Re: [PATCH] kvm: defer huge page recovery vhost task to later

[Keith Busch]

On Fri, Jan 24, 2025 at 10:28:03AM -0500, Paolo Bonzini wrote:
> > Defer the task to after the first VM_RUN call, which occurs after the
> > parent process has forked all its jailed processes. This needs to happen
> > only once for the kvm instance, so this patch introduces infrastructure
> > to do that (Suggested-by Paolo).
> 
> Queued for 6.13; in the end I moved the new data structure to include/linux,
> since it is generally usable and not limited to KVM.

Thanks! I see that you also added the "Fixes" tag that I forgot to
append in the most recent version, so thank you for that.
 
> >  int kvm_arch_post_init_vm(struct kvm *kvm)
> >  {
> > -	return kvm_mmu_post_init_vm(kvm);
> > +	once_init(&kvm->arch.nx_once);
> > +	return 0;
> >  }
> 
> This could have been in kvm_arch_init_vm(), but then the last user of
> kvm_arch_post_init_vm() goes away and more cleanup is in order.  I'll
> post the obvious patch shortly.

Yes, that makes sense. I had a similiar cleanup in the first version
too.