On Fri, Jun 13, 2025 at 1:11 PM avinashlalotra <abinashlalotra@xxxxxxxxx> wrote: > diff --git a/arch/x86/kvm/hyperv.c b/arch/x86/kvm/hyperv.c > index 24f0318c50d7..78bb8d58fe94 100644 > --- a/arch/x86/kvm/hyperv.c > +++ b/arch/x86/kvm/hyperv.c > @@ -2005,7 +2005,7 @@ static u64 kvm_hv_flush_tlb(struct kvm_vcpu *vcpu, struct kvm_hv_hcall *hc) > struct kvm *kvm = vcpu->kvm; > struct hv_tlb_flush_ex flush_ex; > struct hv_tlb_flush flush; > - DECLARE_BITMAP(vcpu_mask, KVM_MAX_VCPUS); > + unsigned long *vcpu_mask; The default KVM_MAX_VCPUS is 1024, which is not too bad; you're probably compiling with CONFIG_MAXSMP and accepting the default limit of 4096. Adding an allocation for every hypercall is not great, I'd rather add it to struct kvm_vcpu_arch* instead. If instead we go for having the allocation, you can use this: unsigned long *vcpu_mask __free(bitmap) = NULL; and avoid changing the returns to gotos everywhere else. Paolo
