On Mon, Sep 23, 2019 at 12:28:35PM +0200, Adrian Moreno wrote:
> On 9/23/19 9:25 AM, Tiwei Bie wrote:
> > On Tue, Sep 17, 2019 at 04:49:00PM +0200, Adrian Moreno wrote:
> >> When IOMMU is enabled the incoming log address is in IOVA space. In that
> >> case, look in IOTLB table and translate the resulting HVA to GPA.
> >>
> >> If IOMMU is not enabled, the incoming log address is already a GPA so no
> >> transformation is needed.
> >>
> >> This change makes page logging work when IOVA_VA is selected in the guest.
> >
> > Besides the log address of the ring, when IOMMU is enabled,
> > the addresses in descriptors are also IOVAs and should be
> > translated to GPAs before doing the dirty page logging.
> >
> Thanks Tiwei. You're right. In fact, it's not the only place where IOVAs are
> assumed to be GPAs, for example in vhost.h:gpa_to_hpa:
>
> /* Convert guest physical address to host physical address */
> static __rte_always_inline rte_iova_t
> gpa_to_hpa(struct virtio_net *dev, uint64_t gpa, uint64_t size)
> {
> uint32_t i;
> struct guest_page *page;
>
> for (i = 0; i < dev->nr_guest_pages; i++) {
> page = &dev->guest_pages[i];
>
> if (gpa >= page->guest_phys_addr &&
> gpa + size < page->guest_phys_addr + page->size) {
> return gpa - page->guest_phys_addr +
> page->host_phys_addr;
> }
> }
>
> return 0;
> }
>
> used in ZERO-COPY mode to check if the IOVA range is continuous in host
> physical
> memory and called buffer IOVAs. If I'm not mistaken this should also be
> translated as:
>
> IOVA --> HVA (iotlb lookup)
> HVA --> GPA (mem_region lookup)
> GPA --> HIOVA (guest_page lookup)
>
> Right?
Yeah, these IOVAs also shouldn't be assumed as GPAs.
Thanks,
Tiwei
