Re: [PATCH] vfio-pci: Allow to mmap sub-page MMIO BARs if the mmio page is exclusive

[Yongji Xie]

On 2016/5/3 10:59, Tian, Kevin wrote:

> > From: Yongji Xie
> > Sent: Wednesday, April 27, 2016 8:22 PM
> >
> > Current vfio-pci implementation disallows to mmap
> > sub-page(size < PAGE_SIZE) MMIO BARs because these BARs' mmio
> > page may be shared with other BARs. This will cause some
> > performance issues when we passthrough a PCI device with
> > this kind of BARs. Guest will be not able to handle the mmio
> > accesses to the BARs which leads to mmio emulations in host.
> >
> > However, not all sub-page BARs will share page with other BARs.
> > We should allow to mmap those sub-page MMIO BARs which we can
> > make sure will not share page with other BARs.
> >
> > This patch adds support for this case. And we also try to use
> > shadow resource to reserve the remaind of the page which hot-add
> > device's BAR might be assigned into.
> 'shadow' usually means you have a corresponding part being
> shadowed, while here looks you mostly want some 'dummy'
> resource for reservation purpose?

Yes., 'dummy' may be better here. And I would also replace
shadow_res/shadow_resources_list with reserved_res/reserved_resources_list.

> > +
> > +		if (!(res->start & ~PAGE_MASK)) {
> > +			/*
> > +			 * Add shadow resource for sub-page bar whose mmio
> > +			 * page is exclusive in case that hot-add device's
> > +			 * bar is assigned into the mem hole.
> > +			 */
> > +			shadow_res = kzalloc(sizeof(*shadow_res), GFP_KERNEL);
> > +			shadow_res->resource.start = res->end + 1;
> > +			shadow_res->resource.end = res->start + PAGE_SIZE - 1;
> What about res->start not page aligned so you end up still having
> a portion before res->start not exclusively reserved?

Do you mean add a 'dummy' resource to reserve the portion
before res->start if res->start not page aligned?

But would it happen that there is a mem hole in the portion
before res->start? The resource should have been assigned
into the hole at the beginning.

Thanks,
Yongji

> > +			shadow_res->resource.flags = res->flags;
> > +			if (request_resource(res->parent,
> > +					&shadow_res->resource)) {
> > +				kfree(shadow_res);
> > +				return false;
> > +			}
> > +			shadow_res->index = index;
> > +			list_add(&shadow_res->res_next,
> > +					&vdev->shadow_resources_list);
> > +			return true;
> Thanks
> Kevin
> --
> To unsubscribe from this list: send the line "unsubscribe kvm" in
> the body of a message to majordomo@xxxxxxxxxxxxxxx
> More majordomo info at  http://vger.kernel.org/majordomo-info.html