Re: [PATCH v10 02/14] KVM: Cache as_id in kvm_memory_slot

[Sean Christopherson]

On Fri, Jul 03, 2020 at 02:41:22PM -0400, Peter Xu wrote:
> On Thu, Jul 02, 2020 at 04:08:49PM -0700, Sean Christopherson wrote:
> > On Mon, Jun 01, 2020 at 07:59:45AM -0400, Peter Xu wrote:
> > > Cache the address space ID just like the slot ID.  It will be used in
> > > order to fill in the dirty ring entries.
> > > 
> > > Suggested-by: Paolo Bonzini <pbonzini@xxxxxxxxxx>
> > > Suggested-by: Sean Christopherson <sean.j.christopherson@xxxxxxxxx>
> > > Signed-off-by: Peter Xu <peterx@xxxxxxxxxx>
> > > ---
> > >  include/linux/kvm_host.h | 1 +
> > >  virt/kvm/kvm_main.c      | 1 +
> > >  2 files changed, 2 insertions(+)
> > > 
> > > diff --git a/include/linux/kvm_host.h b/include/linux/kvm_host.h
> > > index 01276e3d01b9..5e7bbaf7a36b 100644
> > > --- a/include/linux/kvm_host.h
> > > +++ b/include/linux/kvm_host.h
> > > @@ -346,6 +346,7 @@ struct kvm_memory_slot {
> > >  	unsigned long userspace_addr;
> > >  	u32 flags;
> > >  	short id;
> > > +	u16 as_id;
> > >  };
> > >  
> > >  static inline unsigned long kvm_dirty_bitmap_bytes(struct kvm_memory_slot *memslot)
> > > diff --git a/virt/kvm/kvm_main.c b/virt/kvm/kvm_main.c
> > > index 74bdb7bf3295..ebdd98a30e82 100644
> > > --- a/virt/kvm/kvm_main.c
> > > +++ b/virt/kvm/kvm_main.c
> > > @@ -1243,6 +1243,7 @@ int __kvm_set_memory_region(struct kvm *kvm,
> > >  	if (!mem->memory_size)
> > >  		return kvm_delete_memslot(kvm, mem, &old, as_id);
> > 
> > This technically needs to set as_id in the deleted memslot.  I highly doubt
> > it will ever matter from a functionality perspective, but it'd be confusing
> > to encounter a memslot whose as_id did not match that of its owner.
> 
> Yeah it shouldn't matter because as_id is directly passed in to look up the
> pointer of kvm_memslots in kvm_delete_memslot, and memslot->as_id shouldn't be
> further referenced.
> 
> I can add a comment above if this can clarify things a bit:
> 
> +	u16 as_id; /* cache of as_id; only valid if npages != 0 */

Why not just set it?  It's a single line of code, and there's more than one
"shouldn't" in the above.