Re: [PATCH v3 08/20] iommu/sprd: Remove detach_dev callback

[Chunyan Zhang]

On Tue, 29 Nov 2022 at 11:35, Tian, Kevin <kevin.tian@xxxxxxxxx> wrote:
>
> > From: Lu Baolu <baolu.lu@xxxxxxxxxxxxxxx>
> > Sent: Monday, November 28, 2022 2:47 PM
> >
> > The IOMMU driver supports default domain, so the detach_dev op will never
> > be called. Remove it to avoid dead code.
> >
> > Signed-off-by: Lu Baolu <baolu.lu@xxxxxxxxxxxxxxx>
> > ---
> >  drivers/iommu/sprd-iommu.c | 16 ----------------
> >  1 file changed, 16 deletions(-)
> >
> > diff --git a/drivers/iommu/sprd-iommu.c b/drivers/iommu/sprd-iommu.c
> > index 219bfa11f7f4..ae94d74b73f4 100644
> > --- a/drivers/iommu/sprd-iommu.c
> > +++ b/drivers/iommu/sprd-iommu.c
> > @@ -255,21 +255,6 @@ static int sprd_iommu_attach_device(struct
> > iommu_domain *domain,
> >       return 0;
> >  }
> >
> > -static void sprd_iommu_detach_device(struct iommu_domain *domain,
> > -                                          struct device *dev)
> > -{
> > -     struct sprd_iommu_domain *dom = to_sprd_domain(domain);
> > -     struct sprd_iommu_device *sdev = dom->sdev;
> > -     size_t pgt_size = sprd_iommu_pgt_size(domain);
> > -
> > -     if (!sdev)
> > -             return;
> > -
> > -     dma_free_coherent(sdev->dev, pgt_size, dom->pgt_va, dom-
> > >pgt_pa);
> > -     sprd_iommu_hw_en(sdev, false);
> > -     dom->sdev = NULL;
> > -}
> > -
>
> Looks this reveals a bug in this driver (not caused by this removal).
>
> sprd_iommu_attach_device() doesn't check whether the device has
> been already attached to a domain and do auto detach.
>
> It's written in a way that .detach_dev() must be called to free the
> dma buffer but ignores the fact that it's not called when default
> domain support is claimed.
>
> Then the dma buffer allocated for the previous domain was left
> unhandled, causing memory leak.

I'll look into the issue, thanks for pointing this out.

Chunyan