Re: [PATCH RFC 07/12] Documentation: KVM: Drop mention of kvm->lock in SRCU documentation

[Sean Christopherson]

On Wed, May 27, 2026, Ackerley Tng wrote:
> The original comment says that synchronize_srcu(&kvm->srcu) is called
> inside critical sections for kvm->lock, vcpu->mutex and
> kvm->slots_lock. Drop mention of kvm->lock since this is no longer true.

I would *much* rather "fix" this by saying synchronize_srcu() *may* be called
inside blah blah blah.  Because (a) I don't feel like auditing all of KVM to see
if the above is true, (b) KVM's implementation may change again in the future,
and (c) taking kvm->lock inside a kvm->srcu read-side critical section is still
unsafe as we'd end up with ABBA deadlock (well, ABCCA?).

  1. SRCU held, waiting on kvm->lock
  2. kvm->lock held, waiting on vcpu->mutex
  3. vcpu->mutex held, waiting on synchronize_srcu()