Hi all,

Today's linux-next merge of the kvm-x86 tree got a conflict in:

  arch/x86/kvm/svm/svm.c

between commit:

  6b1ca262a943a ("KVM: x86: clarify leave_smm() return value")

from the kvm-fixes tree and commit:

  84dc9fd0354d3 ("KVM: nSVM: Cache all used fields from VMCB12")

from the kvm-x86 tree.

I fixed it up (see below) and can carry the fix as necessary. This
is now fixed as far as linux-next is concerned, but any non trivial
conflicts should be mentioned to your upstream maintainer when your tree
is submitted for merging.  You may also want to consider cooperating
with the maintainer of the conflicting tree to minimise any particularly
complex conflicts.

I suspect this may be flagging an issue with the merge I just posted and
removal of the check for the cache...

diff --cc arch/x86/kvm/svm/svm.c
index e6477affac9a0,3407deac90bd6..0000000000000
--- a/arch/x86/kvm/svm/svm.c
+++ b/arch/x86/kvm/svm/svm.c
@@@ -4880,15 -5030,11 +5030,11 @@@ static int svm_leave_smm(struct kvm_vcp
  	vmcb12 = map.hva;
  	nested_copy_vmcb_control_to_cache(svm, &vmcb12->control);
  	nested_copy_vmcb_save_to_cache(svm, &vmcb12->save);
 -	ret = enter_svm_guest_mode(vcpu, smram64->svm_guest_vmcb_gpa, false);
  
- 	if (nested_svm_check_cached_vmcb12(vcpu) < 0)
- 		goto unmap_save;
- 
- 	if (enter_svm_guest_mode(vcpu, smram64->svm_guest_vmcb_gpa,
- 				 vmcb12, false) != 0)
 -	if (ret)
++	if (enter_svm_guest_mode(vcpu, smram64->svm_guest_vmcb_gpa, false) != 0)
  		goto unmap_save;
  
 +	ret = 0;
  	svm->nested.nested_run_pending = 1;
  
  unmap_save:

On 3/12/26 18:21, Mark Brown wrote:
> diff --cc arch/x86/kvm/svm/svm.c
> index e6477affac9a0,3407deac90bd6..0000000000000
> --- a/arch/x86/kvm/svm/svm.c
> +++ b/arch/x86/kvm/svm/svm.c
> @@@ -4880,15 -5030,11 +5030,11 @@@ static int svm_leave_smm(struct kvm_vcp
>    	vmcb12 = map.hva;
>    	nested_copy_vmcb_control_to_cache(svm, &vmcb12->control);
>    	nested_copy_vmcb_save_to_cache(svm, &vmcb12->save);
>   -	ret = enter_svm_guest_mode(vcpu, smram64->svm_guest_vmcb_gpa, false);
>    
> - 	if (nested_svm_check_cached_vmcb12(vcpu) < 0)
> - 		goto unmap_save;

The right resolution is to keep these two lines...

> -
> - 	if (enter_svm_guest_mode(vcpu, smram64->svm_guest_vmcb_gpa,
> - 				 vmcb12, false) != 0)
>   -	if (ret)
> ++	if (enter_svm_guest_mode(vcpu, smram64->svm_guest_vmcb_gpa, false) != 0)

... while of course this part is okay.

Paolo

>    		goto unmap_save;
>    
>   +	ret = 0;
>    	svm->nested.nested_run_pending = 1;
>    
>    unmap_save:

On Thu, Mar 12, 2026, Paolo Bonzini wrote:
> On 3/12/26 18:21, Mark Brown wrote:
> > diff --cc arch/x86/kvm/svm/svm.c
> > index e6477affac9a0,3407deac90bd6..0000000000000
> > --- a/arch/x86/kvm/svm/svm.c
> > +++ b/arch/x86/kvm/svm/svm.c
> > @@@ -4880,15 -5030,11 +5030,11 @@@ static int svm_leave_smm(struct kvm_vcp
> >    	vmcb12 = map.hva;
> >    	nested_copy_vmcb_control_to_cache(svm, &vmcb12->control);
> >    	nested_copy_vmcb_save_to_cache(svm, &vmcb12->save);
> >   -	ret = enter_svm_guest_mode(vcpu, smram64->svm_guest_vmcb_gpa, false);
> > - 	if (nested_svm_check_cached_vmcb12(vcpu) < 0)
> > - 		goto unmap_save;
> 
> The right resolution is to keep these two lines...
> 
> > -
> > - 	if (enter_svm_guest_mode(vcpu, smram64->svm_guest_vmcb_gpa,
> > - 				 vmcb12, false) != 0)
> >   -	if (ret)
> > ++	if (enter_svm_guest_mode(vcpu, smram64->svm_guest_vmcb_gpa, false) != 0)
> 
> ... while of course this part is okay.

I'm in the process of redoing kvm-x86/next on top of kvm/next, so I'll sort this
out on my end (and double check that I come up with the same resolution as Paolo).

On Thu, Mar 12, 2026 at 06:54:18PM +0100, Paolo Bonzini wrote:
> On 3/12/26 18:21, Mark Brown wrote:

> >   -	ret = enter_svm_guest_mode(vcpu, smram64->svm_guest_vmcb_gpa, false);
> > - 	if (nested_svm_check_cached_vmcb12(vcpu) < 0)
> > - 		goto unmap_save;

> The right resolution is to keep these two lines...

Yeah, that was the bit about the prior resolution being wrong.