Re: [PATCH v2 4/4] KVM: x86: Fix stack-out-of-bounds memory access from ioapic_write_indirect()

[Eduardo Habkost]

On Thu, Aug 26, 2021 at 06:01:15PM +0000, Sean Christopherson wrote:
> On Thu, Aug 26, 2021, Eduardo Habkost wrote:
> > > @@ -918,7 +918,7 @@ static bool kvm_apic_is_broadcast_dest(struct kvm *kvm, struct kvm_lapic **src,
> > >  static inline bool kvm_apic_map_get_dest_lapic(struct kvm *kvm,
> > >                 struct kvm_lapic **src, struct kvm_lapic_irq *irq,
> > >                 struct kvm_apic_map *map, struct kvm_lapic ***dst,
> > > -               unsigned long *bitmap)
> > > +               unsigned long *bitmap64)
> > 
> > You can communicate the expected bitmap size to the compiler
> > without typedefs if using DECLARE_BITMAP inside the function
> > parameter list is acceptable coding style (is it?).
> > 
> > For example, the following would have allowed the compiler to
> > catch the bug you are fixing:
> > 
> > Signed-off-by: Eduardo Habkost <ehabkost@xxxxxxxxxx>
> > ---
> > diff --git a/arch/x86/kvm/lapic.h b/arch/x86/kvm/lapic.h
> > index d7c25d0c1354..e8c64747121a 100644
> > --- a/arch/x86/kvm/lapic.h
> > +++ b/arch/x86/kvm/lapic.h
> > @@ -236,7 +236,7 @@ bool kvm_apic_pending_eoi(struct kvm_vcpu *vcpu, int vector);
> >  void kvm_wait_lapic_expire(struct kvm_vcpu *vcpu);
> >  
> >  void kvm_bitmap_or_dest_vcpus(struct kvm *kvm, struct kvm_lapic_irq *irq,
> > -			      unsigned long *vcpu_bitmap);
> > +			      DECLARE_BITMAP(vcpu_bitmap, KVM_MAX_VCPUS));
> >  
> >  bool kvm_intr_is_single_vcpu_fast(struct kvm *kvm, struct kvm_lapic_irq *irq,
> >  			struct kvm_vcpu **dest_vcpu);
> > diff --git a/arch/x86/kvm/lapic.c b/arch/x86/kvm/lapic.c
> > index 76fb00921203..1df113894cba 100644
> > --- a/arch/x86/kvm/lapic.c
> > +++ b/arch/x86/kvm/lapic.c
> > @@ -1166,7 +1166,7 @@ static int __apic_accept_irq(struct kvm_lapic *apic, int delivery_mode,
> >   * each available vcpu to identify the same.
> >   */
> >  void kvm_bitmap_or_dest_vcpus(struct kvm *kvm, struct kvm_lapic_irq *irq,
> > -			      unsigned long *vcpu_bitmap)
> > +			      DECLARE_BITMAP(vcpu_bitmap, KVM_MAX_VCPUS))
> >  {
> >  	struct kvm_lapic **dest_vcpu = NULL;
> >  	struct kvm_lapic *src = NULL;
> 
> Sadly, that would not have actually caught the bug.  In C++, an array param does
> indeed have a fixed size, but in C an array param is nothing more than syntatic
> sugar that is demoted to a plain ol' pointer.  E.g. gcc-10 and clang-11 both
> happily compile with "DECLARE_BITMAP(vcpu_bitmap, 0)" and the original single
> "unsigned long vcpu_bitmap".  Maybe there are gcc extensions to enforce array
> sizes?  But if there are, they are not (yet) enabled for kernel builds.

The compiler wouldn't have caught it today only because Linux is
compiled with `-Wno-stringop-overflow`.  I have some hope that
eventually the warning will be enabled, as indicated on the
commit message if commit 5a76021c2eff ("gcc-10: disable
'stringop-overflow' warning for now").

Even if the warning isn't enabled, the bitmap size declaration
would be a hint for humans reading the code.

-- 
Eduardo