Re: [PATCH] x86: kvm: introduce CONFIG_KVM_MAX_VCPUS

[Gleb Natapov <gleb@redhat.com>]

On Mon, Sep 16, 2013 at 02:03:33PM +0200, Andrew Jones wrote:
> On Mon, Sep 16, 2013 at 11:47:10AM +0300, Gleb Natapov wrote:
> > On Mon, Sep 16, 2013 at 10:28:20AM +0200, Andrew Jones wrote:
> > > On Sun, Sep 15, 2013 at 12:08:38PM +0300, Gleb Natapov wrote:
> > > > On Sat, Sep 14, 2013 at 02:18:49PM +0200, Andrew Jones wrote:
> > > > > Take CONFIG_KVM_MAX_VCPUS from arm32, but set the default to 255.
> > > > > 
> > > > > Signed-off-by: Andrew Jones <drjones@redhat.com>
> > > > > ---
> > > > >  arch/x86/include/asm/kvm_host.h |  5 +++--
> > > > >  arch/x86/kvm/Kconfig            | 10 ++++++++++
> > > > >  2 files changed, 13 insertions(+), 2 deletions(-)
> > > > > 
> > > > > diff --git a/arch/x86/include/asm/kvm_host.h b/arch/x86/include/asm/kvm_host.h
> > > > > index c76ff74a98f2e..e7e9b523a8f7e 100644
> > > > > --- a/arch/x86/include/asm/kvm_host.h
> > > > > +++ b/arch/x86/include/asm/kvm_host.h
> > > > > @@ -31,8 +31,9 @@
> > > > >  #include <asm/msr-index.h>
> > > > >  #include <asm/asm.h>
> > > > >  
> > > > > -#define KVM_MAX_VCPUS 255
> > > > > -#define KVM_SOFT_MAX_VCPUS 160
> > > > > +#define KVM_MAX_VCPUS CONFIG_KVM_MAX_VCPUS
> > > > > +#define KVM_SOFT_MAX_VCPUS min(160, KVM_MAX_VCPUS)
> > > > > +
> > > > >  #define KVM_USER_MEM_SLOTS 125
> > > > >  /* memory slots that are not exposed to userspace */
> > > > >  #define KVM_PRIVATE_MEM_SLOTS 3
> > > > > diff --git a/arch/x86/kvm/Kconfig b/arch/x86/kvm/Kconfig
> > > > > index a47a3e54b964b..e9532c33527ee 100644
> > > > > --- a/arch/x86/kvm/Kconfig
> > > > > +++ b/arch/x86/kvm/Kconfig
> > > > > @@ -52,6 +52,16 @@ config KVM
> > > > >  
> > > > >  	  If unsure, say N.
> > > > >  
> > > > > +config KVM_MAX_VCPUS
> > > > > +	int "Number maximum supported virtual CPUs per VM"
> > > > > +	depends on KVM
> > > > > +	default 255
> > > > > +	help
> > > > > +	  Static number of max supported virtual CPUs per VM.
> > > > > +
> > > > > +	  Set to a lower number to save some resources. Set to a higher
> > > > > +	  number to test scalability.
> > > > > +
> > > > Maximum this can save is around 2K per VM. This is pretty insignificant
> > > > considering overall memory footprint even smallest VM has.
> > > 
> > > Should I reword this, dropping all 'save resources' verbiage, in order to
> > > avoid sending a message that this option can affect resource consumption?
> > > Or just leave it as it is, because even though it's insignificant, it's
> > > still true and balances out the 'Set to a higher' part.
> > > 
> > I do not think config option is necessary. The overhead is so
> > insignificant that there is no point in additional user visible knob,
> > at least while only 255 vcpu are supported. Is there a reason for anyone
> > to configure less them 255 vcpus here? OTOH what prevents someone from
> > configuring more then 255 vcpus?
> 
> The reason to configure less is to be able to compile a hard limit,
> without having to muck with the source. E.g. if you really don't want to
> allow more than KVM_SOFT_MAX_VCPUS to be configured, then you can compile
> with CONFIG_KVM_MAX_VCPUS == KVM_SOFT_MAX_VCPUS.
> 
> Nothing prevents someone from configuring more than the max in
> userspace, but if they try to create/use more than the max
> (kvm_vm_ioctl_create_vcpu), it'll fail (EINVAL).
> 
I was talking about configuring more then 255 vcus in kernel config. You
should have "range 1 255" there.

> I see this as a step towards getting rid of KVM_SOFT_MAX_VCPUS. I.e.
> compile with whatever maximum limit you want (can support), and return
> online/possible cpus for the recommended number. Only configure the kernel
> with more than your typical maximum (was KVM_SOFT_MAX_VCPUS) for
> development/testing purposes.
> 
The idea behind soft/hard limit was to allow people easily check
scalability without need to recompile the kernel. If custom build will
be needed most people will not bother.

--
			Gleb.