Re: [patch 3/4] KVM: MMU: reload request from GET_DIRTY_LOG path

[Marcelo Tosatti <mtosatti@redhat.com>]

On Tue, Jul 22, 2014 at 05:55:20AM +0800, Xiao Guangrong wrote:
> 
> On Jul 10, 2014, at 3:12 AM, mtosatti@redhat.com wrote:
> 
> > Reload remote vcpus MMU from GET_DIRTY_LOG codepath, before
> > deleting a pinned spte.
> > 
> > Signed-off-by: Marcelo Tosatti <mtosatti@redhat.com>
> > 
> > ---
> > arch/x86/kvm/mmu.c |   29 +++++++++++++++++++++++------
> > 1 file changed, 23 insertions(+), 6 deletions(-)
> > 
> > Index: kvm.pinned-sptes/arch/x86/kvm/mmu.c
> > ===================================================================
> > --- kvm.pinned-sptes.orig/arch/x86/kvm/mmu.c	2014-07-09 11:23:59.290744490 -0300
> > +++ kvm.pinned-sptes/arch/x86/kvm/mmu.c	2014-07-09 11:24:58.449632435 -0300
> > @@ -1208,7 +1208,8 @@
> > *
> > * Return true if tlb need be flushed.
> > */
> > -static bool spte_write_protect(struct kvm *kvm, u64 *sptep, bool pt_protect)
> > +static bool spte_write_protect(struct kvm *kvm, u64 *sptep, bool pt_protect,
> > +			       bool skip_pinned)
> > {
> > 	u64 spte = *sptep;
> > 
> > @@ -1218,6 +1219,22 @@
> > 
> > 	rmap_printk("rmap_write_protect: spte %p %llx\n", sptep, *sptep);
> > 
> > +	if (is_pinned_spte(spte)) {
> > +		/* keep pinned spte intact, mark page dirty again */
> > +		if (skip_pinned) {
> > +			struct kvm_mmu_page *sp;
> > +			gfn_t gfn;
> > +
> > +			sp = page_header(__pa(sptep));
> > +			gfn = kvm_mmu_page_get_gfn(sp, sptep - sp->spt);
> > +
> > +			mark_page_dirty(kvm, gfn);
> > +			return false;
> > +		} else
> > +			mmu_reload_pinned_vcpus(kvm);
> > +	}
> > +
> > +
> > 	if (pt_protect)
> > 		spte &= ~SPTE_MMU_WRITEABLE;
> > 	spte = spte & ~PT_WRITABLE_MASK;
> 
> This is also a window between marking spte readonly and re-ping…
> IIUC, I think all spte spte can not be zapped and write-protected at any time

It is safe because mmu_lock is held by kvm_mmu_slot_remove_write_access
?