Paolo Bonzini <[email protected]> writes:
> Commit 7e2175ebd695 ("KVM: x86: Fix recording of guest steal time
> / preempted status", 2021-11-11) open coded the previous call to
> kvm_map_gfn, but in doing so it dropped the comparison between the cached
> guest physical address and the one in the MSR. This cause an incorrect
> cache hit if the guest modifies the steal time address while the memslots
> remain the same. This can happen with kexec, in which case the steal
> time data is written at the address used by the old kernel instead of
> the old one.
>
> While at it, rename the variable from gfn to gpa since it is a plain
> physical address and not a right-shifted one.
>
> Reported-by: Dave Young <[email protected]>
> Reported-by: Xiaoying Yan <[email protected]>
> Analyzed-by: Dr. David Alan Gilbert <[email protected]>
> Cc: David Woodhouse <[email protected]>
> Cc: [email protected]
> Fixes: 7e2175ebd695 ("KVM: x86: Fix recording of guest steal time / preempted status")
> Signed-off-by: Paolo Bonzini <[email protected]>
> ---
> arch/x86/kvm/x86.c | 6 +++---
> 1 file changed, 3 insertions(+), 3 deletions(-)
>
> diff --git a/arch/x86/kvm/x86.c b/arch/x86/kvm/x86.c
> index e5fa335a4ea7..36dcf18b04bf 100644
> --- a/arch/x86/kvm/x86.c
> +++ b/arch/x86/kvm/x86.c
> @@ -3380,6 +3380,7 @@ static void record_steal_time(struct kvm_vcpu *vcpu)
> struct gfn_to_hva_cache *ghc = &vcpu->arch.st.cache;
> struct kvm_steal_time __user *st;
> struct kvm_memslots *slots;
> + gpa_t gpa = vcpu->arch.st.msr_val & KVM_STEAL_VALID_BITS;
> u64 steal;
> u32 version;
>
> @@ -3397,13 +3398,12 @@ static void record_steal_time(struct kvm_vcpu *vcpu)
> slots = kvm_memslots(vcpu->kvm);
>
> if (unlikely(slots->generation != ghc->generation ||
> + gpa != ghc->gpa ||
> kvm_is_error_hva(ghc->hva) || !ghc->memslot)) {
> - gfn_t gfn = vcpu->arch.st.msr_val & KVM_STEAL_VALID_BITS;
> -
> /* We rely on the fact that it fits in a single page. */
> BUILD_BUG_ON((sizeof(*st) - 1) & KVM_STEAL_VALID_BITS);
>
> - if (kvm_gfn_to_hva_cache_init(vcpu->kvm, ghc, gfn, sizeof(*st)) ||
> + if (kvm_gfn_to_hva_cache_init(vcpu->kvm, ghc, gpa, sizeof(*st)) ||
(It would be nice to somehow get at least a warning when 'gfn_t' is used
instead of 'gpa_t' and vice versa)
> kvm_is_error_hva(ghc->hva) || !ghc->memslot)
> return;
> }
Reviewed-by: Vitaly Kuznetsov <[email protected]>
--
Vitaly
* Vitaly Kuznetsov ([email protected]) wrote:
> Paolo Bonzini <[email protected]> writes:
>
> > Commit 7e2175ebd695 ("KVM: x86: Fix recording of guest steal time
> > / preempted status", 2021-11-11) open coded the previous call to
> > kvm_map_gfn, but in doing so it dropped the comparison between the cached
> > guest physical address and the one in the MSR. This cause an incorrect
> > cache hit if the guest modifies the steal time address while the memslots
> > remain the same. This can happen with kexec, in which case the steal
> > time data is written at the address used by the old kernel instead of
> > the old one.
> >
> > While at it, rename the variable from gfn to gpa since it is a plain
> > physical address and not a right-shifted one.
