Re: [RFC v8 3/5] memory: Add IOMMU_DEVIOTLB_UNMAP IOMMUTLBNotificationType

[Eugenio Perez Martin <eperezma@redhat.com>]

On Thu, Sep 3, 2020 at 1:06 PM Auger Eric <eric.auger@redhat.com> wrote:
>
> Hi Eugenio,
> On 9/3/20 12:13 PM, Eugenio Perez Martin wrote:
> > Hi Eric,
> >
> > On Wed, Sep 2, 2020 at 12:32 PM Auger Eric <eric.auger@redhat.com> wrote:
> >>
> >> Hi Eugenio,
> >>
> >> On 9/1/20 4:26 PM, Eugenio Pérez wrote:
> >>> Adapt intel and vhost to use this new notification type
> >> I think you should explain in the commit message what is the benefice to
> >> introduce this new event type.
> >
> > Will do, thanks!
> >
> >>>
> >>> Signed-off-by: Eugenio Pérez <eperezma@redhat.com>
> >>> ---
> >>>  hw/i386/intel_iommu.c | 2 +-
> >>>  hw/virtio/vhost.c     | 2 +-
> >>>  include/exec/memory.h | 2 ++
> >>>  3 files changed, 4 insertions(+), 2 deletions(-)
> >>>
> >>> diff --git a/hw/i386/intel_iommu.c b/hw/i386/intel_iommu.c
> >>> index 0c4aef5cb5..cdddb089e7 100644
> >>> --- a/hw/i386/intel_iommu.c
> >>> +++ b/hw/i386/intel_iommu.c
> >>> @@ -2468,7 +2468,7 @@ static bool vtd_process_device_iotlb_desc(IntelIOMMUState *s,
> >>>          sz = VTD_PAGE_SIZE;
> >>>      }
> >>>
> >>> -    event.type = IOMMU_NOTIFIER_UNMAP;
> >>> +    event.type = IOMMU_NOTIFIER_DEVIOTLB;
> >> If this is used only for device IOTLB cache invalidation, shouldn't this
> >> be named IOMMU_NOTIFIER_DEVIOTLB_UNMAP to be consistent with the rest?
> >>>      event.entry.target_as = &vtd_dev_as->as;
> >>>      event.entry.addr_mask = sz - 1;
> >>>      event.entry.iova = addr;
> >>> diff --git a/hw/virtio/vhost.c b/hw/virtio/vhost.c
> >>> index 1a1384e7a6..6ca168b47e 100644
> >>> --- a/hw/virtio/vhost.c
> >>> +++ b/hw/virtio/vhost.c
> >>> @@ -729,7 +729,7 @@ static void vhost_iommu_region_add(MemoryListener *listener,
> >>>      iommu_idx = memory_region_iommu_attrs_to_index(iommu_mr,
> >>>                                                     MEMTXATTRS_UNSPECIFIED);
> >>>      iommu_notifier_init(&iommu->n, vhost_iommu_unmap_notify,
> >>> -                        IOMMU_NOTIFIER_UNMAP,
> >>> +                        IOMMU_NOTIFIER_DEVIOTLB,
> >>>                          section->offset_within_region,
> >>>                          int128_get64(end),
> >>>                          iommu_idx);
> >>> diff --git a/include/exec/memory.h b/include/exec/memory.h
> >>> index 8a56707169..215e23973d 100644
> >>> --- a/include/exec/memory.h
> >>> +++ b/include/exec/memory.h
> >>> @@ -87,6 +87,8 @@ typedef enum {
> >>>      IOMMU_NOTIFIER_UNMAP = 0x1,
> >>>      /* Notify entry changes (newly created entries) */
> >>>      IOMMU_NOTIFIER_MAP = 0x2,
> >>> +    /* Notify changes on device IOTLB entries */
> >>> +    IOMMU_NOTIFIER_DEVIOTLB = 0x04,
> >>>  } IOMMUNotifierFlag;
> >>>
> >>>  #define IOMMU_NOTIFIER_ALL (IOMMU_NOTIFIER_MAP | IOMMU_NOTIFIER_UNMAP)
> >> shouldn't we rename this one??
> >>>
> >>
> >
> > Agree, but I'm not sure about the right name. IOMMU_NOTIFIER_ALL_ROOT?
> > IOMMU_NOTIFIER_ALL_REGULAR?
> I would rather name it IOMMU_NOTIFIER_IOTLB_EVENTS versus
> IOMMU_NOTIFIER_DEVIOTLB_EVENTS? This is the cache type that differs,
> isn't it?
>

Ok will propose it.

Thanks!

> Thanks
>
> Eric
> >
> > Thanks!
> >
> >> Thanks
> >>
> >> Eric
> >>
> >
> >
>