When VFIO finishes assigning a device to user space and calls
iommu_group_release_dma_owner() to return the device to kernel, the IOMMU
core will attach the default domain to the device. Unfortunately, some
IOMMU drivers don't support default domain, hence in the end, the core
calls .detach_dev instead.

This adds set_platform_dma iommu ops to make it clear that what it does
is returning control back to the platform DMA ops.

Suggested-by: Jason Gunthorpe <jgg@nvidia.com>
Signed-off-by: Lu Baolu <baolu.lu@linux.intel.com>
---
 include/linux/iommu.h | 4 ++++
 1 file changed, 4 insertions(+)

diff --git a/include/linux/iommu.h b/include/linux/iommu.h
index 68d7d304cdb7..3542461558fa 100644
--- a/include/linux/iommu.h
+++ b/include/linux/iommu.h
@@ -223,6 +223,9 @@ struct iommu_iotlb_gather {
  * @release_device: Remove device from iommu driver handling
  * @probe_finalize: Do final setup work after the device is added to an IOMMU
  *                  group and attached to the groups domain
+ * @set_platform_dma: Returning control back to the platform DMA ops. Only
+ *                    valid for IOMMU drivers which don't support default
+ *                    domain.
  * @device_group: find iommu group for a particular device
  * @get_resv_regions: Request list of reserved regions for a device
  * @of_xlate: add OF master IDs to iommu grouping
@@ -251,6 +254,7 @@ struct iommu_ops {
 	struct iommu_device *(*probe_device)(struct device *dev);
 	void (*release_device)(struct device *dev);
 	void (*probe_finalize)(struct device *dev);
+	void (*set_platform_dma)(struct device *dev);
 	struct iommu_group *(*device_group)(struct device *dev);
 
 	/* Request/Free a list of reserved regions for a device */
-- 
2.34.1

On Mon, Nov 28, 2022 at 02:46:38PM +0800, Lu Baolu wrote:
> When VFIO finishes assigning a device to user space and calls
> iommu_group_release_dma_owner() to return the device to kernel, the IOMMU
> core will attach the default domain to the device. Unfortunately, some
> IOMMU drivers don't support default domain, hence in the end, the core
> calls .detach_dev instead.
> 
> This adds set_platform_dma iommu ops to make it clear that what it does
> is returning control back to the platform DMA ops.
> 
> Suggested-by: Jason Gunthorpe <jgg@nvidia.com>
> Signed-off-by: Lu Baolu <baolu.lu@linux.intel.com>
> ---
>  include/linux/iommu.h | 4 ++++
>  1 file changed, 4 insertions(+)

Reviewed-by: Jason Gunthorpe <jgg@nvidia.com>

> + * @set_platform_dma: Returning control back to the platform DMA ops. Only
> + *                    valid for IOMMU drivers which don't support default
> + *                    domain.

I would add:

  This op is to support old IOMMU drivers, new drivers should use
  default domains, and the common IOMMU DMA ops.

Jason

On 11/28/22 10:11 PM, Jason Gunthorpe wrote:
> On Mon, Nov 28, 2022 at 02:46:38PM +0800, Lu Baolu wrote:
>> When VFIO finishes assigning a device to user space and calls
>> iommu_group_release_dma_owner() to return the device to kernel, the IOMMU
>> core will attach the default domain to the device. Unfortunately, some
>> IOMMU drivers don't support default domain, hence in the end, the core
>> calls .detach_dev instead.
>>
>> This adds set_platform_dma iommu ops to make it clear that what it does
>> is returning control back to the platform DMA ops.
>>
>> Suggested-by: Jason Gunthorpe<jgg@nvidia.com>
>> Signed-off-by: Lu Baolu<baolu.lu@linux.intel.com>
>> ---
>>   include/linux/iommu.h | 4 ++++
>>   1 file changed, 4 insertions(+)
> Reviewed-by: Jason Gunthorpe<jgg@nvidia.com>
> 
>> + * @set_platform_dma: Returning control back to the platform DMA ops. Only
>> + *                    valid for IOMMU drivers which don't support default
>> + *                    domain.
> I would add:
> 
>    This op is to support old IOMMU drivers, new drivers should use
>    default domains, and the common IOMMU DMA ops.

Done. Thank you!

Best regards,
baolu