EINVAL;
> > -        fprintf(stderr, "Number of SMP cpus requested (%d) exceeds max
> > cpus "
> > -                "supported by KVM (%d)\n", smp_cpus, max_vcpus);
> > -        goto err;
> > +        fprintf(stderr,
> > +                "Warning: Number of SMP cpus requested (%d) exceeds "
> > +                "recommended cpus supported by KVM (%d)\n",
> > +                smp_cpus, max_vcpus);
> > +
> > +        max_vcpus = kvm_max_vcpus(s);
> > +        if (smp_cpus > max_vcpus) {
> > +            ret = -EINVAL;
> > +            fprintf(stderr, "Number of SMP cpus requested (%d) exceeds "
> > +                    "max cpus supported by KVM (%d)\n",
> > +                    smp_cpus, max_vcpus);
> > +            goto err;
> > +        }
> 
> You print both error messages when smp_cpus is greater than the max cpus
> supported; is it intentional?
> 

Yup. This way we can inform the user not only that they're greater than
hard-max, but also what the soft-max is. This allows the user to choose
how much lower to adjust when they correct for the hard-max, possibly going
down low enough to avoid blowing the soft-max as well.

> Apart from this, the concept looks good.  However, please over
> qemu-kvm.git's uq/master branch, where we already have Marcelo's patch
> to check max_cpus too against kvm_max_vcpus(s).
> 

OK, will respin on uq/master.

thanks,
drew