Re: [PATCH v11 09/13] iommu/sva: Refactoring iommu_sva_bind/unbind_device()

[Jason Gunthorpe <jgg@nvidia.com>]

On Wed, Aug 17, 2022 at 09:20:20AM +0800, Lu Baolu wrote:
> +
> +/**
> + * iommu_sva_bind_device() - Bind a process address space to a device
> + * @dev: the device
> + * @mm: the mm to bind, caller must hold a reference to mm_users
> + *
> + * Create a bond between device and address space, allowing the device to access
> + * the mm using the returned PASID. If a bond already exists between @device and
> + * @mm, it is returned and an additional reference is taken. Caller must call
> + * iommu_sva_unbind_device() to release each reference.
> + *
> + * iommu_dev_enable_feature(dev, IOMMU_DEV_FEAT_SVA) must be called first, to
> + * initialize the required SVA features.
> + *
> + * On error, returns an ERR_PTR value.
> + */
> +struct iommu_sva *iommu_sva_bind_device(struct device *dev, struct mm_struct *mm)
> +{
> +	struct iommu_domain *domain;
> +	struct iommu_sva *bond;

This is called handle below, pick one name please

> +	ioasid_t max_pasids;
> +	int ret;
> +
> +	max_pasids = dev->iommu->max_pasids;
> +	if (!max_pasids)
> +		return ERR_PTR(-EOPNOTSUPP);
> +
> +	/* Allocate mm->pasid if necessary. */
> +	ret = iommu_sva_alloc_pasid(mm, 1, max_pasids - 1);
> +	if (ret)
> +		return ERR_PTR(ret);
> +
> +	bond = kzalloc(sizeof(*bond), GFP_KERNEL);
> +	if (!bond)
> +		return ERR_PTR(-ENOMEM);
> +
> +	mutex_lock(&iommu_sva_lock);
> +	/* Search for an existing domain. */
> +	domain = iommu_get_domain_for_dev_pasid(dev, mm->pasid);
> +	if (domain) {

This isn't safe, or sane. A driver could have attached something to
this PASID that is not a SVA domain and thus not protected by the
iommu_sva_lock.

At a minimum you should add a type match to
iommu_get_domain_for_dev_pasid(), eg to confirm it is a SVA domain and
do that check under the xa_lock of the pasid xarray.

And then the general idea is that SVA domain attach/detach must hold
this janky global lock. 

> +		refcount_inc(&domain->users);

This atomic is always processed under the iommu_sva_lock, so it
doesn't need to be an atomic anymore.

Otherwise this design looks OK to me too

Jason