Re: [PATCH 6/7] iommu: Use right way to retrieve iommu_ops

[Jason Gunthorpe <jgg@nvidia.com>]

On Mon, Jan 24, 2022 at 03:11:01PM +0800, Lu Baolu wrote:
> The common iommu_ops is hooked to both device and domain. When a helper
> has both device and domain pointer, the way to get the iommu_ops looks
> messy in iommu core. This sorts out the way to get iommu_ops. The device
> related helpers go through device pointer, while the domain related ones
> go through domain pointer.
> 
> Signed-off-by: Lu Baolu <baolu.lu@linux.intel.com>
>  include/linux/iommu.h |  8 ++++++++
>  drivers/iommu/iommu.c | 25 ++++++++++++++-----------
>  2 files changed, 22 insertions(+), 11 deletions(-)
> 
> diff --git a/include/linux/iommu.h b/include/linux/iommu.h
> index aa5486243892..111b3e9c79bb 100644
> +++ b/include/linux/iommu.h
> @@ -385,6 +385,14 @@ static inline void iommu_iotlb_gather_init(struct iommu_iotlb_gather *gather)
>  	};
>  }
>  
> +static inline const struct iommu_ops *dev_iommu_ops_get(struct device *dev)
> +{
> +	if (dev && dev->iommu && dev->iommu->iommu_dev)
> +		return dev->iommu->iommu_dev->ops;
> +
> +	return NULL;

What is the purpose of this helper?

> +	const struct iommu_ops *ops = dev_iommu_ops_get(dev);
>  	struct iommu_domain *domain = group->default_domain;
>  	struct iommu_resv_region *entry;
>  	struct list_head mappings;
> @@ -785,8 +786,8 @@ static int iommu_create_device_direct_mappings(struct iommu_group *group,
>  		dma_addr_t start, end, addr;
>  		size_t map_size = 0;
>  
> -		if (domain->ops->apply_resv_region)
> -			domain->ops->apply_resv_region(dev, domain, entry);
> +		if (ops->apply_resv_region)
> +			ops->apply_resv_region(dev, domain, entry);

Here we call it and don't check for NULL? So why did we check the
interior pointers in the helper?

> @@ -831,8 +832,10 @@ static int iommu_create_device_direct_mappings(struct iommu_group *group,
>  static bool iommu_is_attach_deferred(struct iommu_domain *domain,
>  				     struct device *dev)
>  {
> -	if (domain->ops->is_attach_deferred)
> -		return domain->ops->is_attach_deferred(domain, dev);
> +	const struct iommu_ops *ops = dev_iommu_ops_get(dev);
> +
> +	if (ops->is_attach_deferred)
> +		return ops->is_attach_deferred(domain, dev);

Same here, at least return false if ops is null..
  
> @@ -1251,10 +1254,10 @@ int iommu_page_response(struct device *dev,
>  	struct iommu_fault_event *evt;
>  	struct iommu_fault_page_request *prm;
>  	struct dev_iommu *param = dev->iommu;
> +	const struct iommu_ops *ops = dev_iommu_ops_get(dev);
>  	bool has_pasid = msg->flags & IOMMU_PAGE_RESP_PASID_VALID;
> -	struct iommu_domain *domain = iommu_get_domain_for_dev(dev);
>  
> -	if (!domain || !domain->ops->page_response)
> +	if (!ops || !ops->page_response)
>  		return -ENODEV;
>  
>  	if (!param || !param->fault_param)
> @@ -1295,7 +1298,7 @@ int iommu_page_response(struct device *dev,
>  			msg->pasid = 0;
>  		}
>  
> -		ret = domain->ops->page_response(dev, evt, msg);
> +		ret = ops->page_response(dev, evt, msg);
>  		list_del(&evt->list);
>  		kfree(evt);
>  		break;

Feels weird that page_response is not connected to a domain, the fault
originated from a domain after all. I would say this op should be
moved to the domain and the caller should provide the a pointer to the
domain that originated the fault.

Ideally since only some domain's will be configured to handle faults
at all - domains that can't do this should have a NULL page_response
op, even if other domains created by the same device driver could
handle page_response..

> @@ -1758,10 +1761,10 @@ static int __iommu_group_dma_attach(struct iommu_group *group)
>  
>  static int iommu_group_do_probe_finalize(struct device *dev, void *data)
>  {
> -	struct iommu_domain *domain = data;
> +	const struct iommu_ops *ops = dev_iommu_ops_get(dev);
>  
> -	if (domain->ops->probe_finalize)
> -		domain->ops->probe_finalize(dev);
> +	if (ops->probe_finalize)
> +		ops->probe_finalize(dev);

This is an oddball one too, it is finishing setting up the default
domain for a device? Several drivers seem to recover the default
domain in their implementations..

Jason