Right now, svm_hv_vmcb_dirty_nested_enlightenments has an incorrect dereference of vmcb->control.reserved_sw before the vmcb is checked for being non-NULL. The compiler is usually sinking the dereference after the check; instead of doing this ourselves in the source, ensure that svm_hv_vmcb_dirty_nested_enlightenments is only called with a non-NULL VMCB. Reported-by: Dan Carpenter <dan.carpenter@oracle.com> Cc: Vineeth Pillai <viremana@linux.microsoft.com> Signed-off-by: Paolo Bonzini <pbonzini@redhat.com> [Untested for now due to issues with my AMD machine. - Paolo] --- arch/x86/kvm/svm/svm.c | 4 ++-- arch/x86/kvm/svm/svm_onhyperv.h | 2 +- 2 files changed, 3 insertions(+), 3 deletions(-) diff --git a/arch/x86/kvm/svm/svm.c b/arch/x86/kvm/svm/svm.c index 9a6987549e1b..4bcb95bb8ed7 100644 --- a/arch/x86/kvm/svm/svm.c +++ b/arch/x86/kvm/svm/svm.c @@ -1406,8 +1406,6 @@ static int svm_create_vcpu(struct kvm_vcpu *vcpu) goto error_free_vmsa_page; } - svm_vcpu_init_msrpm(vcpu, svm->msrpm); - svm->vmcb01.ptr = page_address(vmcb01_page); svm->vmcb01.pa = __sme_set(page_to_pfn(vmcb01_page) << PAGE_SHIFT); @@ -1419,6 +1417,8 @@ static int svm_create_vcpu(struct kvm_vcpu *vcpu) svm_switch_vmcb(svm, &svm->vmcb01); init_vmcb(vcpu); + svm_vcpu_init_msrpm(vcpu, svm->msrpm); + svm_init_osvw(vcpu); vcpu->arch.microcode_version = 0x01000065; diff --git a/arch/x86/kvm/svm/svm_onhyperv.h b/arch/x86/kvm/svm/svm_onhyperv.h index 9b9a55abc29f..c53b8bf8d013 100644 --- a/arch/x86/kvm/svm/svm_onhyperv.h +++ b/arch/x86/kvm/svm/svm_onhyperv.h @@ -89,7 +89,7 @@ static inline void svm_hv_vmcb_dirty_nested_enlightenments( * as we mark it dirty unconditionally towards end of vcpu * init phase. */ - if (vmcb && vmcb_is_clean(vmcb, VMCB_HV_NESTED_ENLIGHTENMENTS) && + if (vmcb_is_clean(vmcb, VMCB_HV_NESTED_ENLIGHTENMENTS) && hve->hv_enlightenments_control.msr_bitmap) vmcb_mark_dirty(vmcb, VMCB_HV_NESTED_ENLIGHTENMENTS); } -- 2.27.0
Right now, svm_hv_vmcb_dirty_nested_enlightenments has an incorrect dereference of vmcb->control.reserved_sw before the vmcb is checked for being non-NULL. The compiler is usually sinking the dereference after the check; instead of doing this ourselves in the source, ensure that svm_hv_vmcb_dirty_nested_enlightenments is only called with a non-NULL VMCB. Reported-by: Dan Carpenter <dan.carpenter@oracle.com> Cc: Vineeth Pillai <viremana@linux.microsoft.com> Signed-off-by: Paolo Bonzini <pbonzini@redhat.com> [Untested for now due to issues with my AMD machine. - Paolo] --- arch/x86/kvm/svm/svm.c | 4 ++-- arch/x86/kvm/svm/svm_onhyperv.h | 2 +- 2 files changed, 3 insertions(+), 3 deletions(-) diff --git a/arch/x86/kvm/svm/svm.c b/arch/x86/kvm/svm/svm.c index 9a6987549e1b..4bcb95bb8ed7 100644 --- a/arch/x86/kvm/svm/svm.c +++ b/arch/x86/kvm/svm/svm.c @@ -1406,8 +1406,6 @@ static int svm_create_vcpu(struct kvm_vcpu *vcpu) goto error_free_vmsa_page; } - svm_vcpu_init_msrpm(vcpu, svm->msrpm); - svm->vmcb01.ptr = page_address(vmcb01_page); svm->vmcb01.pa = __sme_set(page_to_pfn(vmcb01_page) << PAGE_SHIFT); @@ -1419,6 +1417,8 @@ static int svm_create_vcpu(struct kvm_vcpu *vcpu) svm_switch_vmcb(svm, &svm->vmcb01); init_vmcb(vcpu); + svm_vcpu_init_msrpm(vcpu, svm->msrpm); + svm_init_osvw(vcpu); vcpu->arch.microcode_version = 0x01000065; diff --git a/arch/x86/kvm/svm/svm_onhyperv.h b/arch/x86/kvm/svm/svm_onhyperv.h index 9b9a55abc29f..c53b8bf8d013 100644 --- a/arch/x86/kvm/svm/svm_onhyperv.h +++ b/arch/x86/kvm/svm/svm_onhyperv.h @@ -89,7 +89,7 @@ static inline void svm_hv_vmcb_dirty_nested_enlightenments( * as we mark it dirty unconditionally towards end of vcpu * init phase. */ - if (vmcb && vmcb_is_clean(vmcb, VMCB_HV_NESTED_ENLIGHTENMENTS) && + if (vmcb_is_clean(vmcb, VMCB_HV_NESTED_ENLIGHTENMENTS) && hve->hv_enlightenments_control.msr_bitmap) vmcb_mark_dirty(vmcb, VMCB_HV_NESTED_ENLIGHTENMENTS); } -- 2.27.0
Hi Paulo,
Thanks a lot for fixing this.
>
> diff --git a/arch/x86/kvm/svm/svm_onhyperv.h b/arch/x86/kvm/svm/svm_onhyperv.h
> index 9b9a55abc29f..c53b8bf8d013 100644
> --- a/arch/x86/kvm/svm/svm_onhyperv.h
> +++ b/arch/x86/kvm/svm/svm_onhyperv.h
> @@ -89,7 +89,7 @@ static inline void svm_hv_vmcb_dirty_nested_enlightenments(
> * as we mark it dirty unconditionally towards end of vcpu
> * init phase.
> */
> - if (vmcb && vmcb_is_clean(vmcb, VMCB_HV_NESTED_ENLIGHTENMENTS) &&
> + if (vmcb_is_clean(vmcb, VMCB_HV_NESTED_ENLIGHTENMENTS) &&
> hve->hv_enlightenments_control.msr_bitmap)
> vmcb_mark_dirty(vmcb, VMCB_HV_NESTED_ENLIGHTENMENTS);
> }
The changes looks good to me. Could you please remove the above comment
as well while you are at it.
Many Thanks,
Vineeth
Paolo Bonzini <pbonzini@redhat.com> writes: > Right now, svm_hv_vmcb_dirty_nested_enlightenments has an incorrect > dereference of vmcb->control.reserved_sw before the vmcb is checked > for being non-NULL. The compiler is usually sinking the dereference > after the check; instead of doing this ourselves in the source, > ensure that svm_hv_vmcb_dirty_nested_enlightenments is only called > with a non-NULL VMCB. > > Reported-by: Dan Carpenter <dan.carpenter@oracle.com> > Cc: Vineeth Pillai <viremana@linux.microsoft.com> > Signed-off-by: Paolo Bonzini <pbonzini@redhat.com> > [Untested for now due to issues with my AMD machine. - Paolo] At least this doesn't seem to break kvm-amd on bare metal, so Tested-by: Vitaly Kuznetsov <vkuznets@redhat.com> > --- > arch/x86/kvm/svm/svm.c | 4 ++-- > arch/x86/kvm/svm/svm_onhyperv.h | 2 +- > 2 files changed, 3 insertions(+), 3 deletions(-) > > diff --git a/arch/x86/kvm/svm/svm.c b/arch/x86/kvm/svm/svm.c > index 9a6987549e1b..4bcb95bb8ed7 100644 > --- a/arch/x86/kvm/svm/svm.c > +++ b/arch/x86/kvm/svm/svm.c > @@ -1406,8 +1406,6 @@ static int svm_create_vcpu(struct kvm_vcpu *vcpu) > goto error_free_vmsa_page; > } > > - svm_vcpu_init_msrpm(vcpu, svm->msrpm); > - > svm->vmcb01.ptr = page_address(vmcb01_page); > svm->vmcb01.pa = __sme_set(page_to_pfn(vmcb01_page) << PAGE_SHIFT); > > @@ -1419,6 +1417,8 @@ static int svm_create_vcpu(struct kvm_vcpu *vcpu) > svm_switch_vmcb(svm, &svm->vmcb01); > init_vmcb(vcpu); > > + svm_vcpu_init_msrpm(vcpu, svm->msrpm); > + > svm_init_osvw(vcpu); > vcpu->arch.microcode_version = 0x01000065; > > diff --git a/arch/x86/kvm/svm/svm_onhyperv.h b/arch/x86/kvm/svm/svm_onhyperv.h > index 9b9a55abc29f..c53b8bf8d013 100644 > --- a/arch/x86/kvm/svm/svm_onhyperv.h > +++ b/arch/x86/kvm/svm/svm_onhyperv.h > @@ -89,7 +89,7 @@ static inline void svm_hv_vmcb_dirty_nested_enlightenments( > * as we mark it dirty unconditionally towards end of vcpu > * init phase. > */ > - if (vmcb && vmcb_is_clean(vmcb, VMCB_HV_NESTED_ENLIGHTENMENTS) && > + if (vmcb_is_clean(vmcb, VMCB_HV_NESTED_ENLIGHTENMENTS) && > hve->hv_enlightenments_control.msr_bitmap) > vmcb_mark_dirty(vmcb, VMCB_HV_NESTED_ENLIGHTENMENTS); > } -- Vitaly
On 7/27/2021 11:23 AM, Vitaly Kuznetsov wrote:
> Paolo Bonzini <pbonzini@redhat.com> writes:
>
>> Right now, svm_hv_vmcb_dirty_nested_enlightenments has an incorrect
>> dereference of vmcb->control.reserved_sw before the vmcb is checked
>> for being non-NULL. The compiler is usually sinking the dereference
>> after the check; instead of doing this ourselves in the source,
>> ensure that svm_hv_vmcb_dirty_nested_enlightenments is only called
>> with a non-NULL VMCB.
>>
>> Reported-by: Dan Carpenter <dan.carpenter@oracle.com>
>> Cc: Vineeth Pillai <viremana@linux.microsoft.com>
>> Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
>> [Untested for now due to issues with my AMD machine. - Paolo]
Finally got hold of an AMD machine and tested nested virt: windows on
linux on
windows with the patches applied. Did basic boot and minimal verification.
Tested-by: Vineeth Pillai <viremana@linux.microsoft.com>
Thanks,
Vineeth
On 28/07/21 22:18, Vineeth Pillai wrote:
>
> On 7/27/2021 11:23 AM, Vitaly Kuznetsov wrote:
>> Paolo Bonzini <pbonzini@redhat.com> writes:
>>
>>> Right now, svm_hv_vmcb_dirty_nested_enlightenments has an incorrect
>>> dereference of vmcb->control.reserved_sw before the vmcb is checked
>>> for being non-NULL. The compiler is usually sinking the dereference
>>> after the check; instead of doing this ourselves in the source,
>>> ensure that svm_hv_vmcb_dirty_nested_enlightenments is only called
>>> with a non-NULL VMCB.
>>>
>>> Reported-by: Dan Carpenter <dan.carpenter@oracle.com>
>>> Cc: Vineeth Pillai <viremana@linux.microsoft.com>
>>> Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
>>> [Untested for now due to issues with my AMD machine. - Paolo]
> Finally got hold of an AMD machine and tested nested virt: windows on
> linux on
> windows with the patches applied. Did basic boot and minimal verification.
>
> Tested-by: Vineeth Pillai <viremana@linux.microsoft.com>
Thanks! In the meanwhile I had fixed my machine too. :)
Paolo