Re: [PATCH v3 3/4] x86/kvm: add max number of vcpus for hyperv emulation

[Juergen Gross]

On 17.11.21 21:50, Sean Christopherson wrote:
> On Tue, Nov 16, 2021, Juergen Gross wrote:
> > When emulating Hyperv the theoretical maximum of vcpus supported is
> > 4096, as this is the architectural limit for sending IPIs via the PV
> > interface.
> >
> > For restricting the actual supported number of vcpus for that case
> > introduce another define KVM_MAX_HYPERV_VCPUS and set it to 1024, like
> > today's KVM_MAX_VCPUS. Make both values unsigned ones as this will be
> > needed later.
> >
> > The actual number of supported vcpus for Hyperv emulation will be the
> > lower value of both defines.
> >
> > This is a preparation for a future boot parameter support of the max
> > number of vcpus for a KVM guest.
> >
> > Signed-off-by: Juergen Gross <jgross@xxxxxxxx>
> > ---
> > V3:
> > - new patch
> > ---
> >   arch/x86/include/asm/kvm_host.h |  3 ++-
> >   arch/x86/kvm/hyperv.c           | 15 ++++++++-------
> >   2 files changed, 10 insertions(+), 8 deletions(-)
> >
> > diff --git a/arch/x86/include/asm/kvm_host.h b/arch/x86/include/asm/kvm_host.h
> > index 886930ec8264..8ea03ff01c45 100644
> > --- a/arch/x86/include/asm/kvm_host.h
> > +++ b/arch/x86/include/asm/kvm_host.h
> > @@ -38,7 +38,8 @@
> >   
> >   #define __KVM_HAVE_ARCH_VCPU_DEBUGFS
> >   
> > -#define KVM_MAX_VCPUS 1024
> > +#define KVM_MAX_VCPUS 1024U
> > +#define KVM_MAX_HYPERV_VCPUS 1024U
>
> I don't see any reason to put this in kvm_host.h, it should never be used outside
> of hyperv.c.

Okay, fine with me.

> >   #define KVM_MAX_VCPU_IDS kvm_max_vcpu_ids()
> >   /* memory slots that are not exposed to userspace */
> >   #define KVM_PRIVATE_MEM_SLOTS 3
> > diff --git a/arch/x86/kvm/hyperv.c b/arch/x86/kvm/hyperv.c
> > index 4a555f32885a..c0fa837121f1 100644
> > --- a/arch/x86/kvm/hyperv.c
> > +++ b/arch/x86/kvm/hyperv.c
> > @@ -41,7 +41,7 @@
> >   /* "Hv#1" signature */
> >   #define HYPERV_CPUID_SIGNATURE_EAX 0x31237648
> >   
> > -#define KVM_HV_MAX_SPARSE_VCPU_SET_BITS DIV_ROUND_UP(KVM_MAX_VCPUS, 64)
> > +#define KVM_HV_MAX_SPARSE_VCPU_SET_BITS DIV_ROUND_UP(KVM_MAX_HYPERV_VCPUS, 64)
> >   
> >   static void stimer_mark_pending(struct kvm_vcpu_hv_stimer *stimer,
> >   				bool vcpu_kick);
> > @@ -166,7 +166,7 @@ static struct kvm_vcpu *get_vcpu_by_vpidx(struct kvm *kvm, u32 vpidx)
> >   	struct kvm_vcpu *vcpu = NULL;
> >   	int i;
> >   
> > -	if (vpidx >= KVM_MAX_VCPUS)
> > +	if (vpidx >= min(KVM_MAX_VCPUS, KVM_MAX_HYPERV_VCPUS))
>
> IMO, this is conceptually wrong.  KVM should refuse to allow Hyper-V to be enabled
> if the max number of vCPUs exceeds what can be supported, or should refuse to create

TBH, I wasn't sure where to put this test. Is there a guaranteed
sequence of ioctl()s regarding vcpu creation (or setting the max
number of vcpus) and the Hyper-V enabling?

> the vCPUs.  I agree it makes sense to add a Hyper-V specific limit, since there are
> Hyper-V structures that have a hard limit, but detection of violations should be a
> BUILD_BUG_ON, not a silent failure at runtime.

A BUILD_BUG_ON won't be possible with KVM_MAX_VCPUS being selecteble via
boot parameter.


Juergen

Attachment: OpenPGP_0xB0DE9DD628BF132F.asc
Description: OpenPGP public key

Attachment: OpenPGP_signature
Description: OpenPGP digital signature