[2/4] media: omap3isp: Handle ARM dma_iommu_mapping

It's no longer practical for the OMAP IOMMU driver to trick
arm_setup_iommu_dma_ops() into ignoring its presence, so let's use the
same tactic as other IOMMU API users on 32-bit ARM and explicitly kick
the arch code's dma_iommu_mapping out of the way to avoid problems.

Fixes: 4720287c7bf7 ("iommu: Remove struct iommu_ops *iommu from arch_setup_dma_ops()")
Signed-off-by: Robin Murphy <robin.murphy@arm.com>
---
 drivers/media/platform/ti/omap3isp/isp.c | 7 +++++++
 1 file changed, 7 insertions(+)

Hi

On Mon, Oct 28, 2024 at 05:58:36PM +0000, Robin Murphy wrote:
> It's no longer practical for the OMAP IOMMU driver to trick
> arm_setup_iommu_dma_ops() into ignoring its presence, so let's use the
> same tactic as other IOMMU API users on 32-bit ARM and explicitly kick
> the arch code's dma_iommu_mapping out of the way to avoid problems.
> 
> Fixes: 4720287c7bf7 ("iommu: Remove struct iommu_ops *iommu from arch_setup_dma_ops()")
> Signed-off-by: Robin Murphy <robin.murphy@arm.com>
> ---
>  drivers/media/platform/ti/omap3isp/isp.c | 7 +++++++
>  1 file changed, 7 insertions(+)
> 
> diff --git a/drivers/media/platform/ti/omap3isp/isp.c b/drivers/media/platform/ti/omap3isp/isp.c
> index 91101ba88ef0..b2210841a320 100644
> --- a/drivers/media/platform/ti/omap3isp/isp.c
> +++ b/drivers/media/platform/ti/omap3isp/isp.c
> @@ -1961,6 +1961,13 @@ static int isp_attach_iommu(struct isp_device *isp)
>  	struct dma_iommu_mapping *mapping;
>  	int ret;
>  
> +	/* We always want to replace any default mapping from the arch code */
> +	mapping = to_dma_iommu_mapping(isp->dev);
> +	if (mapping) {
> +		arm_iommu_detach_device(isp->dev);
> +		arm_iommu_release_mapping(mapping);
> +	}
> +
>  	/*
>  	 * Create the ARM mapping, used by the ARM DMA mapping core to allocate
>  	 * VAs. This will allocate a corresponding IOMMU domain.
> -- 
> 2.39.2.101.g768bb238c484.dirty
> 

I have finally found time to test this patch on the Nokia N900 and can
confirm it is working fine.

I was wondering - is there a reason that it is not merged yet? I tested
on 6.14-rc2, which did not have it, and notice it is also not in
linux-next.

If it helps:

Tested-by: Sicelo A. Mhlongo <absicsz@gmail.com>

CC'ing Sakari.

Sakari, would you pick this patch ?

On Sat, Feb 15, 2025 at 08:57:24AM +0200, Sicelo wrote:
> On Mon, Oct 28, 2024 at 05:58:36PM +0000, Robin Murphy wrote:
> > It's no longer practical for the OMAP IOMMU driver to trick
> > arm_setup_iommu_dma_ops() into ignoring its presence, so let's use the
> > same tactic as other IOMMU API users on 32-bit ARM and explicitly kick
> > the arch code's dma_iommu_mapping out of the way to avoid problems.
> > 
> > Fixes: 4720287c7bf7 ("iommu: Remove struct iommu_ops *iommu from arch_setup_dma_ops()")
> > Signed-off-by: Robin Murphy <robin.murphy@arm.com>
> > ---
> >  drivers/media/platform/ti/omap3isp/isp.c | 7 +++++++
> >  1 file changed, 7 insertions(+)
> > 
> > diff --git a/drivers/media/platform/ti/omap3isp/isp.c b/drivers/media/platform/ti/omap3isp/isp.c
> > index 91101ba88ef0..b2210841a320 100644
> > --- a/drivers/media/platform/ti/omap3isp/isp.c
> > +++ b/drivers/media/platform/ti/omap3isp/isp.c
> > @@ -1961,6 +1961,13 @@ static int isp_attach_iommu(struct isp_device *isp)
> >  	struct dma_iommu_mapping *mapping;
> >  	int ret;
> >  
> > +	/* We always want to replace any default mapping from the arch code */
> > +	mapping = to_dma_iommu_mapping(isp->dev);
> > +	if (mapping) {
> > +		arm_iommu_detach_device(isp->dev);
> > +		arm_iommu_release_mapping(mapping);
> > +	}
> > +
> >  	/*
> >  	 * Create the ARM mapping, used by the ARM DMA mapping core to allocate
> >  	 * VAs. This will allocate a corresponding IOMMU domain.
> > -- 
> > 2.39.2.101.g768bb238c484.dirty
> > 
> 
> I have finally found time to test this patch on the Nokia N900 and can
> confirm it is working fine.
> 
> I was wondering - is there a reason that it is not merged yet? I tested
> on 6.14-rc2, which did not have it, and notice it is also not in
> linux-next.
> 
> If it helps:
> 
> Tested-by: Sicelo A. Mhlongo <absicsz@gmail.com>