Re: [PATCH v3 02/27] KVM, vfio: remove symbol_get(kvm_get_kvm_safe) from vfio

[sashiko-bot@kernel.org]

Thank you for your contribution! Sashiko AI review found 1 potential issue(s) to consider:
- [Low] Build failure in arch/x86/kvm/svm/sev.c due to missing `users_count` struct field update.
--

commit d29c547487163d60419f67e02b219cd59361c39b
Author: Paolo Bonzini <pbonzini@redhat.com>

KVM, vfio: remove symbol_get(kvm_get_kvm_safe) from vfio

This removes the deprecated symbol_get usage for kvm_get_kvm_safe in VFIO
by making the function inline. As part of this, the function is moved to
kvm_types.h and the users_count field in struct kvm is moved into a new
kvm_refcount struct.

[ ... ]

> diff --git a/include/linux/kvm_host.h b/include/linux/kvm_host.h
> index 4c14aee1fb063..8a068cbf12ff2 100644
> --- a/include/linux/kvm_host.h
> +++ b/include/linux/kvm_host.h
> @@ -768,6 +768,9 @@ struct kvm_memslots {
>  };
>  
>  struct kvm {
> +	/* Must be the first field, see function definitions in kvm_types.h.  */
> +	struct kvm_refcount rc;
> +
>  #ifdef KVM_HAVE_MMU_RWLOCK
>  	rwlock_t mmu_lock;
>  #else
> @@ -831,7 +834,6 @@ struct kvm {
>  	struct list_head ioeventfds;
>  	struct kvm_vm_stat stat;
>  	struct kvm_arch arch;
> -	refcount_t users_count;

[Severity: Low]
Does removing users_count from struct kvm break the build for x86 with AMD
SEV?

Looking at kvm_lockdep_assert_sev_lock_held() in arch/x86/kvm/svm/sev.c, it
still references this directly:

arch/x86/kvm/svm/sev.c:kvm_lockdep_assert_sev_lock_held() {
    ...
	/*
	 * Querying SEV+ support is safe if there are no other references, i.e.
	 * if concurrent initialization of SEV+ is impossible.
	 */
	if (!refcount_read(&kvm->users_count))
		return;
    ...
}

Should this reference be updated to use kvm->rc.users_count like the other
files modified in this patch?

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/20260529155050.2902245-1-seiden@linux.ibm.com?part=2