> >
> > Reported-by: Dave Young <[email protected]>
> > Reported-by: Xiaoying Yan <[email protected]>
> > Analyzed-by: Dr. David Alan Gilbert <[email protected]>
> > Cc: David Woodhouse <[email protected]>
> > Cc: [email protected]
> > Fixes: 7e2175ebd695 ("KVM: x86: Fix recording of guest steal time / preempted status")
> > Signed-off-by: Paolo Bonzini <[email protected]>
> > ---
> > arch/x86/kvm/x86.c | 6 +++---
> > 1 file changed, 3 insertions(+), 3 deletions(-)
> >
> > diff --git a/arch/x86/kvm/x86.c b/arch/x86/kvm/x86.c
> > index e5fa335a4ea7..36dcf18b04bf 100644
> > --- a/arch/x86/kvm/x86.c
> > +++ b/arch/x86/kvm/x86.c
> > @@ -3380,6 +3380,7 @@ static void record_steal_time(struct kvm_vcpu *vcpu)
> > struct gfn_to_hva_cache *ghc = &vcpu->arch.st.cache;
> > struct kvm_steal_time __user *st;
> > struct kvm_memslots *slots;
> > + gpa_t gpa = vcpu->arch.st.msr_val & KVM_STEAL_VALID_BITS;
> > u64 steal;
> > u32 version;
> >
> > @@ -3397,13 +3398,12 @@ static void record_steal_time(struct kvm_vcpu *vcpu)
> > slots = kvm_memslots(vcpu->kvm);
> >
> > if (unlikely(slots->generation != ghc->generation ||
> > + gpa != ghc->gpa ||
> > kvm_is_error_hva(ghc->hva) || !ghc->memslot)) {
> > - gfn_t gfn = vcpu->arch.st.msr_val & KVM_STEAL_VALID_BITS;
> > -
> > /* We rely on the fact that it fits in a single page. */
> > BUILD_BUG_ON((sizeof(*st) - 1) & KVM_STEAL_VALID_BITS);
> >
> > - if (kvm_gfn_to_hva_cache_init(vcpu->kvm, ghc, gfn, sizeof(*st)) ||
> > + if (kvm_gfn_to_hva_cache_init(vcpu->kvm, ghc, gpa, sizeof(*st)) ||
>
> (It would be nice to somehow get at least a warning when 'gfn_t' is used
> instead of 'gpa_t' and vice versa)
Can't sparse be taught to do that?
Dave
> > kvm_is_error_hva(ghc->hva) || !ghc->memslot)
> > return;
> > }
>
> Reviewed-by: Vitaly Kuznetsov <[email protected]>
>
> --
> Vitaly
>
--
Dr. David Alan Gilbert / [email protected] / Manchester, UK
On Thu, Aug 04, 2022, Dr. David Alan Gilbert wrote:
> * Vitaly Kuznetsov ([email protected]) wrote:
> > Paolo Bonzini <[email protected]> writes:
> > > - gfn_t gfn = vcpu->arch.st.msr_val & KVM_STEAL_VALID_BITS;
> > > -
> > > /* We rely on the fact that it fits in a single page. */
> > > BUILD_BUG_ON((sizeof(*st) - 1) & KVM_STEAL_VALID_BITS);
> > >
> > > - if (kvm_gfn_to_hva_cache_init(vcpu->kvm, ghc, gfn, sizeof(*st)) ||
> > > + if (kvm_gfn_to_hva_cache_init(vcpu->kvm, ghc, gpa, sizeof(*st)) ||
> >
> > (It would be nice to somehow get at least a warning when 'gfn_t' is used
> > instead of 'gpa_t' and vice versa)
>
> Can't sparse be taught to do that?
Hmm, it probably could, but the result would likely be a mess. E.g. anything that
shifts the GPA on-demand will require explicit casts to make sparse happy.
This particular case is solvable without sparse, e.g. WARN if gpa[11:0]!=0, or
even better rework the function to actually take a @gfn and then WARN if the
incoming gfn would yield an illegal gpa.