Re: [PATCH 1/6] kvm.git mmu notifier patch

[Marcelo Tosatti <mtosatti@redhat.com>]

On Mon, Jul 28, 2008 at 06:11:04PM -0300, Marcelo Tosatti wrote:
> > +static int kvm_unmap_rmapp(struct kvm *kvm, unsigned long *rmapp)
> > +{
> > +	u64 *spte;
> > +	int need_tlb_flush = 0;
> > +
> > +	while ((spte = rmap_next(kvm, rmapp, NULL))) {
                                            ^^^^^^^^

Yeah, I can't read. :-)


> > +		BUG_ON(!(*spte & PT_PRESENT_MASK));
> > +		rmap_printk("kvm_rmap_unmap_hva: spte %p %llx\n", spte, *spte);
> > +		rmap_remove(kvm, spte);
> > +		set_shadow_pte(spte, shadow_trap_nonpresent_pte);
> > +		need_tlb_flush = 1;
> > +	}
> 
> Is rmap_remove() safe during rmap_next() iteration? Don't you need to
> restart the rmap walk from the beginning?
> 
> Remember the discussion around commit
> 6597ca09e6c0e5aec7ffd2b8ab48c671d3c28414.