Re: [PATCH v3 2/6] vfio/type1: Check reserve region conflict and update iova list
From: Alex Williamson
Date: Mon Feb 19 2018 - 14:43:33 EST
On Mon, 19 Feb 2018 10:00:53 +0000
Shameerali Kolothum Thodi <shameerali.kolothum.thodi@xxxxxxxxxx> wrote:
> > -----Original Message-----
> > From: Alex Williamson [mailto:alex.williamson@xxxxxxxxxx]
> > Sent: Friday, February 16, 2018 9:18 PM
> >
> > On Thu, 15 Feb 2018 09:45:00 +0000
> > Shameer Kolothum <shameerali.kolothum.thodi@xxxxxxxxxx> wrote:
> > > + iommu_get_group_resv_regions(iommu_group, &group_resv_regions);
> > > +
> > > + if (vfio_iommu_resv_conflict(iommu, &group_resv_regions)) {
> > > + ret = -EINVAL;
> > > + goto out_detach;
> > > + }
> > > +
> > > /* Get a copy of the current iova list and work on it */
> > > INIT_LIST_HEAD(&iova_copy);
> > > ret = vfio_iommu_get_iova_copy(iommu, &iova_copy);
> > > @@ -1437,6 +1512,10 @@ static int vfio_iommu_type1_attach_group(void
> > *iommu_data,
> > > if (ret)
> > > goto out_detach;
> > >
> > > + ret = vfio_iommu_resv_exclude(&iova_copy, &group_resv_regions);
> > > + if (ret)
> > > + goto out_detach;
> > > +
> > > resv_msi = vfio_iommu_has_sw_msi(iommu_group, &resv_msi_base);
> > >
> > > INIT_LIST_HEAD(&domain->group_list);
> > > @@ -1497,6 +1576,9 @@ static int vfio_iommu_type1_attach_group(void
> > *iommu_data,
> > > /* Delete the old one and insert new iova list */
> > > vfio_iommu_insert_iova_copy(iommu, &iova_copy);
> > >
> > > + list_for_each_entry_safe(resv, resv_next, &group_resv_regions, list)
> > > + kfree(resv);
> >
> > list_del() here and below, also this can be done after the mutex unlock.
>
> Ok. I thought that as the reserved regions are local to this function, list_del() is
> not required. Same for the iova_copy in the first patch as well(which I missed to
> comment there).
What you have works (afaik), it just seems sloppy to me to have free'd
entries in the list, even as the destructor, as this is not a
performance critical path. Is this more common elsewhere in the kernel
than I suspect? Thanks,
Alex