RE: [PATCH v3 1/8] iommu: Add iopf domain attach/detach/replace interface

[Tian, Kevin]

> From: Lu Baolu <baolu.lu@xxxxxxxxxxxxxxx>
> Sent: Monday, January 22, 2024 3:39 PM
> 
> There is a slight difference between iopf domains and non-iopf domains.
> In the latter, references to domains occur between attach and detach;
> While in the former, due to the existence of asynchronous iopf handling
> paths, references to the domain may occur after detach, which leads to
> potential UAF issues.

Does UAF still exist if iommu driver follows the guidance you just added
to iopf_queue_remove_device()?

it clearly says that the driver needs to disable IOMMU PRI reception,
remove device from iopf queue and disable PRI on the device.

presumably those are all about what needs to be done in the detach
operation. Then once detach completes there should be no more
reference to the domain from the iopf path?

> 
> +struct iopf_attach_cookie {
> +	struct iommu_domain *domain;
> +	struct device *dev;
> +	unsigned int pasid;
> +	refcount_t users;
> +
> +	void *private;
> +	void (*release)(struct iopf_attach_cookie *cookie);
> +};

this cookie has nothing specific to iopf.

it may makes more sense to build a generic iommu_attach_device_cookie()
helper so the same object can be reused in future other usages too.

within iommu core it can check domain iopf handler and this generic cookie
to update iopf specific data e.g. the pasid_cookie xarray.