Re: [PATCH v2] vfio: Use file-based reference counting for KVM

[Steffen Eiden]

On Thu, Sep 03, 2026 at 10:20:39AM -0700, Sean Christopherson wrote:
> On Thu, Sep 03, 2026, Steffen Eiden wrote:
> > diff --git a/virt/kvm/kvm_main.c b/virt/kvm/kvm_main.c
> > index 65eb26a0520d..34b4c43908b3 100644
> > --- a/virt/kvm/kvm_main.c
> > +++ b/virt/kvm/kvm_main.c
> > @@ -1314,7 +1314,7 @@ void kvm_get_kvm(struct kvm *kvm)
> >  {
> >  	refcount_inc(&kvm->users_count);
> >  }
> > -EXPORT_SYMBOL_GPL(kvm_get_kvm);
> > +EXPORT_SYMBOL_FOR_KVM_INTERNAL(kvm_get_kvm);
> >  
> >  /*
> >   * Make sure the vm is not during destruction, which is a safe version of
> > @@ -1324,14 +1324,14 @@ bool kvm_get_kvm_safe(struct kvm *kvm)
> >  {
> >  	return refcount_inc_not_zero(&kvm->users_count);
> >  }
> > -EXPORT_SYMBOL_GPL(kvm_get_kvm_safe);
> > +EXPORT_SYMBOL_FOR_KVM_INTERNAL(kvm_get_kvm_safe);
> >  
> >  void kvm_put_kvm(struct kvm *kvm)
> >  {
> >  	if (refcount_dec_and_test(&kvm->users_count))
> >  		kvm_destroy_vm(kvm);
> >  }
> > -EXPORT_SYMBOL_GPL(kvm_put_kvm);
> > +EXPORT_SYMBOL_FOR_KVM_INTERNAL(kvm_put_kvm);
> 
> Please isolate the export changes.  They don't *need* to happen at the same time
> as the VFIO changes, i.e. can be done on top.
> 
> >  /*
> >   * Used to put a reference that was taken on behalf of an object associated
> > @@ -1352,6 +1352,8 @@ static int kvm_vm_release(struct inode *inode, struct file *filp)
> >  
> >  	kvm_irqfd_release(kvm);
> >  
> > +	WRITE_ONCE(kvm->file, NULL);
> 
> Please move tracking the file in "struct kvm" to its own patch as well.
> 
> > +
> >  	kvm_put_kvm(kvm);
> >  	return 0;
> >  }
> > @@ -5496,6 +5498,15 @@ bool file_is_kvm(struct file *file)
> >  }
> >  EXPORT_SYMBOL_FOR_KVM_INTERNAL(file_is_kvm);
> >  
> > +struct kvm *file_to_kvm(struct file *file)
> > +{
> > +	if (!file_is_kvm(file))
> 
> file_is_kvm() should exist at the end of this series.  The only reason to ever
> use file_is_kvm() is in advance of getting at "struct kvm", i.e. this should be
> open coded in file_to_kvm().  Though as I suggested in the other subtread, it
> woudl be kvm_file_to_kvm_fn(), i.e. kvm_file_to_kvm_x86() or kvm_file_to_kvm_s390().
> 
> And those changes can and should be done as prep work, i.e. in separate patches,
> e.g. to end up with something like:
> 
>   1. Add kvm_file_to_kvm_fn() and use it on x86, i.e. replace the use of
>      file_is_kvm() in arch/x86/kvm/svm/sev.c.
>   2. Add kvm->file tracking.
>   3. Switch VFIO to tracking the file.
>   4. s/EXPORT_SYMBOL_GPL/EXPORT_SYMBOL_FOR_KVM_INTERNAL on the get/put APIs.
>   5. Enable kvm_file_to_kvm_fn() on s390 and use kvm_file_to_kvm_s390() in the
>      relevant code to prepare for s390+arm64.


I sent the changes incorporated in the main series[1] in 5 patches:

KVM: Introduce file_to_kvm_<arch>() infrastructure 	- adds the macros
KVM: Add file back-pointer to struct kvm		- adds the file field
KVM: x86: Use file_to_kvm_x86() in SEV
KVM/vfio: Use file-based reference counting for KVM	- the swotch in one go
KVM: Restrict kvm_get_kvm/kvm_put_kvm export to internal KVM modules

I needed to do the switch in one go (i.e. merge 5->3) otherwhise there would be conflicts.

Note, that in addition, I moved the symbol verification makefilestuff to
a shared loaction and reused it for s390 in separate patches in the
series.

	Steffen

[1] https://lore.kernel.org/all/20260918133107.1042730-1-seiden@xxxxxxxxxxxxx