[RFC PATCH] iommu: map reserved memory as cacheable if device is coherent

[RFC PATCH] iommu: map reserved memory as cacheable if device is coherent

[laurentiu.tudor]

From: Laurentiu Tudor <laurentiu.tudor@nxp.com>

Check if the device is marked as DMA coherent in the DT and if so,
map its reserved memory as cacheable in the IOMMU.

Signed-off-by: Laurentiu Tudor <laurentiu.tudor@nxp.com>
---
 drivers/iommu/of_iommu.c | 3 +++
 1 file changed, 3 insertions(+)

diff --git a/drivers/iommu/of_iommu.c b/drivers/iommu/of_iommu.c
index 40f57d293a79..ccb912049b88 100644
--- a/drivers/iommu/of_iommu.c
+++ b/drivers/iommu/of_iommu.c
@@ -254,6 +254,9 @@ void of_iommu_get_resv_regions(struct device *dev, struct list_head *list)
 				phys_addr_t iova;
 				size_t length;
 
+				if (of_dma_is_coherent(dev->of_node))
+					prot |= IOMMU_CACHE;
+
 				maps = of_translate_dma_region(np, maps, &iova, &length);
 				type = iommu_resv_region_get_type(dev, &phys, iova, length);
 
-- 
2.17.1

Re: [RFC PATCH] iommu: map reserved memory as cacheable if device is coherent

[Jason Gunthorpe]

On Tue, Jul 25, 2023 at 05:00:55PM +0300, laurentiu.tudor@nxp.com wrote:
> From: Laurentiu Tudor <laurentiu.tudor@nxp.com>
> 
> Check if the device is marked as DMA coherent in the DT and if so,
> map its reserved memory as cacheable in the IOMMU.
> 
> Signed-off-by: Laurentiu Tudor <laurentiu.tudor@nxp.com>
> ---
>  drivers/iommu/of_iommu.c | 3 +++
>  1 file changed, 3 insertions(+)

This only makes sense for IOMMU_RESV_DIRECT - is that your use case?
Maybe mention that in the commit message

And add a fixes for the commit that introduces IOMMU_RESV_DIRECT into
of_iommu.c

Jason

Re: [RFC PATCH] iommu: map reserved memory as cacheable if device is coherent

[Laurentiu Tudor]

On 7/26/2023 3:22 PM, Jason Gunthorpe wrote:
> On Tue, Jul 25, 2023 at 05:00:55PM +0300, laurentiu.tudor@nxp.com wrote:
>> From: Laurentiu Tudor <laurentiu.tudor@nxp.com>
>>
>> Check if the device is marked as DMA coherent in the DT and if so,
>> map its reserved memory as cacheable in the IOMMU.
>>
>> Signed-off-by: Laurentiu Tudor <laurentiu.tudor@nxp.com>
>> ---
>>   drivers/iommu/of_iommu.c | 3 +++
>>   1 file changed, 3 insertions(+)
> 
> This only makes sense for IOMMU_RESV_DIRECT - is that your use case?
> Maybe mention that in the commit message

Yep. Will respin with an improved commit message.

> And add a fixes for the commit that introduces IOMMU_RESV_DIRECT into
> of_iommu.c

Would it make more sense to mention the commit [1] that actually added 
this support for mapping reserved memory in IOMMU?

[1] 
https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=a5bf3cfce8cb77d9d24613ab52d520896f83dd48

---
Thanks & Best Regards, Laurentiu

Re: [RFC PATCH] iommu: map reserved memory as cacheable if device is coherent

[Jason Gunthorpe]

On Wed, Jul 26, 2023 at 03:29:47PM +0300, Laurentiu Tudor wrote:
> 
> 
> On 7/26/2023 3:22 PM, Jason Gunthorpe wrote:
> > On Tue, Jul 25, 2023 at 05:00:55PM +0300, laurentiu.tudor@nxp.com wrote:
> > > From: Laurentiu Tudor <laurentiu.tudor@nxp.com>
> > > 
> > > Check if the device is marked as DMA coherent in the DT and if so,
> > > map its reserved memory as cacheable in the IOMMU.
> > > 
> > > Signed-off-by: Laurentiu Tudor <laurentiu.tudor@nxp.com>
> > > ---
> > >   drivers/iommu/of_iommu.c | 3 +++
> > >   1 file changed, 3 insertions(+)
> > 
> > This only makes sense for IOMMU_RESV_DIRECT - is that your use case?
> > Maybe mention that in the commit message
> 
> Yep. Will respin with an improved commit message.
> 
> > And add a fixes for the commit that introduces IOMMU_RESV_DIRECT into
> > of_iommu.c
> 
> Would it make more sense to mention the commit [1] that actually added this
> support for mapping reserved memory in IOMMU?
> 
> [1] https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=a5bf3cfce8cb77d9d24613ab52d520896f83dd48

Yes, that should be marked Fixes: - it added IOMMU_RESV_DIRECT without
forming the PROT properly

Jason