[PATCH] KVM: x86: fix shift out of bounds reported by UBSAN

[PATCH] KVM: x86: fix shift out of bounds reported by UBSAN

[Paolo Bonzini]

Since we know that e >= s, we can reassociate the left shift,
changing the shifted number from 1 to 2 in exchange for
decreasing the right hand side by 1.

Reported-by: syzbot+e87846c48bf72bc85311@syzkaller.appspotmail.com
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
---
 arch/x86/kvm/mmu.h | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/arch/x86/kvm/mmu.h b/arch/x86/kvm/mmu.h
index 9c4a9c8e43d9..581925e476d6 100644
--- a/arch/x86/kvm/mmu.h
+++ b/arch/x86/kvm/mmu.h
@@ -49,7 +49,7 @@ static inline u64 rsvd_bits(int s, int e)
 	if (e < s)
 		return 0;
 
-	return ((1ULL << (e - s + 1)) - 1) << s;
+	return ((2ULL << (e - s)) - 1) << s;
 }
 
 void kvm_mmu_set_mmio_spte_mask(u64 mmio_value, u64 access_mask);
-- 
2.26.2

Re: [PATCH] KVM: x86: fix shift out of bounds reported by UBSAN

[Sean Christopherson]

On Tue, Dec 22, 2020, Paolo Bonzini wrote:
> Since we know that e >= s, we can reassociate the left shift,
> changing the shifted number from 1 to 2 in exchange for
> decreasing the right hand side by 1.

I assume the edge case is that this ends up as `(1ULL << 64) - 1` and overflows
SHL's max shift count of 63 when s=0 and e=63?  If so, that should be called
out.  If it's something else entirely, then an explanation is definitely in
order.

> Reported-by: syzbot+e87846c48bf72bc85311@syzkaller.appspotmail.com
> Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
> ---
>  arch/x86/kvm/mmu.h | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
> 
> diff --git a/arch/x86/kvm/mmu.h b/arch/x86/kvm/mmu.h
> index 9c4a9c8e43d9..581925e476d6 100644
> --- a/arch/x86/kvm/mmu.h
> +++ b/arch/x86/kvm/mmu.h
> @@ -49,7 +49,7 @@ static inline u64 rsvd_bits(int s, int e)
>  	if (e < s)
>  		return 0;

Maybe add a commment?  Again assuming my guess about the edge case is on point.

	/*
	 * Use 2ULL to incorporate the necessary +1 in the shift; adding +1 in
	 * the shift count will overflow SHL's max shift of 63 if s=0 and e=63.
	 */

> -	return ((1ULL << (e - s + 1)) - 1) << s;
> +	return ((2ULL << (e - s)) - 1) << s;
>  }
>  
>  void kvm_mmu_set_mmio_spte_mask(u64 mmio_value, u64 access_mask);
> -- 
> 2.26.2
> 

RE: [PATCH] KVM: x86: fix shift out of bounds reported by UBSAN

[David Laight]

From: Sean Christopherson
> Sent: 22 December 2020 18:13
> 
> On Tue, Dec 22, 2020, Paolo Bonzini wrote:
> > Since we know that e >= s, we can reassociate the left shift,
> > changing the shifted number from 1 to 2 in exchange for
> > decreasing the right hand side by 1.
> 
> I assume the edge case is that this ends up as `(1ULL << 64) - 1` and overflows
> SHL's max shift count of 63 when s=0 and e=63?  If so, that should be called
> out.  If it's something else entirely, then an explanation is definitely in
> order.
> 
> > Reported-by: syzbot+e87846c48bf72bc85311@syzkaller.appspotmail.com
> > Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
> > ---
> >  arch/x86/kvm/mmu.h | 2 +-
> >  1 file changed, 1 insertion(+), 1 deletion(-)
> >
> > diff --git a/arch/x86/kvm/mmu.h b/arch/x86/kvm/mmu.h
> > index 9c4a9c8e43d9..581925e476d6 100644
> > --- a/arch/x86/kvm/mmu.h
> > +++ b/arch/x86/kvm/mmu.h
> > @@ -49,7 +49,7 @@ static inline u64 rsvd_bits(int s, int e)
> >  	if (e < s)
> >  		return 0;
> 
> Maybe add a commment?  Again assuming my guess about the edge case is on point.
> 
> 	/*
> 	 * Use 2ULL to incorporate the necessary +1 in the shift; adding +1 in
> 	 * the shift count will overflow SHL's max shift of 63 if s=0 and e=63.
> 	 */

A comment of the desired output value would be more use.
I think it is:
	return 'e-s' ones followed by 's' zeros without shifting by 64.

> > -	return ((1ULL << (e - s + 1)) - 1) << s;
> > +	return ((2ULL << (e - s)) - 1) << s;

	David

-
Registered Address Lakeside, Bramley Road, Mount Farm, Milton Keynes, MK1 1PT, UK
Registration No: 1397386 (Wales)

Re: [PATCH] KVM: x86: fix shift out of bounds reported by UBSAN

[Paolo Bonzini]

On 22/12/20 19:31, David Laight wrote:
>> 	/*
>> 	 * Use 2ULL to incorporate the necessary +1 in the shift; adding +1 in
>> 	 * the shift count will overflow SHL's max shift of 63 if s=0 and e=63.
>> 	 */
> A comment of the desired output value would be more use.
> I think it is:
> 	return 'e-s' ones followed by 's' zeros without shifting by 64.
> 

What about a mix of the two:

	/*
	 * Return 'e-s' ones followed by 's' zeros.  Note that the
	 * apparently obvious 1ULL << (e - s + 1) can shift by 64 if
	 * s=0 and e=63, which is undefined behavior.
	 */

Paolo

Re: [PATCH] KVM: x86: fix shift out of bounds reported by UBSAN

[Sean Christopherson]

On Tue, Dec 22, 2020, Paolo Bonzini wrote:
> On 22/12/20 19:31, David Laight wrote:
> > > 	/*
> > > 	 * Use 2ULL to incorporate the necessary +1 in the shift; adding +1 in
> > > 	 * the shift count will overflow SHL's max shift of 63 if s=0 and e=63.
> > > 	 */
> > A comment of the desired output value would be more use.
> > I think it is:
> > 	return 'e-s' ones followed by 's' zeros without shifting by 64.
> > 
> 
> What about a mix of the two:
> 
> 	/*
> 	 * Return 'e-s' ones followed by 's' zeros.  Note that the
> 	 * apparently obvious 1ULL << (e - s + 1) can shift by 64 if
> 	 * s=0 and e=63, which is undefined behavior.
> 	 */

Works for me, thanks!