Re: [PATCH 1/2] iommu/vt-d: Use rbtree to track iommu probed devices

[Baolu Lu]

On 2024/2/19 10:45, Ethan Zhao wrote:
> > @@ -4264,25 +4313,34 @@ static struct iommu_device 
> > *intel_iommu_probe_device(struct device *dev)
> >       }
> >       dev_iommu_priv_set(dev, info);
> > +    ret = device_rbtree_insert(iommu, info);
> > +    if (ret)
> > +        goto free;
> >       if (sm_supported(iommu) && !dev_is_real_dma_subdevice(dev)) {
> >           ret = intel_pasid_alloc_table(dev);
> >           if (ret) {
> >               dev_err(dev, "PASID table allocation failed\n");
> > -            kfree(info);
> > -            return ERR_PTR(ret);
> > +            goto clear_rbtree;
> >           }
> >       }
> >       intel_iommu_debugfs_create_dev(info);
> >       return &iommu->iommu;
> > +clear_rbtree:
> > +    device_rbtree_remove(info);
> > +free:
> > +    kfree(info);
> > +
> > +    return ERR_PTR(ret);
> >   }
> >   static void intel_iommu_release_device(struct device *dev)
> >   {
> >       struct device_domain_info *info = dev_iommu_priv_get(dev);
> > +    device_rbtree_remove(info);
>
> Perhpas too early here to remove dev from the rbtree, if it is wanted in
> devTLB invalidation steps in intel_pasid_tear_down_entry().

Perhaps the caller of device_rbtree_find() should not depend on the
order in the release_device callback. For the device TLB invalidation
timed-out case, probably it could be checked in this way:

    struct device *dev = device_rbtree_find(iommu, ite_sid);
    if (!dev || !pci_device_is_present(to_pci_dev(dev)))
        return -ETIMEDOUT;

Best regards,
baolu